diff options
| author | Gadi Haber <gadi.haber@intel.com> | 2017-06-27 15:05:13 +0000 |
|---|---|---|
| committer | Gadi Haber <gadi.haber@intel.com> | 2017-06-27 15:05:13 +0000 |
| commit | 13759a7ed62a362bc3d7455da8b96279e545cdc6 (patch) | |
| tree | 2090b37ec2372cd62bbda9fca89ec5c157b44be7 /llvm/test | |
| parent | a179d25b99fec680d2430a07b6a35254c548e298 (diff) | |
| download | bcm5719-llvm-13759a7ed62a362bc3d7455da8b96279e545cdc6.tar.gz bcm5719-llvm-13759a7ed62a362bc3d7455da8b96279e545cdc6.zip | |
Updated and extended the information about each instruction in HSW and SNB to include the following data:
•static latency
•number of uOps from which the instructions consists
•all ports used by the instruction
Reviewers:
RKSimon
zvi
aymanmus
m_zuckerman
Differential Revision: https://reviews.llvm.org/D33897
llvm-svn: 306414
Diffstat (limited to 'llvm/test')
31 files changed, 4339 insertions, 3632 deletions
diff --git a/llvm/test/CodeGen/X86/avx-schedule.ll b/llvm/test/CodeGen/X86/avx-schedule.ll index 47e95fe31bd..6fe5f4c4da2 100644 --- a/llvm/test/CodeGen/X86/avx-schedule.ll +++ b/llvm/test/CodeGen/X86/avx-schedule.ll @@ -10,14 +10,14 @@ define <4 x double> @test_addpd(<4 x double> %a0, <4 x double> %a1, <4 x double> ; SANDY-LABEL: test_addpd: ; SANDY: # BB#0: ; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vaddpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vaddpd (%rdi), %ymm0, %ymm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_addpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vaddpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vaddpd (%rdi), %ymm0, %ymm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_addpd: ; BTVER2: # BB#0: @@ -40,14 +40,14 @@ define <8 x float> @test_addps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a ; SANDY-LABEL: test_addps: ; SANDY: # BB#0: ; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vaddps (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vaddps (%rdi), %ymm0, %ymm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_addps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vaddps (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vaddps (%rdi), %ymm0, %ymm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_addps: ; BTVER2: # BB#0: @@ -70,14 +70,14 @@ define <4 x double> @test_addsubpd(<4 x double> %a0, <4 x double> %a1, <4 x doub ; SANDY-LABEL: test_addsubpd: ; SANDY: # BB#0: ; SANDY-NEXT: vaddsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vaddsubpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vaddsubpd (%rdi), %ymm0, %ymm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_addsubpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vaddsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vaddsubpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vaddsubpd (%rdi), %ymm0, %ymm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_addsubpd: ; BTVER2: # BB#0: @@ -101,14 +101,14 @@ define <8 x float> @test_addsubps(<8 x float> %a0, <8 x float> %a1, <8 x float> ; SANDY-LABEL: test_addsubps: ; SANDY: # BB#0: ; SANDY-NEXT: vaddsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vaddsubps (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vaddsubps (%rdi), %ymm0, %ymm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_addsubps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vaddsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vaddsubps (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vaddsubps (%rdi), %ymm0, %ymm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_addsubps: ; BTVER2: # BB#0: @@ -131,17 +131,17 @@ declare <8 x float> @llvm.x86.avx.addsub.ps.256(<8 x float>, <8 x float>) nounwi define <4 x double> @test_andnotpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) { ; SANDY-LABEL: test_andnotpd: ; SANDY: # BB#0: -; SANDY-NEXT: vandnpd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SANDY-NEXT: vandnpd (%rdi), %ymm0, %ymm0 # sched: [5:0.50] +; SANDY-NEXT: vandnpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00] +; SANDY-NEXT: vandnpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00] ; SANDY-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_andnotpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vandnpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: vandnpd (%rdi), %ymm0, %ymm0 # sched: [5:1.00] +; HASWELL-NEXT: vandnpd (%rdi), %ymm0, %ymm0 # sched: [1:1.00] ; HASWELL-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_andnotpd: ; BTVER2: # BB#0: @@ -172,17 +172,17 @@ define <4 x double> @test_andnotpd(<4 x double> %a0, <4 x double> %a1, <4 x doub define <8 x float> @test_andnotps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) { ; SANDY-LABEL: test_andnotps: ; SANDY: # BB#0: -; SANDY-NEXT: vandnps %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SANDY-NEXT: vandnps (%rdi), %ymm0, %ymm0 # sched: [5:0.50] +; SANDY-NEXT: vandnps %ymm1, %ymm0, %ymm0 # sched: [1:1.00] +; SANDY-NEXT: vandnps (%rdi), %ymm0, %ymm0 # sched: [8:1.00] ; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_andnotps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vandnps %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: vandnps (%rdi), %ymm0, %ymm0 # sched: [5:1.00] +; HASWELL-NEXT: vandnps (%rdi), %ymm0, %ymm0 # sched: [1:1.00] ; HASWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_andnotps: ; BTVER2: # BB#0: @@ -216,14 +216,14 @@ define <4 x double> @test_andpd(<4 x double> %a0, <4 x double> %a1, <4 x double> ; SANDY-NEXT: vandpd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] ; SANDY-NEXT: vandpd (%rdi), %ymm0, %ymm0 # sched: [5:0.50] ; SANDY-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_andpd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vandpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: vandpd (%rdi), %ymm0, %ymm0 # sched: [5:1.00] +; HASWELL-NEXT: vandpd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] +; HASWELL-NEXT: vandpd (%rdi), %ymm0, %ymm0 # sched: [5:0.50] ; HASWELL-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_andpd: ; BTVER2: # BB#0: @@ -255,14 +255,14 @@ define <8 x float> @test_andps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a ; SANDY-NEXT: vandps %ymm1, %ymm0, %ymm0 # sched: [1:0.33] ; SANDY-NEXT: vandps (%rdi), %ymm0, %ymm0 # sched: [5:0.50] ; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_andps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vandps %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: vandps (%rdi), %ymm0, %ymm0 # sched: [5:1.00] +; HASWELL-NEXT: vandps %ymm1, %ymm0, %ymm0 # sched: [1:0.33] +; HASWELL-NEXT: vandps (%rdi), %ymm0, %ymm0 # sched: [5:0.50] ; HASWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_andps: ; BTVER2: # BB#0: @@ -291,17 +291,17 @@ define <8 x float> @test_andps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a define <4 x double> @test_blendpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) { ; SANDY-LABEL: test_blendpd: ; SANDY: # BB#0: -; SANDY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3] sched: [1:0.50] +; SANDY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3] sched: [1:1.00] ; SANDY-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],mem[1,2],ymm0[3] sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],mem[1,2],ymm0[3] sched: [8:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_blendpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3] sched: [1:0.33] ; HASWELL-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],mem[1,2],ymm0[3] sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],mem[1,2],ymm0[3] sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_blendpd: ; BTVER2: # BB#0: @@ -326,15 +326,15 @@ define <4 x double> @test_blendpd(<4 x double> %a0, <4 x double> %a1, <4 x doubl define <8 x float> @test_blendps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) { ; SANDY-LABEL: test_blendps: ; SANDY: # BB#0: -; SANDY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3,4,5,6,7] sched: [1:0.50] -; SANDY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3],mem[4,5,6],ymm0[7] sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3,4,5,6,7] sched: [1:1.00] +; SANDY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3],mem[4,5,6],ymm0[7] sched: [8:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_blendps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3,4,5,6,7] sched: [1:0.33] -; HASWELL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3],mem[4,5,6],ymm0[7] sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3],mem[4,5,6],ymm0[7] sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_blendps: ; BTVER2: # BB#0: @@ -356,15 +356,15 @@ define <8 x float> @test_blendps(<8 x float> %a0, <8 x float> %a1, <8 x float> * define <4 x double> @test_blendvpd(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, <4 x double> *%a3) { ; SANDY-LABEL: test_blendvpd: ; SANDY: # BB#0: -; SANDY-NEXT: vblendvpd %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:1.00] -; SANDY-NEXT: vblendvpd %ymm2, (%rdi), %ymm0, %ymm0 # sched: [6:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vblendvpd %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:2.00] +; SANDY-NEXT: vblendvpd %ymm2, (%rdi), %ymm0, %ymm0 # sched: [9:2.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_blendvpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vblendvpd %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:2.00] -; HASWELL-NEXT: vblendvpd %ymm2, (%rdi), %ymm0, %ymm0 # sched: [6:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vblendvpd %ymm2, (%rdi), %ymm0, %ymm0 # sched: [2:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_blendvpd: ; BTVER2: # BB#0: @@ -387,15 +387,15 @@ declare <4 x double> @llvm.x86.avx.blendv.pd.256(<4 x double>, <4 x double>, <4 define <8 x float> @test_blendvps(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, <8 x float> *%a3) { ; SANDY-LABEL: test_blendvps: ; SANDY: # BB#0: -; SANDY-NEXT: vblendvps %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:1.00] -; SANDY-NEXT: vblendvps %ymm2, (%rdi), %ymm0, %ymm0 # sched: [6:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vblendvps %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:2.00] +; SANDY-NEXT: vblendvps %ymm2, (%rdi), %ymm0, %ymm0 # sched: [9:2.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_blendvps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vblendvps %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:2.00] -; HASWELL-NEXT: vblendvps %ymm2, (%rdi), %ymm0, %ymm0 # sched: [6:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vblendvps %ymm2, (%rdi), %ymm0, %ymm0 # sched: [2:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_blendvps: ; BTVER2: # BB#0: @@ -418,13 +418,13 @@ declare <8 x float> @llvm.x86.avx.blendv.ps.256(<8 x float>, <8 x float>, <8 x f define <8 x float> @test_broadcastf128(<4 x float> *%a0) { ; SANDY-LABEL: test_broadcastf128: ; SANDY: # BB#0: -; SANDY-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1] sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1] sched: [3:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_broadcastf128: ; HASWELL: # BB#0: -; HASWELL-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1] sched: [4:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1] sched: [?:5.000000e-01] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_broadcastf128: ; BTVER2: # BB#0: @@ -443,13 +443,13 @@ define <8 x float> @test_broadcastf128(<4 x float> *%a0) { define <4 x double> @test_broadcastsd_ymm(double *%a0) { ; SANDY-LABEL: test_broadcastsd_ymm: ; SANDY: # BB#0: -; SANDY-NEXT: vbroadcastsd (%rdi), %ymm0 # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vbroadcastsd (%rdi), %ymm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_broadcastsd_ymm: ; HASWELL: # BB#0: -; HASWELL-NEXT: vbroadcastsd (%rdi), %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vbroadcastsd (%rdi), %ymm0 # sched: [?:5.000000e-01] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_broadcastsd_ymm: ; BTVER2: # BB#0: @@ -469,13 +469,13 @@ define <4 x double> @test_broadcastsd_ymm(double *%a0) { define <4 x float> @test_broadcastss(float *%a0) { ; SANDY-LABEL: test_broadcastss: ; SANDY: # BB#0: -; SANDY-NEXT: vbroadcastss (%rdi), %xmm0 # sched: [4:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vbroadcastss (%rdi), %xmm0 # sched: [6:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_broadcastss: ; HASWELL: # BB#0: -; HASWELL-NEXT: vbroadcastss (%rdi), %xmm0 # sched: [4:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vbroadcastss (%rdi), %xmm0 # sched: [?:5.000000e-01] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_broadcastss: ; BTVER2: # BB#0: @@ -496,12 +496,12 @@ define <8 x float> @test_broadcastss_ymm(float *%a0) { ; SANDY-LABEL: test_broadcastss_ymm: ; SANDY: # BB#0: ; SANDY-NEXT: vbroadcastss (%rdi), %ymm0 # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_broadcastss_ymm: ; HASWELL: # BB#0: ; HASWELL-NEXT: vbroadcastss (%rdi), %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_broadcastss_ymm: ; BTVER2: # BB#0: @@ -521,17 +521,17 @@ define <8 x float> @test_broadcastss_ymm(float *%a0) { define <4 x double> @test_cmppd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) { ; SANDY-LABEL: test_cmppd: ; SANDY: # BB#0: -; SANDY-NEXT: vcmpeqpd %ymm1, %ymm0, %ymm1 # sched: [3:1.00] -; SANDY-NEXT: vcmpeqpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: vorpd %ymm0, %ymm1, %ymm0 # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vcmpeqpd (%rdi), %ymm0, %ymm2 # sched: [9:1.00] +; SANDY-NEXT: vcmpeqpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] +; SANDY-NEXT: vorpd %ymm2, %ymm0, %ymm0 # sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cmppd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcmpeqpd %ymm1, %ymm0, %ymm1 # sched: [3:1.00] -; HASWELL-NEXT: vcmpeqpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00] +; HASWELL-NEXT: vcmpeqpd (%rdi), %ymm0, %ymm0 # sched: [3:1.00] ; HASWELL-NEXT: vorpd %ymm0, %ymm1, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cmppd: ; BTVER2: # BB#0: @@ -559,17 +559,17 @@ define <4 x double> @test_cmppd(<4 x double> %a0, <4 x double> %a1, <4 x double> define <8 x float> @test_cmpps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) { ; SANDY-LABEL: test_cmpps: ; SANDY: # BB#0: -; SANDY-NEXT: vcmpeqps %ymm1, %ymm0, %ymm1 # sched: [3:1.00] -; SANDY-NEXT: vcmpeqps (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: vorps %ymm0, %ymm1, %ymm0 # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vcmpeqps (%rdi), %ymm0, %ymm2 # sched: [9:1.00] +; SANDY-NEXT: vcmpeqps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] +; SANDY-NEXT: vorps %ymm2, %ymm0, %ymm0 # sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cmpps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcmpeqps %ymm1, %ymm0, %ymm1 # sched: [3:1.00] -; HASWELL-NEXT: vcmpeqps (%rdi), %ymm0, %ymm0 # sched: [7:1.00] +; HASWELL-NEXT: vcmpeqps (%rdi), %ymm0, %ymm0 # sched: [3:1.00] ; HASWELL-NEXT: vorps %ymm0, %ymm1, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cmpps: ; BTVER2: # BB#0: @@ -598,16 +598,16 @@ define <4 x double> @test_cvtdq2pd(<4 x i32> %a0, <4 x i32> *%a1) { ; SANDY-LABEL: test_cvtdq2pd: ; SANDY: # BB#0: ; SANDY-NEXT: vcvtdq2pd %xmm0, %ymm0 # sched: [4:1.00] -; SANDY-NEXT: vcvtdq2pd (%rdi), %ymm1 # sched: [8:1.00] +; SANDY-NEXT: vcvtdq2pd (%rdi), %ymm1 # sched: [10:1.00] ; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtdq2pd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvtdq2pd %xmm0, %ymm0 # sched: [6:1.00] -; HASWELL-NEXT: vcvtdq2pd (%rdi), %ymm1 # sched: [8:1.00] +; HASWELL-NEXT: vcvtdq2pd (%rdi), %ymm1 # sched: [6:1.00] ; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtdq2pd: ; BTVER2: # BB#0: @@ -632,19 +632,19 @@ define <4 x double> @test_cvtdq2pd(<4 x i32> %a0, <4 x i32> *%a1) { define <8 x float> @test_cvtdq2ps(<8 x i32> %a0, <8 x i32> *%a1) { ; SANDY-LABEL: test_cvtdq2ps: ; SANDY: # BB#0: -; SANDY-NEXT: vcvtdq2ps %ymm0, %ymm0 # sched: [4:1.00] -; SANDY-NEXT: vmovaps (%rdi), %xmm1 # sched: [4:0.50] -; SANDY-NEXT: vinsertf128 $1, 16(%rdi), %ymm1, %ymm1 # sched: [5:1.00] -; SANDY-NEXT: vcvtdq2ps %ymm1, %ymm1 # sched: [4:1.00] +; SANDY-NEXT: vcvtdq2ps %ymm0, %ymm0 # sched: [3:1.00] +; SANDY-NEXT: vmovaps (%rdi), %xmm1 # sched: [6:0.50] +; SANDY-NEXT: vinsertf128 $1, 16(%rdi), %ymm1, %ymm1 # sched: [7:1.00] +; SANDY-NEXT: vcvtdq2ps %ymm1, %ymm1 # sched: [3:1.00] ; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtdq2ps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vcvtdq2ps %ymm0, %ymm0 # sched: [4:1.00] -; HASWELL-NEXT: vcvtdq2ps (%rdi), %ymm1 # sched: [8:1.00] +; HASWELL-NEXT: vcvtdq2ps %ymm0, %ymm0 # sched: [3:1.00] +; HASWELL-NEXT: vcvtdq2ps (%rdi), %ymm1 # sched: [3:1.00] ; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtdq2ps: ; BTVER2: # BB#0: @@ -669,17 +669,17 @@ define <8 x float> @test_cvtdq2ps(<8 x i32> %a0, <8 x i32> *%a1) { define <8 x i32> @test_cvtpd2dq(<4 x double> %a0, <4 x double> *%a1) { ; SANDY-LABEL: test_cvtpd2dq: ; SANDY: # BB#0: -; SANDY-NEXT: vcvttpd2dq %ymm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vcvttpd2dqy (%rdi), %xmm1 # sched: [7:1.00] +; SANDY-NEXT: vcvttpd2dq %ymm0, %xmm0 # sched: [4:1.00] +; SANDY-NEXT: vcvttpd2dqy (%rdi), %xmm1 # sched: [11:1.00] ; SANDY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtpd2dq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvttpd2dq %ymm0, %xmm0 # sched: [6:1.00] -; HASWELL-NEXT: vcvttpd2dqy (%rdi), %xmm1 # sched: [10:1.00] +; HASWELL-NEXT: vcvttpd2dqy (%rdi), %xmm1 # sched: [7:1.00] ; HASWELL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtpd2dq: ; BTVER2: # BB#0: @@ -704,17 +704,17 @@ define <8 x i32> @test_cvtpd2dq(<4 x double> %a0, <4 x double> *%a1) { define <8 x float> @test_cvtpd2ps(<4 x double> %a0, <4 x double> *%a1) { ; SANDY-LABEL: test_cvtpd2ps: ; SANDY: # BB#0: -; SANDY-NEXT: vcvtpd2ps %ymm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vcvtpd2psy (%rdi), %xmm1 # sched: [7:1.00] +; SANDY-NEXT: vcvtpd2ps %ymm0, %xmm0 # sched: [4:1.00] +; SANDY-NEXT: vcvtpd2psy (%rdi), %xmm1 # sched: [11:1.00] ; SANDY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtpd2ps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vcvtpd2ps %ymm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: vcvtpd2psy (%rdi), %xmm1 # sched: [9:1.00] +; HASWELL-NEXT: vcvtpd2ps %ymm0, %xmm0 # sched: [6:1.00] +; HASWELL-NEXT: vcvtpd2psy (%rdi), %xmm1 # sched: [7:1.00] ; HASWELL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtpd2ps: ; BTVER2: # BB#0: @@ -741,15 +741,15 @@ define <8 x i32> @test_cvtps2dq(<8 x float> %a0, <8 x float> *%a1) { ; SANDY: # BB#0: ; SANDY-NEXT: vcvttps2dq %ymm0, %ymm0 # sched: [3:1.00] ; SANDY-NEXT: vcvttps2dq (%rdi), %ymm1 # sched: [7:1.00] -; SANDY-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtps2dq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvttps2dq %ymm0, %ymm0 # sched: [3:1.00] ; HASWELL-NEXT: vcvttps2dq (%rdi), %ymm1 # sched: [7:1.00] ; HASWELL-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtps2dq: ; BTVER2: # BB#0: @@ -774,15 +774,15 @@ define <8 x i32> @test_cvtps2dq(<8 x float> %a0, <8 x float> *%a1) { define <4 x double> @test_divpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) { ; SANDY-LABEL: test_divpd: ; SANDY: # BB#0: -; SANDY-NEXT: vdivpd %ymm1, %ymm0, %ymm0 # sched: [12:1.00] -; SANDY-NEXT: vdivpd (%rdi), %ymm0, %ymm0 # sched: [16:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vdivpd %ymm1, %ymm0, %ymm0 # sched: [45:3.00] +; SANDY-NEXT: vdivpd (%rdi), %ymm0, %ymm0 # sched: [52:3.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_divpd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vdivpd %ymm1, %ymm0, %ymm0 # sched: [27:2.00] -; HASWELL-NEXT: vdivpd (%rdi), %ymm0, %ymm0 # sched: [31:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vdivpd %ymm1, %ymm0, %ymm0 # sched: [35:2.00] +; HASWELL-NEXT: vdivpd (%rdi), %ymm0, %ymm0 # sched: [35:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_divpd: ; BTVER2: # BB#0: @@ -804,15 +804,15 @@ define <4 x double> @test_divpd(<4 x double> %a0, <4 x double> %a1, <4 x double> define <8 x float> @test_divps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) { ; SANDY-LABEL: test_divps: ; SANDY: # BB#0: -; SANDY-NEXT: vdivps %ymm1, %ymm0, %ymm0 # sched: [12:1.00] -; SANDY-NEXT: vdivps (%rdi), %ymm0, %ymm0 # sched: [16:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vdivps %ymm1, %ymm0, %ymm0 # sched: [29:3.00] +; SANDY-NEXT: vdivps (%rdi), %ymm0, %ymm0 # sched: [36:3.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_divps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vdivps %ymm1, %ymm0, %ymm0 # sched: [19:2.00] -; HASWELL-NEXT: vdivps (%rdi), %ymm0, %ymm0 # sched: [23:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vdivps %ymm1, %ymm0, %ymm0 # sched: [21:2.00] +; HASWELL-NEXT: vdivps (%rdi), %ymm0, %ymm0 # sched: [21:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_divps: ; BTVER2: # BB#0: @@ -834,15 +834,15 @@ define <8 x float> @test_divps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a define <8 x float> @test_dpps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) { ; SANDY-LABEL: test_dpps: ; SANDY: # BB#0: -; SANDY-NEXT: vdpps $7, %ymm1, %ymm0, %ymm0 # sched: [3:1.00] +; SANDY-NEXT: vdpps $7, %ymm1, %ymm0, %ymm0 # sched: [12:2.00] ; SANDY-NEXT: vdpps $7, (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_dpps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vdpps $7, %ymm1, %ymm0, %ymm0 # sched: [14:2.00] -; HASWELL-NEXT: vdpps $7, (%rdi), %ymm0, %ymm0 # sched: [18:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vdpps $7, (%rdi), %ymm0, %ymm0 # sched: [14:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_dpps: ; BTVER2: # BB#0: @@ -866,16 +866,16 @@ define <4 x float> @test_extractf128(<8 x float> %a0, <8 x float> %a1, <4 x floa ; SANDY-LABEL: test_extractf128: ; SANDY: # BB#0: ; SANDY-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [1:1.00] -; SANDY-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [1:1.00] +; SANDY-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [5:1.00] ; SANDY-NEXT: vzeroupper # sched: [?:0.000000e+00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_extractf128: ; HASWELL: # BB#0: ; HASWELL-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [4:1.00] -; HASWELL-NEXT: vzeroupper # sched: [1:0.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: vzeroupper # sched: [4:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_extractf128: ; BTVER2: # BB#0: @@ -900,13 +900,13 @@ define <4 x double> @test_haddpd(<4 x double> %a0, <4 x double> %a1, <4 x double ; SANDY: # BB#0: ; SANDY-NEXT: vhaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] ; SANDY-NEXT: vhaddpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_haddpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vhaddpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00] ; HASWELL-NEXT: vhaddpd (%rdi), %ymm0, %ymm0 # sched: [9:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_haddpd: ; BTVER2: # BB#0: @@ -929,15 +929,15 @@ declare <4 x double> @llvm.x86.avx.hadd.pd.256(<4 x double>, <4 x double>) nounw define <8 x float> @test_haddps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) { ; SANDY-LABEL: test_haddps: ; SANDY: # BB#0: -; SANDY-NEXT: vhaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vhaddps (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vhaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00] +; SANDY-NEXT: vhaddps (%rdi), %ymm0, %ymm0 # sched: [12:2.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_haddps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vhaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00] -; HASWELL-NEXT: vhaddps (%rdi), %ymm0, %ymm0 # sched: [9:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vhaddps (%rdi), %ymm0, %ymm0 # sched: [5:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_haddps: ; BTVER2: # BB#0: @@ -960,15 +960,15 @@ declare <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float>, <8 x float>) nounwind define <4 x double> @test_hsubpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) { ; SANDY-LABEL: test_hsubpd: ; SANDY: # BB#0: -; SANDY-NEXT: vhsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vhsubpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vhsubpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00] +; SANDY-NEXT: vhsubpd (%rdi), %ymm0, %ymm0 # sched: [12:2.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_hsubpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vhsubpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00] -; HASWELL-NEXT: vhsubpd (%rdi), %ymm0, %ymm0 # sched: [9:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vhsubpd (%rdi), %ymm0, %ymm0 # sched: [5:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_hsubpd: ; BTVER2: # BB#0: @@ -991,15 +991,15 @@ declare <4 x double> @llvm.x86.avx.hsub.pd.256(<4 x double>, <4 x double>) nounw define <8 x float> @test_hsubps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) { ; SANDY-LABEL: test_hsubps: ; SANDY: # BB#0: -; SANDY-NEXT: vhsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vhsubps (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vhsubps %ymm1, %ymm0, %ymm0 # sched: [5:2.00] +; SANDY-NEXT: vhsubps (%rdi), %ymm0, %ymm0 # sched: [12:2.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_hsubps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vhsubps %ymm1, %ymm0, %ymm0 # sched: [5:2.00] -; HASWELL-NEXT: vhsubps (%rdi), %ymm0, %ymm0 # sched: [9:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vhsubps (%rdi), %ymm0, %ymm0 # sched: [5:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_hsubps: ; BTVER2: # BB#0: @@ -1023,16 +1023,16 @@ define <8 x float> @test_insertf128(<8 x float> %a0, <4 x float> %a1, <4 x float ; SANDY-LABEL: test_insertf128: ; SANDY: # BB#0: ; SANDY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 # sched: [1:1.00] -; SANDY-NEXT: vinsertf128 $1, (%rdi), %ymm0, %ymm0 # sched: [5:1.00] +; SANDY-NEXT: vinsertf128 $1, (%rdi), %ymm0, %ymm0 # sched: [7:1.00] ; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_insertf128: ; HASWELL: # BB#0: ; HASWELL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 # sched: [3:1.00] -; HASWELL-NEXT: vinsertf128 $1, (%rdi), %ymm0, %ymm0 # sched: [3:1.00] +; HASWELL-NEXT: vinsertf128 $1, (%rdi), %ymm0, %ymm0 # sched: [1:0.50] ; HASWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_insertf128: ; BTVER2: # BB#0: @@ -1059,13 +1059,13 @@ define <8 x float> @test_insertf128(<8 x float> %a0, <4 x float> %a1, <4 x float define <32 x i8> @test_lddqu(i8* %a0) { ; SANDY-LABEL: test_lddqu: ; SANDY: # BB#0: -; SANDY-NEXT: vlddqu (%rdi), %ymm0 # sched: [4:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vlddqu (%rdi), %ymm0 # sched: [6:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_lddqu: ; HASWELL: # BB#0: -; HASWELL-NEXT: vlddqu (%rdi), %ymm0 # sched: [4:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vlddqu (%rdi), %ymm0 # sched: [?:5.000000e-01] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_lddqu: ; BTVER2: # BB#0: @@ -1084,17 +1084,17 @@ declare <32 x i8> @llvm.x86.avx.ldu.dq.256(i8*) nounwind readonly define <2 x double> @test_maskmovpd(i8* %a0, <2 x i64> %a1, <2 x double> %a2) { ; SANDY-LABEL: test_maskmovpd: ; SANDY: # BB#0: -; SANDY-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2 # sched: [?:0.000000e+00] -; SANDY-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi) # sched: [?:0.000000e+00] +; SANDY-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2 # sched: [8:2.00] +; SANDY-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi) # sched: [5:1.00] ; SANDY-NEXT: vmovapd %xmm2, %xmm0 # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_maskmovpd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2 # sched: [4:2.00] -; HASWELL-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi) # sched: [13:1.00] +; HASWELL-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2 # sched: [2:2.00] +; HASWELL-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi) # sched: [4:1.00] ; HASWELL-NEXT: vmovapd %xmm2, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_maskmovpd: ; BTVER2: # BB#0: @@ -1119,29 +1119,29 @@ declare void @llvm.x86.avx.maskstore.pd(i8*, <2 x i64>, <2 x double>) nounwind define <4 x double> @test_maskmovpd_ymm(i8* %a0, <4 x i64> %a1, <4 x double> %a2) { ; SANDY-LABEL: test_maskmovpd_ymm: ; SANDY: # BB#0: -; SANDY-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [?:0.000000e+00] -; SANDY-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) # sched: [?:0.000000e+00] +; SANDY-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [5:1.00] +; SANDY-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) ; SANDY-NEXT: vmovapd %ymm2, %ymm0 # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_maskmovpd_ymm: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [4:2.00] -; HASWELL-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) # sched: [14:1.00] +; HASWELL-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [4:1.00] +; HASWELL-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) ; HASWELL-NEXT: vmovapd %ymm2, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_maskmovpd_ymm: ; BTVER2: # BB#0: ; BTVER2-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [?:0.000000e+00] -; BTVER2-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) # sched: [?:0.000000e+00] +; BTVER2-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) ; BTVER2-NEXT: vmovapd %ymm2, %ymm0 # sched: [1:0.50] ; BTVER2-NEXT: retq # sched: [4:1.00] ; ; ZNVER1-LABEL: test_maskmovpd_ymm: ; ZNVER1: # BB#0: ; ZNVER1-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [?:0.000000e+00] -; ZNVER1-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) # sched: [?:0.000000e+00] +; ZNVER1-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) ; ZNVER1-NEXT: vmovapd %ymm2, %ymm0 # sched: [1:0.50] ; ZNVER1-NEXT: retq # sched: [4:1.00] %1 = call <4 x double> @llvm.x86.avx.maskload.pd.256(i8* %a0, <4 x i64> %a1) @@ -1154,17 +1154,17 @@ declare void @llvm.x86.avx.maskstore.pd.256(i8*, <4 x i64>, <4 x double>) nounwi define <4 x float> @test_maskmovps(i8* %a0, <4 x i32> %a1, <4 x float> %a2) { ; SANDY-LABEL: test_maskmovps: ; SANDY: # BB#0: -; SANDY-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2 # sched: [?:0.000000e+00] -; SANDY-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi) # sched: [?:0.000000e+00] +; SANDY-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2 # sched: [8:2.00] +; SANDY-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi) # sched: [5:1.00] ; SANDY-NEXT: vmovaps %xmm2, %xmm0 # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_maskmovps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2 # sched: [4:2.00] -; HASWELL-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi) # sched: [13:1.00] +; HASWELL-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2 # sched: [2:2.00] +; HASWELL-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi) # sched: [4:1.00] ; HASWELL-NEXT: vmovaps %xmm2, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_maskmovps: ; BTVER2: # BB#0: @@ -1189,29 +1189,29 @@ declare void @llvm.x86.avx.maskstore.ps(i8*, <4 x i32>, <4 x float>) nounwind define <8 x float> @test_maskmovps_ymm(i8* %a0, <8 x i32> %a1, <8 x float> %a2) { ; SANDY-LABEL: test_maskmovps_ymm: ; SANDY: # BB#0: -; SANDY-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [?:0.000000e+00] -; SANDY-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) # sched: [?:0.000000e+00] +; SANDY-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [1:0.50] +; SANDY-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) ; SANDY-NEXT: vmovaps %ymm2, %ymm0 # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_maskmovps_ymm: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [4:2.00] -; HASWELL-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) # sched: [14:1.00] +; HASWELL-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [1:0.50] +; HASWELL-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) ; HASWELL-NEXT: vmovaps %ymm2, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_maskmovps_ymm: ; BTVER2: # BB#0: ; BTVER2-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [?:0.000000e+00] -; BTVER2-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) # sched: [?:0.000000e+00] +; BTVER2-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) ; BTVER2-NEXT: vmovaps %ymm2, %ymm0 # sched: [1:0.50] ; BTVER2-NEXT: retq # sched: [4:1.00] ; ; ZNVER1-LABEL: test_maskmovps_ymm: ; ZNVER1: # BB#0: ; ZNVER1-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [?:0.000000e+00] -; ZNVER1-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) # sched: [?:0.000000e+00] +; ZNVER1-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) ; ZNVER1-NEXT: vmovaps %ymm2, %ymm0 # sched: [1:0.50] ; ZNVER1-NEXT: retq # sched: [4:1.00] %1 = call <8 x float> @llvm.x86.avx.maskload.ps.256(i8* %a0, <8 x i32> %a1) @@ -1225,14 +1225,14 @@ define <4 x double> @test_maxpd(<4 x double> %a0, <4 x double> %a1, <4 x double> ; SANDY-LABEL: test_maxpd: ; SANDY: # BB#0: ; SANDY-NEXT: vmaxpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vmaxpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmaxpd (%rdi), %ymm0, %ymm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_maxpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmaxpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vmaxpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmaxpd (%rdi), %ymm0, %ymm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_maxpd: ; BTVER2: # BB#0: @@ -1256,14 +1256,14 @@ define <8 x float> @test_maxps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a ; SANDY-LABEL: test_maxps: ; SANDY: # BB#0: ; SANDY-NEXT: vmaxps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vmaxps (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmaxps (%rdi), %ymm0, %ymm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_maxps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmaxps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vmaxps (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmaxps (%rdi), %ymm0, %ymm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_maxps: ; BTVER2: # BB#0: @@ -1288,13 +1288,13 @@ define <4 x double> @test_minpd(<4 x double> %a0, <4 x double> %a1, <4 x double> ; SANDY: # BB#0: ; SANDY-NEXT: vminpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] ; SANDY-NEXT: vminpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_minpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vminpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] ; HASWELL-NEXT: vminpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_minpd: ; BTVER2: # BB#0: @@ -1319,13 +1319,13 @@ define <8 x float> @test_minps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a ; SANDY: # BB#0: ; SANDY-NEXT: vminps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] ; SANDY-NEXT: vminps (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_minps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vminps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] ; HASWELL-NEXT: vminps (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_minps: ; BTVER2: # BB#0: @@ -1348,17 +1348,17 @@ declare <8 x float> @llvm.x86.avx.min.ps.256(<8 x float>, <8 x float>) nounwind define <4 x double> @test_movapd(<4 x double> *%a0, <4 x double> *%a1) { ; SANDY-LABEL: test_movapd: ; SANDY: # BB#0: -; SANDY-NEXT: vmovapd (%rdi), %ymm0 # sched: [4:0.50] +; SANDY-NEXT: vmovapd (%rdi), %ymm0 # sched: [7:0.50] ; SANDY-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vmovapd %ymm0, (%rsi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovapd %ymm0, (%rsi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movapd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovapd (%rdi), %ymm0 # sched: [4:0.50] +; HASWELL-NEXT: vmovapd (%rdi), %ymm0 # sched: [?:5.000000e-01] ; HASWELL-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vmovapd %ymm0, (%rsi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovapd %ymm0, (%rsi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movapd: ; BTVER2: # BB#0: @@ -1382,17 +1382,17 @@ define <4 x double> @test_movapd(<4 x double> *%a0, <4 x double> *%a1) { define <8 x float> @test_movaps(<8 x float> *%a0, <8 x float> *%a1) { ; SANDY-LABEL: test_movaps: ; SANDY: # BB#0: -; SANDY-NEXT: vmovaps (%rdi), %ymm0 # sched: [4:0.50] +; SANDY-NEXT: vmovaps (%rdi), %ymm0 # sched: [7:0.50] ; SANDY-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vmovaps %ymm0, (%rsi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovaps %ymm0, (%rsi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movaps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovaps (%rdi), %ymm0 # sched: [4:0.50] +; HASWELL-NEXT: vmovaps (%rdi), %ymm0 # sched: [?:5.000000e-01] ; HASWELL-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vmovaps %ymm0, (%rsi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovaps %ymm0, (%rsi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movaps: ; BTVER2: # BB#0: @@ -1417,16 +1417,16 @@ define <4 x double> @test_movddup(<4 x double> %a0, <4 x double> *%a1) { ; SANDY-LABEL: test_movddup: ; SANDY: # BB#0: ; SANDY-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2] sched: [1:1.00] -; SANDY-NEXT: vmovddup {{.*#+}} ymm1 = mem[0,0,2,2] sched: [4:0.50] +; SANDY-NEXT: vmovddup {{.*#+}} ymm1 = mem[0,0,2,2] sched: [7:0.50] ; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movddup: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2] sched: [1:1.00] -; HASWELL-NEXT: vmovddup {{.*#+}} ymm1 = mem[0,0,2,2] sched: [4:0.50] +; HASWELL-NEXT: vmovddup {{.*#+}} ymm1 = mem[0,0,2,2] sched: [?:5.000000e-01] ; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movddup: ; BTVER2: # BB#0: @@ -1451,15 +1451,15 @@ define <4 x double> @test_movddup(<4 x double> %a0, <4 x double> *%a1) { define i32 @test_movmskpd(<4 x double> %a0) { ; SANDY-LABEL: test_movmskpd: ; SANDY: # BB#0: -; SANDY-NEXT: vmovmskpd %ymm0, %eax # sched: [1:0.33] +; SANDY-NEXT: vmovmskpd %ymm0, %eax # sched: [2:1.00] ; SANDY-NEXT: vzeroupper # sched: [?:0.000000e+00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movmskpd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovmskpd %ymm0, %eax # sched: [2:1.00] -; HASWELL-NEXT: vzeroupper # sched: [1:0.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovmskpd %ymm0, %eax # sched: [3:1.00] +; HASWELL-NEXT: vzeroupper # sched: [4:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movmskpd: ; BTVER2: # BB#0: @@ -1479,15 +1479,15 @@ declare i32 @llvm.x86.avx.movmsk.pd.256(<4 x double>) nounwind readnone define i32 @test_movmskps(<8 x float> %a0) { ; SANDY-LABEL: test_movmskps: ; SANDY: # BB#0: -; SANDY-NEXT: vmovmskps %ymm0, %eax # sched: [1:0.33] +; SANDY-NEXT: vmovmskps %ymm0, %eax # sched: [3:1.00] ; SANDY-NEXT: vzeroupper # sched: [?:0.000000e+00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movmskps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovmskps %ymm0, %eax # sched: [2:1.00] -; HASWELL-NEXT: vzeroupper # sched: [1:0.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovmskps %ymm0, %eax # sched: [5:1.00] +; HASWELL-NEXT: vzeroupper # sched: [4:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movmskps: ; BTVER2: # BB#0: @@ -1508,14 +1508,14 @@ define <4 x double> @test_movntpd(<4 x double> %a0, <4 x double> *%a1) { ; SANDY-LABEL: test_movntpd: ; SANDY: # BB#0: ; SANDY-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vmovntpd %ymm0, (%rdi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovntpd %ymm0, (%rdi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movntpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vmovntpd %ymm0, (%rdi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovntpd %ymm0, (%rdi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movntpd: ; BTVER2: # BB#0: @@ -1537,14 +1537,14 @@ define <8 x float> @test_movntps(<8 x float> %a0, <8 x float> *%a1) { ; SANDY-LABEL: test_movntps: ; SANDY: # BB#0: ; SANDY-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vmovntps %ymm0, (%rdi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovntps %ymm0, (%rdi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movntps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vmovntps %ymm0, (%rdi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovntps %ymm0, (%rdi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movntps: ; BTVER2: # BB#0: @@ -1566,16 +1566,16 @@ define <8 x float> @test_movshdup(<8 x float> %a0, <8 x float> *%a1) { ; SANDY-LABEL: test_movshdup: ; SANDY: # BB#0: ; SANDY-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7] sched: [1:1.00] -; SANDY-NEXT: vmovshdup {{.*#+}} ymm1 = mem[1,1,3,3,5,5,7,7] sched: [4:0.50] +; SANDY-NEXT: vmovshdup {{.*#+}} ymm1 = mem[1,1,3,3,5,5,7,7] sched: [7:0.50] ; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movshdup: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7] sched: [1:1.00] -; HASWELL-NEXT: vmovshdup {{.*#+}} ymm1 = mem[1,1,3,3,5,5,7,7] sched: [4:0.50] +; HASWELL-NEXT: vmovshdup {{.*#+}} ymm1 = mem[1,1,3,3,5,5,7,7] sched: [?:5.000000e-01] ; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movshdup: ; BTVER2: # BB#0: @@ -1601,16 +1601,16 @@ define <8 x float> @test_movsldup(<8 x float> %a0, <8 x float> *%a1) { ; SANDY-LABEL: test_movsldup: ; SANDY: # BB#0: ; SANDY-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6] sched: [1:1.00] -; SANDY-NEXT: vmovsldup {{.*#+}} ymm1 = mem[0,0,2,2,4,4,6,6] sched: [4:0.50] +; SANDY-NEXT: vmovsldup {{.*#+}} ymm1 = mem[0,0,2,2,4,4,6,6] sched: [7:0.50] ; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movsldup: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6] sched: [1:1.00] -; HASWELL-NEXT: vmovsldup {{.*#+}} ymm1 = mem[0,0,2,2,4,4,6,6] sched: [4:0.50] +; HASWELL-NEXT: vmovsldup {{.*#+}} ymm1 = mem[0,0,2,2,4,4,6,6] sched: [?:5.000000e-01] ; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movsldup: ; BTVER2: # BB#0: @@ -1635,19 +1635,19 @@ define <8 x float> @test_movsldup(<8 x float> %a0, <8 x float> *%a1) { define <4 x double> @test_movupd(<4 x double> *%a0, <4 x double> *%a1) { ; SANDY-LABEL: test_movupd: ; SANDY: # BB#0: -; SANDY-NEXT: vmovups (%rdi), %xmm0 # sched: [4:0.50] -; SANDY-NEXT: vinsertf128 $1, 16(%rdi), %ymm0, %ymm0 # sched: [5:1.00] +; SANDY-NEXT: vmovups (%rdi), %xmm0 # sched: [6:0.50] +; SANDY-NEXT: vinsertf128 $1, 16(%rdi), %ymm0, %ymm0 # sched: [7:1.00] ; SANDY-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vextractf128 $1, %ymm0, 16(%rsi) # sched: [1:1.00] -; SANDY-NEXT: vmovupd %xmm0, (%rsi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vextractf128 $1, %ymm0, 16(%rsi) # sched: [5:1.00] +; SANDY-NEXT: vmovupd %xmm0, (%rsi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movupd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovupd (%rdi), %ymm0 # sched: [4:0.50] +; HASWELL-NEXT: vmovupd (%rdi), %ymm0 # sched: [?:5.000000e-01] ; HASWELL-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vmovupd %ymm0, (%rsi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovupd %ymm0, (%rsi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movupd: ; BTVER2: # BB#0: @@ -1671,19 +1671,19 @@ define <4 x double> @test_movupd(<4 x double> *%a0, <4 x double> *%a1) { define <8 x float> @test_movups(<8 x float> *%a0, <8 x float> *%a1) { ; SANDY-LABEL: test_movups: ; SANDY: # BB#0: -; SANDY-NEXT: vmovups (%rdi), %xmm0 # sched: [4:0.50] -; SANDY-NEXT: vinsertf128 $1, 16(%rdi), %ymm0, %ymm0 # sched: [5:1.00] +; SANDY-NEXT: vmovups (%rdi), %xmm0 # sched: [6:0.50] +; SANDY-NEXT: vinsertf128 $1, 16(%rdi), %ymm0, %ymm0 # sched: [7:1.00] ; SANDY-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vextractf128 $1, %ymm0, 16(%rsi) # sched: [1:1.00] -; SANDY-NEXT: vmovups %xmm0, (%rsi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vextractf128 $1, %ymm0, 16(%rsi) # sched: [5:1.00] +; SANDY-NEXT: vmovups %xmm0, (%rsi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movups: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovups (%rdi), %ymm0 # sched: [4:0.50] +; HASWELL-NEXT: vmovups (%rdi), %ymm0 # sched: [?:5.000000e-01] ; HASWELL-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vmovups %ymm0, (%rsi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovups %ymm0, (%rsi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movups: ; BTVER2: # BB#0: @@ -1708,14 +1708,14 @@ define <4 x double> @test_mulpd(<4 x double> %a0, <4 x double> %a1, <4 x double> ; SANDY-LABEL: test_mulpd: ; SANDY: # BB#0: ; SANDY-NEXT: vmulpd %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; SANDY-NEXT: vmulpd (%rdi), %ymm0, %ymm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmulpd (%rdi), %ymm0, %ymm0 # sched: [12:2.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_mulpd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmulpd %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: vmulpd (%rdi), %ymm0, %ymm0 # sched: [9:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmulpd %ymm1, %ymm0, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: vmulpd (%rdi), %ymm0, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_mulpd: ; BTVER2: # BB#0: @@ -1738,14 +1738,14 @@ define <8 x float> @test_mulps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a ; SANDY-LABEL: test_mulps: ; SANDY: # BB#0: ; SANDY-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; SANDY-NEXT: vmulps (%rdi), %ymm0, %ymm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmulps (%rdi), %ymm0, %ymm0 # sched: [12:2.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_mulps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: vmulps (%rdi), %ymm0, %ymm0 # sched: [9:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: vmulps (%rdi), %ymm0, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_mulps: ; BTVER2: # BB#0: @@ -1767,17 +1767,17 @@ define <8 x float> @test_mulps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a define <4 x double> @orpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) { ; SANDY-LABEL: orpd: ; SANDY: # BB#0: -; SANDY-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SANDY-NEXT: vorpd (%rdi), %ymm0, %ymm0 # sched: [5:0.50] +; SANDY-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00] +; SANDY-NEXT: vorpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00] ; SANDY-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: orpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: vorpd (%rdi), %ymm0, %ymm0 # sched: [5:1.00] +; HASWELL-NEXT: vorpd (%rdi), %ymm0, %ymm0 # sched: [1:1.00] ; HASWELL-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: orpd: ; BTVER2: # BB#0: @@ -1806,17 +1806,17 @@ define <4 x double> @orpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) define <8 x float> @test_orps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) { ; SANDY-LABEL: test_orps: ; SANDY: # BB#0: -; SANDY-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SANDY-NEXT: vorps (%rdi), %ymm0, %ymm0 # sched: [5:0.50] +; SANDY-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00] +; SANDY-NEXT: vorps (%rdi), %ymm0, %ymm0 # sched: [8:1.00] ; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_orps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: vorps (%rdi), %ymm0, %ymm0 # sched: [5:1.00] +; HASWELL-NEXT: vorps (%rdi), %ymm0, %ymm0 # sched: [1:1.00] ; HASWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_orps: ; BTVER2: # BB#0: @@ -1846,16 +1846,16 @@ define <2 x double> @test_permilpd(<2 x double> %a0, <2 x double> *%a1) { ; SANDY-LABEL: test_permilpd: ; SANDY: # BB#0: ; SANDY-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] sched: [1:1.00] -; SANDY-NEXT: vpermilpd {{.*#+}} xmm1 = mem[1,0] sched: [5:1.00] +; SANDY-NEXT: vpermilpd {{.*#+}} xmm1 = mem[1,0] sched: [7:1.00] ; SANDY-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_permilpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] sched: [1:1.00] -; HASWELL-NEXT: vpermilpd {{.*#+}} xmm1 = mem[1,0] sched: [5:1.00] +; HASWELL-NEXT: vpermilpd {{.*#+}} xmm1 = mem[1,0] sched: [1:1.00] ; HASWELL-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_permilpd: ; BTVER2: # BB#0: @@ -1880,17 +1880,17 @@ define <2 x double> @test_permilpd(<2 x double> %a0, <2 x double> *%a1) { define <4 x double> @test_permilpd_ymm(<4 x double> %a0, <4 x double> *%a1) { ; SANDY-LABEL: test_permilpd_ymm: ; SANDY: # BB#0: -; SANDY-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3] sched: [1:1.00] +; SANDY-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3] sched: [8:1.00] ; SANDY-NEXT: vpermilpd {{.*#+}} ymm1 = mem[1,0,2,3] sched: [5:1.00] ; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_permilpd_ymm: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3] sched: [1:1.00] ; HASWELL-NEXT: vpermilpd {{.*#+}} ymm1 = mem[1,0,2,3] sched: [5:1.00] ; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_permilpd_ymm: ; BTVER2: # BB#0: @@ -1916,16 +1916,16 @@ define <4 x float> @test_permilps(<4 x float> %a0, <4 x float> *%a1) { ; SANDY-LABEL: test_permilps: ; SANDY: # BB#0: ; SANDY-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0] sched: [1:1.00] -; SANDY-NEXT: vpermilps {{.*#+}} xmm1 = mem[3,2,1,0] sched: [5:1.00] +; SANDY-NEXT: vpermilps {{.*#+}} xmm1 = mem[3,2,1,0] sched: [7:1.00] ; SANDY-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_permilps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0] sched: [1:1.00] -; HASWELL-NEXT: vpermilps {{.*#+}} xmm1 = mem[3,2,1,0] sched: [5:1.00] +; HASWELL-NEXT: vpermilps {{.*#+}} xmm1 = mem[3,2,1,0] sched: [1:1.00] ; HASWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_permilps: ; BTVER2: # BB#0: @@ -1950,17 +1950,17 @@ define <4 x float> @test_permilps(<4 x float> %a0, <4 x float> *%a1) { define <8 x float> @test_permilps_ymm(<8 x float> %a0, <8 x float> *%a1) { ; SANDY-LABEL: test_permilps_ymm: ; SANDY: # BB#0: -; SANDY-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00] +; SANDY-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [8:1.00] ; SANDY-NEXT: vpermilps {{.*#+}} ymm1 = mem[3,2,1,0,7,6,5,4] sched: [5:1.00] ; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_permilps_ymm: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00] ; HASWELL-NEXT: vpermilps {{.*#+}} ymm1 = mem[3,2,1,0,7,6,5,4] sched: [5:1.00] ; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_permilps_ymm: ; BTVER2: # BB#0: @@ -1986,14 +1986,14 @@ define <2 x double> @test_permilvarpd(<2 x double> %a0, <2 x i64> %a1, <2 x i64> ; SANDY-LABEL: test_permilvarpd: ; SANDY: # BB#0: ; SANDY-NEXT: vpermilpd %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; SANDY-NEXT: vpermilpd (%rdi), %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpermilpd (%rdi), %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_permilvarpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpermilpd %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpermilpd (%rdi), %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpermilpd (%rdi), %xmm0, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_permilvarpd: ; BTVER2: # BB#0: @@ -2018,13 +2018,13 @@ define <4 x double> @test_permilvarpd_ymm(<4 x double> %a0, <4 x i64> %a1, <4 x ; SANDY: # BB#0: ; SANDY-NEXT: vpermilpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00] ; SANDY-NEXT: vpermilpd (%rdi), %ymm0, %ymm0 # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_permilvarpd_ymm: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpermilpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00] ; HASWELL-NEXT: vpermilpd (%rdi), %ymm0, %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_permilvarpd_ymm: ; BTVER2: # BB#0: @@ -2048,14 +2048,14 @@ define <4 x float> @test_permilvarps(<4 x float> %a0, <4 x i32> %a1, <4 x i32> * ; SANDY-LABEL: test_permilvarps: ; SANDY: # BB#0: ; SANDY-NEXT: vpermilps %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; SANDY-NEXT: vpermilps (%rdi), %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpermilps (%rdi), %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_permilvarps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpermilps %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpermilps (%rdi), %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpermilps (%rdi), %xmm0, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_permilvarps: ; BTVER2: # BB#0: @@ -2080,13 +2080,13 @@ define <8 x float> @test_permilvarps_ymm(<8 x float> %a0, <8 x i32> %a1, <8 x i3 ; SANDY: # BB#0: ; SANDY-NEXT: vpermilps %ymm1, %ymm0, %ymm0 # sched: [1:1.00] ; SANDY-NEXT: vpermilps (%rdi), %ymm0, %ymm0 # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_permilvarps_ymm: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpermilps %ymm1, %ymm0, %ymm0 # sched: [1:1.00] ; HASWELL-NEXT: vpermilps (%rdi), %ymm0, %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_permilvarps_ymm: ; BTVER2: # BB#0: @@ -2112,14 +2112,14 @@ define <8 x float> @test_rcpps(<8 x float> %a0, <8 x float> *%a1) { ; SANDY-NEXT: vrcpps %ymm0, %ymm0 # sched: [5:1.00] ; SANDY-NEXT: vrcpps (%rdi), %ymm1 # sched: [9:1.00] ; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_rcpps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrcpps (%rdi), %ymm1 # sched: [11:2.00] -; HASWELL-NEXT: vrcpps %ymm0, %ymm0 # sched: [7:2.00] +; HASWELL-NEXT: vrcpps %ymm0, %ymm0 # sched: [11:2.00] ; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_rcpps: ; BTVER2: # BB#0: @@ -2148,14 +2148,14 @@ define <4 x double> @test_roundpd(<4 x double> %a0, <4 x double> *%a1) { ; SANDY-NEXT: vroundpd $7, %ymm0, %ymm0 # sched: [3:1.00] ; SANDY-NEXT: vroundpd $7, (%rdi), %ymm1 # sched: [7:1.00] ; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_roundpd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vroundpd $7, %ymm0, %ymm0 # sched: [6:2.00] -; HASWELL-NEXT: vroundpd $7, (%rdi), %ymm1 # sched: [10:2.00] +; HASWELL-NEXT: vroundpd $7, %ymm0, %ymm0 # sched: [3:1.00] +; HASWELL-NEXT: vroundpd $7, (%rdi), %ymm1 # sched: [7:1.00] ; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_roundpd: ; BTVER2: # BB#0: @@ -2184,14 +2184,14 @@ define <8 x float> @test_roundps(<8 x float> %a0, <8 x float> *%a1) { ; SANDY-NEXT: vroundps $7, %ymm0, %ymm0 # sched: [3:1.00] ; SANDY-NEXT: vroundps $7, (%rdi), %ymm1 # sched: [7:1.00] ; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_roundps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vroundps $7, %ymm0, %ymm0 # sched: [6:2.00] -; HASWELL-NEXT: vroundps $7, (%rdi), %ymm1 # sched: [10:2.00] +; HASWELL-NEXT: vroundps $7, %ymm0, %ymm0 # sched: [3:1.00] +; HASWELL-NEXT: vroundps $7, (%rdi), %ymm1 # sched: [7:1.00] ; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_roundps: ; BTVER2: # BB#0: @@ -2217,17 +2217,17 @@ declare <8 x float> @llvm.x86.avx.round.ps.256(<8 x float>, i32) nounwind readno define <8 x float> @test_rsqrtps(<8 x float> %a0, <8 x float> *%a1) { ; SANDY-LABEL: test_rsqrtps: ; SANDY: # BB#0: -; SANDY-NEXT: vrsqrtps %ymm0, %ymm0 # sched: [5:1.00] -; SANDY-NEXT: vrsqrtps (%rdi), %ymm1 # sched: [9:1.00] +; SANDY-NEXT: vrsqrtps (%rdi), %ymm1 # sched: [14:3.00] +; SANDY-NEXT: vrsqrtps %ymm0, %ymm0 # sched: [7:3.00] ; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_rsqrtps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrsqrtps (%rdi), %ymm1 # sched: [11:2.00] -; HASWELL-NEXT: vrsqrtps %ymm0, %ymm0 # sched: [7:2.00] +; HASWELL-NEXT: vrsqrtps %ymm0, %ymm0 # sched: [11:2.00] ; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_rsqrtps: ; BTVER2: # BB#0: @@ -2254,16 +2254,16 @@ define <4 x double> @test_shufpd(<4 x double> %a0, <4 x double> %a1, <4 x double ; SANDY-LABEL: test_shufpd: ; SANDY: # BB#0: ; SANDY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[3] sched: [1:1.00] -; SANDY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],mem[0],ymm1[2],mem[3] sched: [5:1.00] +; SANDY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],mem[0],ymm1[2],mem[3] sched: [8:1.00] ; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_shufpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[3] sched: [1:1.00] -; HASWELL-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],mem[0],ymm1[2],mem[3] sched: [5:1.00] +; HASWELL-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],mem[0],ymm1[2],mem[3] sched: [1:1.00] ; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_shufpd: ; BTVER2: # BB#0: @@ -2289,14 +2289,14 @@ define <8 x float> @test_shufps(<8 x float> %a0, <8 x float> %a1, <8 x float> *% ; SANDY-LABEL: test_shufps: ; SANDY: # BB#0: ; SANDY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4] sched: [1:1.00] -; SANDY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,3],mem[0,0],ymm0[4,7],mem[4,4] sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,3],mem[0,0],ymm0[4,7],mem[4,4] sched: [8:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_shufps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4] sched: [1:1.00] -; HASWELL-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,3],mem[0,0],ymm0[4,7],mem[4,4] sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,3],mem[0,0],ymm0[4,7],mem[4,4] sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_shufps: ; BTVER2: # BB#0: @@ -2318,17 +2318,17 @@ define <8 x float> @test_shufps(<8 x float> %a0, <8 x float> %a1, <8 x float> *% define <4 x double> @test_sqrtpd(<4 x double> %a0, <4 x double> *%a1) { ; SANDY-LABEL: test_sqrtpd: ; SANDY: # BB#0: -; SANDY-NEXT: vsqrtpd %ymm0, %ymm0 # sched: [15:1.00] -; SANDY-NEXT: vsqrtpd (%rdi), %ymm1 # sched: [19:1.00] +; SANDY-NEXT: vsqrtpd (%rdi), %ymm1 # sched: [52:3.00] +; SANDY-NEXT: vsqrtpd %ymm0, %ymm0 # sched: [45:3.00] ; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_sqrtpd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vsqrtpd (%rdi), %ymm1 # sched: [32:2.00] -; HASWELL-NEXT: vsqrtpd %ymm0, %ymm0 # sched: [28:2.00] +; HASWELL-NEXT: vsqrtpd (%rdi), %ymm1 # sched: [35:2.00] +; HASWELL-NEXT: vsqrtpd %ymm0, %ymm0 # sched: [35:2.00] ; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_sqrtpd: ; BTVER2: # BB#0: @@ -2354,17 +2354,17 @@ declare <4 x double> @llvm.x86.avx.sqrt.pd.256(<4 x double>) nounwind readnone define <8 x float> @test_sqrtps(<8 x float> %a0, <8 x float> *%a1) { ; SANDY-LABEL: test_sqrtps: ; SANDY: # BB#0: -; SANDY-NEXT: vsqrtps %ymm0, %ymm0 # sched: [15:1.00] -; SANDY-NEXT: vsqrtps (%rdi), %ymm1 # sched: [19:1.00] +; SANDY-NEXT: vsqrtps (%rdi), %ymm1 # sched: [36:3.00] +; SANDY-NEXT: vsqrtps %ymm0, %ymm0 # sched: [29:3.00] ; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_sqrtps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vsqrtps (%rdi), %ymm1 # sched: [23:2.00] -; HASWELL-NEXT: vsqrtps %ymm0, %ymm0 # sched: [19:2.00] +; HASWELL-NEXT: vsqrtps (%rdi), %ymm1 # sched: [21:2.00] +; HASWELL-NEXT: vsqrtps %ymm0, %ymm0 # sched: [21:2.00] ; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_sqrtps: ; BTVER2: # BB#0: @@ -2391,14 +2391,14 @@ define <4 x double> @test_subpd(<4 x double> %a0, <4 x double> %a1, <4 x double> ; SANDY-LABEL: test_subpd: ; SANDY: # BB#0: ; SANDY-NEXT: vsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vsubpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vsubpd (%rdi), %ymm0, %ymm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_subpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vsubpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vsubpd (%rdi), %ymm0, %ymm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_subpd: ; BTVER2: # BB#0: @@ -2421,14 +2421,14 @@ define <8 x float> @test_subps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a ; SANDY-LABEL: test_subps: ; SANDY: # BB#0: ; SANDY-NEXT: vsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vsubps (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vsubps (%rdi), %ymm0, %ymm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_subps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vsubps (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vsubps (%rdi), %ymm0, %ymm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_subps: ; BTVER2: # BB#0: @@ -2451,20 +2451,20 @@ define i32 @test_testpd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) { ; SANDY-LABEL: test_testpd: ; SANDY: # BB#0: ; SANDY-NEXT: xorl %eax, %eax # sched: [1:0.33] -; SANDY-NEXT: vtestpd %xmm1, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: setb %al # sched: [1:0.33] -; SANDY-NEXT: vtestpd (%rdi), %xmm0 # sched: [5:0.50] +; SANDY-NEXT: vtestpd %xmm1, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: setb %al # sched: [1:1.00] +; SANDY-NEXT: vtestpd (%rdi), %xmm0 # sched: [7:1.00] ; SANDY-NEXT: adcl $0, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_testpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: xorl %eax, %eax # sched: [1:0.25] -; HASWELL-NEXT: vtestpd %xmm1, %xmm0 # sched: [1:0.33] -; HASWELL-NEXT: setb %al # sched: [1:0.50] -; HASWELL-NEXT: vtestpd (%rdi), %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: adcl $0, %eax # sched: [2:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vtestpd %xmm1, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: setb %al # sched: [1:1.00] +; HASWELL-NEXT: vtestpd (%rdi), %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: adcl $0, %eax # sched: [1:0.25] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_testpd: ; BTVER2: # BB#0: @@ -2495,22 +2495,22 @@ define i32 @test_testpd_ymm(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a ; SANDY-LABEL: test_testpd_ymm: ; SANDY: # BB#0: ; SANDY-NEXT: xorl %eax, %eax # sched: [1:0.33] -; SANDY-NEXT: vtestpd %ymm1, %ymm0 # sched: [1:0.33] -; SANDY-NEXT: setb %al # sched: [1:0.33] -; SANDY-NEXT: vtestpd (%rdi), %ymm0 # sched: [5:0.50] +; SANDY-NEXT: vtestpd %ymm1, %ymm0 # sched: [1:1.00] +; SANDY-NEXT: setb %al # sched: [1:1.00] +; SANDY-NEXT: vtestpd (%rdi), %ymm0 # sched: [8:1.00] ; SANDY-NEXT: adcl $0, %eax # sched: [1:0.33] ; SANDY-NEXT: vzeroupper # sched: [?:0.000000e+00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_testpd_ymm: ; HASWELL: # BB#0: ; HASWELL-NEXT: xorl %eax, %eax # sched: [1:0.25] -; HASWELL-NEXT: vtestpd %ymm1, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: setb %al # sched: [1:0.50] -; HASWELL-NEXT: vtestpd (%rdi), %ymm0 # sched: [5:0.50] -; HASWELL-NEXT: adcl $0, %eax # sched: [2:0.50] -; HASWELL-NEXT: vzeroupper # sched: [1:0.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vtestpd %ymm1, %ymm0 # sched: [1:1.00] +; HASWELL-NEXT: setb %al # sched: [1:1.00] +; HASWELL-NEXT: vtestpd (%rdi), %ymm0 # sched: [1:1.00] +; HASWELL-NEXT: adcl $0, %eax # sched: [1:0.25] +; HASWELL-NEXT: vzeroupper # sched: [4:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_testpd_ymm: ; BTVER2: # BB#0: @@ -2542,20 +2542,20 @@ define i32 @test_testps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) { ; SANDY-LABEL: test_testps: ; SANDY: # BB#0: ; SANDY-NEXT: xorl %eax, %eax # sched: [1:0.33] -; SANDY-NEXT: vtestps %xmm1, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: setb %al # sched: [1:0.33] -; SANDY-NEXT: vtestps (%rdi), %xmm0 # sched: [5:0.50] +; SANDY-NEXT: vtestps %xmm1, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: setb %al # sched: [1:1.00] +; SANDY-NEXT: vtestps (%rdi), %xmm0 # sched: [7:1.00] ; SANDY-NEXT: adcl $0, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_testps: ; HASWELL: # BB#0: ; HASWELL-NEXT: xorl %eax, %eax # sched: [1:0.25] -; HASWELL-NEXT: vtestps %xmm1, %xmm0 # sched: [1:0.33] -; HASWELL-NEXT: setb %al # sched: [1:0.50] -; HASWELL-NEXT: vtestps (%rdi), %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: adcl $0, %eax # sched: [2:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vtestps %xmm1, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: setb %al # sched: [1:1.00] +; HASWELL-NEXT: vtestps (%rdi), %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: adcl $0, %eax # sched: [1:0.25] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_testps: ; BTVER2: # BB#0: @@ -2586,22 +2586,22 @@ define i32 @test_testps_ymm(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) ; SANDY-LABEL: test_testps_ymm: ; SANDY: # BB#0: ; SANDY-NEXT: xorl %eax, %eax # sched: [1:0.33] -; SANDY-NEXT: vtestps %ymm1, %ymm0 # sched: [1:0.33] -; SANDY-NEXT: setb %al # sched: [1:0.33] -; SANDY-NEXT: vtestps (%rdi), %ymm0 # sched: [5:0.50] +; SANDY-NEXT: vtestps %ymm1, %ymm0 # sched: [1:1.00] +; SANDY-NEXT: setb %al # sched: [1:1.00] +; SANDY-NEXT: vtestps (%rdi), %ymm0 # sched: [8:1.00] ; SANDY-NEXT: adcl $0, %eax # sched: [1:0.33] ; SANDY-NEXT: vzeroupper # sched: [?:0.000000e+00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_testps_ymm: ; HASWELL: # BB#0: ; HASWELL-NEXT: xorl %eax, %eax # sched: [1:0.25] -; HASWELL-NEXT: vtestps %ymm1, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: setb %al # sched: [1:0.50] -; HASWELL-NEXT: vtestps (%rdi), %ymm0 # sched: [5:0.50] -; HASWELL-NEXT: adcl $0, %eax # sched: [2:0.50] -; HASWELL-NEXT: vzeroupper # sched: [1:0.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vtestps %ymm1, %ymm0 # sched: [1:1.00] +; HASWELL-NEXT: setb %al # sched: [1:1.00] +; HASWELL-NEXT: vtestps (%rdi), %ymm0 # sched: [1:1.00] +; HASWELL-NEXT: adcl $0, %eax # sched: [1:0.25] +; HASWELL-NEXT: vzeroupper # sched: [4:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_testps_ymm: ; BTVER2: # BB#0: @@ -2635,14 +2635,14 @@ define <4 x double> @test_unpckhpd(<4 x double> %a0, <4 x double> %a1, <4 x doub ; SANDY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00] ; SANDY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],mem[1],ymm1[3],mem[3] sched: [5:1.00] ; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_unpckhpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00] ; HASWELL-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],mem[1],ymm1[3],mem[3] sched: [5:1.00] ; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_unpckhpd: ; BTVER2: # BB#0: @@ -2669,13 +2669,13 @@ define <8 x float> @test_unpckhps(<8 x float> %a0, <8 x float> %a1, <8 x float> ; SANDY: # BB#0: ; SANDY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00] ; SANDY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_unpckhps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00] ; HASWELL-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_unpckhps: ; BTVER2: # BB#0: @@ -2698,16 +2698,16 @@ define <4 x double> @test_unpcklpd(<4 x double> %a0, <4 x double> %a1, <4 x doub ; SANDY-LABEL: test_unpcklpd: ; SANDY: # BB#0: ; SANDY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00] -; SANDY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] sched: [5:1.00] +; SANDY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] sched: [8:1.00] ; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_unpcklpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00] -; HASWELL-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] sched: [5:1.00] +; HASWELL-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] sched: [1:1.00] ; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_unpcklpd: ; BTVER2: # BB#0: @@ -2733,14 +2733,14 @@ define <8 x float> @test_unpcklps(<8 x float> %a0, <8 x float> %a1, <8 x float> ; SANDY-LABEL: test_unpcklps: ; SANDY: # BB#0: ; SANDY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00] -; SANDY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_unpcklps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00] -; HASWELL-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_unpcklps: ; BTVER2: # BB#0: @@ -2765,14 +2765,14 @@ define <4 x double> @test_xorpd(<4 x double> %a0, <4 x double> %a1, <4 x double> ; SANDY-NEXT: vxorpd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] ; SANDY-NEXT: vxorpd (%rdi), %ymm0, %ymm0 # sched: [5:0.50] ; SANDY-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_xorpd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vxorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: vxorpd (%rdi), %ymm0, %ymm0 # sched: [5:1.00] +; HASWELL-NEXT: vxorpd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] +; HASWELL-NEXT: vxorpd (%rdi), %ymm0, %ymm0 # sched: [5:0.50] ; HASWELL-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_xorpd: ; BTVER2: # BB#0: @@ -2804,14 +2804,14 @@ define <8 x float> @test_xorps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a ; SANDY-NEXT: vxorps %ymm1, %ymm0, %ymm0 # sched: [1:0.33] ; SANDY-NEXT: vxorps (%rdi), %ymm0, %ymm0 # sched: [5:0.50] ; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_xorps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vxorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: vxorps (%rdi), %ymm0, %ymm0 # sched: [5:1.00] +; HASWELL-NEXT: vxorps %ymm1, %ymm0, %ymm0 # sched: [1:0.33] +; HASWELL-NEXT: vxorps (%rdi), %ymm0, %ymm0 # sched: [5:0.50] ; HASWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_xorps: ; BTVER2: # BB#0: @@ -2841,12 +2841,12 @@ define void @test_zeroall() { ; SANDY-LABEL: test_zeroall: ; SANDY: # BB#0: ; SANDY-NEXT: vzeroall # sched: [?:0.000000e+00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_zeroall: ; HASWELL: # BB#0: -; HASWELL-NEXT: vzeroall # sched: [1:0.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vzeroall # sched: [16:16.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_zeroall: ; BTVER2: # BB#0: @@ -2866,12 +2866,12 @@ define void @test_zeroupper() { ; SANDY-LABEL: test_zeroupper: ; SANDY: # BB#0: ; SANDY-NEXT: vzeroupper # sched: [?:0.000000e+00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_zeroupper: ; HASWELL: # BB#0: -; HASWELL-NEXT: vzeroupper # sched: [1:0.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vzeroupper # sched: [4:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_zeroupper: ; BTVER2: # BB#0: diff --git a/llvm/test/CodeGen/X86/avx2-intrinsics-x86.ll b/llvm/test/CodeGen/X86/avx2-intrinsics-x86.ll index 52e37dbf269..a82c51747a2 100644 --- a/llvm/test/CodeGen/X86/avx2-intrinsics-x86.ll +++ b/llvm/test/CodeGen/X86/avx2-intrinsics-x86.ll @@ -1619,10 +1619,10 @@ define <8 x float> @test_gather_mask(<8 x float> %a0, float* %a, <8 x i32> %idx ; ; AVX512VL-LABEL: test_gather_mask: ; AVX512VL: ## BB#0: -; AVX512VL-NEXT: movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x04] -; AVX512VL-NEXT: vmovaps %ymm2, %ymm3 ## EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xda] -; AVX512VL-NEXT: vgatherdps %ymm3, (%eax,%ymm1,4), %ymm0 ## encoding: [0xc4,0xe2,0x65,0x92,0x04,0x88] ; AVX512VL-NEXT: movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x08] +; AVX512VL-NEXT: movl {{[0-9]+}}(%esp), %ecx ## encoding: [0x8b,0x4c,0x24,0x04] +; AVX512VL-NEXT: vmovaps %ymm2, %ymm3 ## EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xda] +; AVX512VL-NEXT: vgatherdps %ymm3, (%ecx,%ymm1,4), %ymm0 ## encoding: [0xc4,0xe2,0x65,0x92,0x04,0x89] ; AVX512VL-NEXT: vmovups %ymm2, (%eax) ## EVEX TO VEX Compression encoding: [0xc5,0xfc,0x11,0x10] ; AVX512VL-NEXT: retl ## encoding: [0xc3] %a_i8 = bitcast float* %a to i8* diff --git a/llvm/test/CodeGen/X86/avx2-schedule.ll b/llvm/test/CodeGen/X86/avx2-schedule.ll index 042bc217b97..ca5a132f4f4 100644 --- a/llvm/test/CodeGen/X86/avx2-schedule.ll +++ b/llvm/test/CodeGen/X86/avx2-schedule.ll @@ -9,7 +9,7 @@ define <32 x i8> @test_pabsb(<32 x i8> %a0, <32 x i8> *%a1) { ; HASWELL-NEXT: vpabsb %ymm0, %ymm0 # sched: [1:0.50] ; HASWELL-NEXT: vpabsb (%rdi), %ymm1 # sched: [5:0.50] ; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_pabsb: ; ZNVER1: # BB#0: @@ -29,9 +29,9 @@ define <8 x i32> @test_pabsd(<8 x i32> %a0, <8 x i32> *%a1) { ; HASWELL-LABEL: test_pabsd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpabsd %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpabsd (%rdi), %ymm1 # sched: [5:0.50] +; HASWELL-NEXT: vpabsd (%rdi), %ymm1 # sched: [1:0.50] ; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_pabsd: ; ZNVER1: # BB#0: @@ -51,9 +51,9 @@ define <16 x i16> @test_pabsw(<16 x i16> %a0, <16 x i16> *%a1) { ; HASWELL-LABEL: test_pabsw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpabsw %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpabsw (%rdi), %ymm1 # sched: [5:0.50] +; HASWELL-NEXT: vpabsw (%rdi), %ymm1 # sched: [1:0.50] ; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_pabsw: ; ZNVER1: # BB#0: @@ -74,7 +74,7 @@ define <32 x i8> @test_paddb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { ; HASWELL: # BB#0: ; HASWELL-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] ; HASWELL-NEXT: vpaddb (%rdi), %ymm0, %ymm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_paddb: ; ZNVER1: # BB#0: @@ -92,7 +92,7 @@ define <8 x i32> @test_paddd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { ; HASWELL: # BB#0: ; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] ; HASWELL-NEXT: vpaddd (%rdi), %ymm0, %ymm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_paddd: ; ZNVER1: # BB#0: @@ -109,8 +109,8 @@ define <4 x i64> @test_paddq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { ; HASWELL-LABEL: test_paddq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddq (%rdi), %ymm0, %ymm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpaddq (%rdi), %ymm0, %ymm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_paddq: ; ZNVER1: # BB#0: @@ -128,7 +128,7 @@ define <16 x i16> @test_paddw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { ; HASWELL: # BB#0: ; HASWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] ; HASWELL-NEXT: vpaddw (%rdi), %ymm0, %ymm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_paddw: ; ZNVER1: # BB#0: @@ -145,9 +145,9 @@ define <4 x i64> @test_pand(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { ; HASWELL-LABEL: test_pand: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpand %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: vpand (%rdi), %ymm0, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: vpand (%rdi), %ymm0, %ymm0 # sched: [1:0.50] ; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_pand: ; ZNVER1: # BB#0: @@ -166,9 +166,9 @@ define <4 x i64> @test_pandn(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { ; HASWELL-LABEL: test_pandn: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpandn %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: vpandn (%rdi), %ymm0, %ymm1 # sched: [5:0.50] +; HASWELL-NEXT: vpandn (%rdi), %ymm0, %ymm1 # sched: [1:0.50] ; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_pandn: ; ZNVER1: # BB#0: @@ -190,7 +190,7 @@ define <8 x i32> @test_pmulld(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmulld %ymm1, %ymm0, %ymm0 # sched: [10:2.00] ; HASWELL-NEXT: vpmulld (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_pmulld: ; ZNVER1: # BB#0: @@ -207,8 +207,8 @@ define <16 x i16> @test_pmullw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) ; HASWELL-LABEL: test_pmullw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmullw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: vpmullw (%rdi), %ymm0, %ymm0 # sched: [9:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpmullw (%rdi), %ymm0, %ymm0 # sched: [5:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_pmullw: ; ZNVER1: # BB#0: @@ -225,9 +225,9 @@ define <4 x i64> @test_por(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { ; HASWELL-LABEL: test_por: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: vpor (%rdi), %ymm0, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: vpor (%rdi), %ymm0, %ymm0 # sched: [1:0.50] ; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_por: ; ZNVER1: # BB#0: @@ -246,8 +246,8 @@ define <32 x i8> @test_psubb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { ; HASWELL-LABEL: test_psubb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsubb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubb (%rdi), %ymm0, %ymm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpsubb (%rdi), %ymm0, %ymm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_psubb: ; ZNVER1: # BB#0: @@ -264,8 +264,8 @@ define <8 x i32> @test_psubd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { ; HASWELL-LABEL: test_psubd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubd (%rdi), %ymm0, %ymm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpsubd (%rdi), %ymm0, %ymm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_psubd: ; ZNVER1: # BB#0: @@ -282,8 +282,8 @@ define <4 x i64> @test_psubq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { ; HASWELL-LABEL: test_psubq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsubq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubq (%rdi), %ymm0, %ymm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpsubq (%rdi), %ymm0, %ymm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_psubq: ; ZNVER1: # BB#0: @@ -300,8 +300,8 @@ define <16 x i16> @test_psubw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { ; HASWELL-LABEL: test_psubw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsubw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubw (%rdi), %ymm0, %ymm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpsubw (%rdi), %ymm0, %ymm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_psubw: ; ZNVER1: # BB#0: @@ -318,9 +318,9 @@ define <4 x i64> @test_pxor(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { ; HASWELL-LABEL: test_pxor: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpxor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: vpxor (%rdi), %ymm0, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: vpxor (%rdi), %ymm0, %ymm0 # sched: [1:0.50] ; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; ZNVER1-LABEL: test_pxor: ; ZNVER1: # BB#0: diff --git a/llvm/test/CodeGen/X86/avx2-vector-shifts.ll b/llvm/test/CodeGen/X86/avx2-vector-shifts.ll index 45a1cd97503..958c7746e29 100644 --- a/llvm/test/CodeGen/X86/avx2-vector-shifts.ll +++ b/llvm/test/CodeGen/X86/avx2-vector-shifts.ll @@ -381,6 +381,7 @@ define <4 x i32> @srl_trunc_and_v4i64(<4 x i32> %x, <4 x i64> %y) nounwind { ; X32-NEXT: vpsrlvd %xmm1, %xmm0, %xmm0 ; X32-NEXT: vzeroupper ; X32-NEXT: retl +; X32-NEXT: ## -- End function ; ; X64-LABEL: srl_trunc_and_v4i64: ; X64: ## BB#0: @@ -391,6 +392,7 @@ define <4 x i32> @srl_trunc_and_v4i64(<4 x i32> %x, <4 x i64> %y) nounwind { ; X64-NEXT: vpsrlvd %xmm1, %xmm0, %xmm0 ; X64-NEXT: vzeroupper ; X64-NEXT: retq +; X64-NEXT: ## -- End function %and = and <4 x i64> %y, <i64 8, i64 8, i64 8, i64 8> %trunc = trunc <4 x i64> %and to <4 x i32> %sra = lshr <4 x i32> %x, %trunc @@ -412,6 +414,7 @@ define <8 x i16> @shl_8i16(<8 x i16> %r, <8 x i16> %a) nounwind { ; X32-NEXT: ## kill: %XMM0<def> %XMM0<kill> %YMM0<kill> ; X32-NEXT: vzeroupper ; X32-NEXT: retl +; X32-NEXT: ## -- End function ; ; X64-LABEL: shl_8i16: ; X64: ## BB#0: @@ -423,6 +426,7 @@ define <8 x i16> @shl_8i16(<8 x i16> %r, <8 x i16> %a) nounwind { ; X64-NEXT: ## kill: %XMM0<def> %XMM0<kill> %YMM0<kill> ; X64-NEXT: vzeroupper ; X64-NEXT: retq +; X64-NEXT: ## -- End function %shl = shl <8 x i16> %r, %a ret <8 x i16> %shl } @@ -434,13 +438,14 @@ define <16 x i16> @shl_16i16(<16 x i16> %r, <16 x i16> %a) nounwind { ; X32-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15] ; X32-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15] ; X32-NEXT: vpsllvd %ymm3, %ymm4, %ymm3 -; X32-NEXT: vpsrld $16, %ymm3, %ymm3 ; X32-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11] ; X32-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11] ; X32-NEXT: vpsllvd %ymm1, %ymm0, %ymm0 +; X32-NEXT: vpsrld $16, %ymm3, %ymm1 ; X32-NEXT: vpsrld $16, %ymm0, %ymm0 -; X32-NEXT: vpackusdw %ymm3, %ymm0, %ymm0 +; X32-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 ; X32-NEXT: retl +; X32-NEXT: ## -- End function ; ; X64-LABEL: shl_16i16: ; X64: ## BB#0: @@ -448,13 +453,14 @@ define <16 x i16> @shl_16i16(<16 x i16> %r, <16 x i16> %a) nounwind { ; X64-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15] ; X64-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15] ; X64-NEXT: vpsllvd %ymm3, %ymm4, %ymm3 -; X64-NEXT: vpsrld $16, %ymm3, %ymm3 ; X64-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11] ; X64-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11] ; X64-NEXT: vpsllvd %ymm1, %ymm0, %ymm0 +; X64-NEXT: vpsrld $16, %ymm3, %ymm1 ; X64-NEXT: vpsrld $16, %ymm0, %ymm0 -; X64-NEXT: vpackusdw %ymm3, %ymm0, %ymm0 +; X64-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 ; X64-NEXT: retq +; X64-NEXT: ## -- End function %shl = shl <16 x i16> %r, %a ret <16 x i16> %shl } @@ -474,6 +480,7 @@ define <32 x i8> @shl_32i8(<32 x i8> %r, <32 x i8> %a) nounwind { ; X32-NEXT: vpaddb %ymm1, %ymm1, %ymm1 ; X32-NEXT: vpblendvb %ymm1, %ymm2, %ymm0, %ymm0 ; X32-NEXT: retl +; X32-NEXT: ## -- End function ; ; X64-LABEL: shl_32i8: ; X64: ## BB#0: @@ -489,6 +496,7 @@ define <32 x i8> @shl_32i8(<32 x i8> %r, <32 x i8> %a) nounwind { ; X64-NEXT: vpaddb %ymm1, %ymm1, %ymm1 ; X64-NEXT: vpblendvb %ymm1, %ymm2, %ymm0, %ymm0 ; X64-NEXT: retq +; X64-NEXT: ## -- End function %shl = shl <32 x i8> %r, %a ret <32 x i8> %shl } @@ -504,6 +512,7 @@ define <8 x i16> @ashr_8i16(<8 x i16> %r, <8 x i16> %a) nounwind { ; X32-NEXT: ## kill: %XMM0<def> %XMM0<kill> %YMM0<kill> ; X32-NEXT: vzeroupper ; X32-NEXT: retl +; X32-NEXT: ## -- End function ; ; X64-LABEL: ashr_8i16: ; X64: ## BB#0: @@ -515,6 +524,7 @@ define <8 x i16> @ashr_8i16(<8 x i16> %r, <8 x i16> %a) nounwind { ; X64-NEXT: ## kill: %XMM0<def> %XMM0<kill> %YMM0<kill> ; X64-NEXT: vzeroupper ; X64-NEXT: retq +; X64-NEXT: ## -- End function %ashr = ashr <8 x i16> %r, %a ret <8 x i16> %ashr } @@ -526,13 +536,14 @@ define <16 x i16> @ashr_16i16(<16 x i16> %r, <16 x i16> %a) nounwind { ; X32-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15] ; X32-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15] ; X32-NEXT: vpsravd %ymm3, %ymm4, %ymm3 -; X32-NEXT: vpsrld $16, %ymm3, %ymm3 ; X32-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11] ; X32-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11] ; X32-NEXT: vpsravd %ymm1, %ymm0, %ymm0 +; X32-NEXT: vpsrld $16, %ymm3, %ymm1 ; X32-NEXT: vpsrld $16, %ymm0, %ymm0 -; X32-NEXT: vpackusdw %ymm3, %ymm0, %ymm0 +; X32-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 ; X32-NEXT: retl +; X32-NEXT: ## -- End function ; ; X64-LABEL: ashr_16i16: ; X64: ## BB#0: @@ -540,13 +551,14 @@ define <16 x i16> @ashr_16i16(<16 x i16> %r, <16 x i16> %a) nounwind { ; X64-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15] ; X64-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15] ; X64-NEXT: vpsravd %ymm3, %ymm4, %ymm3 -; X64-NEXT: vpsrld $16, %ymm3, %ymm3 ; X64-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11] ; X64-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11] ; X64-NEXT: vpsravd %ymm1, %ymm0, %ymm0 +; X64-NEXT: vpsrld $16, %ymm3, %ymm1 ; X64-NEXT: vpsrld $16, %ymm0, %ymm0 -; X64-NEXT: vpackusdw %ymm3, %ymm0, %ymm0 +; X64-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 ; X64-NEXT: retq +; X64-NEXT: ## -- End function %ashr = ashr <16 x i16> %r, %a ret <16 x i16> %ashr } @@ -579,6 +591,7 @@ define <32 x i8> @ashr_32i8(<32 x i8> %r, <32 x i8> %a) nounwind { ; X32-NEXT: vpsrlw $8, %ymm0, %ymm0 ; X32-NEXT: vpackuswb %ymm2, %ymm0, %ymm0 ; X32-NEXT: retl +; X32-NEXT: ## -- End function ; ; X64-LABEL: ashr_32i8: ; X64: ## BB#0: @@ -607,6 +620,7 @@ define <32 x i8> @ashr_32i8(<32 x i8> %r, <32 x i8> %a) nounwind { ; X64-NEXT: vpsrlw $8, %ymm0, %ymm0 ; X64-NEXT: vpackuswb %ymm2, %ymm0, %ymm0 ; X64-NEXT: retq +; X64-NEXT: ## -- End function %ashr = ashr <32 x i8> %r, %a ret <32 x i8> %ashr } @@ -622,6 +636,7 @@ define <8 x i16> @lshr_8i16(<8 x i16> %r, <8 x i16> %a) nounwind { ; X32-NEXT: ## kill: %XMM0<def> %XMM0<kill> %YMM0<kill> ; X32-NEXT: vzeroupper ; X32-NEXT: retl +; X32-NEXT: ## -- End function ; ; X64-LABEL: lshr_8i16: ; X64: ## BB#0: @@ -633,6 +648,7 @@ define <8 x i16> @lshr_8i16(<8 x i16> %r, <8 x i16> %a) nounwind { ; X64-NEXT: ## kill: %XMM0<def> %XMM0<kill> %YMM0<kill> ; X64-NEXT: vzeroupper ; X64-NEXT: retq +; X64-NEXT: ## -- End function %lshr = lshr <8 x i16> %r, %a ret <8 x i16> %lshr } @@ -644,13 +660,14 @@ define <16 x i16> @lshr_16i16(<16 x i16> %r, <16 x i16> %a) nounwind { ; X32-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15] ; X32-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15] ; X32-NEXT: vpsrlvd %ymm3, %ymm4, %ymm3 -; X32-NEXT: vpsrld $16, %ymm3, %ymm3 ; X32-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11] ; X32-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11] ; X32-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0 +; X32-NEXT: vpsrld $16, %ymm3, %ymm1 ; X32-NEXT: vpsrld $16, %ymm0, %ymm0 -; X32-NEXT: vpackusdw %ymm3, %ymm0, %ymm0 +; X32-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 ; X32-NEXT: retl +; X32-NEXT: ## -- End function ; ; X64-LABEL: lshr_16i16: ; X64: ## BB#0: @@ -658,13 +675,14 @@ define <16 x i16> @lshr_16i16(<16 x i16> %r, <16 x i16> %a) nounwind { ; X64-NEXT: vpunpckhwd {{.*#+}} ymm3 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15] ; X64-NEXT: vpunpckhwd {{.*#+}} ymm4 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15] ; X64-NEXT: vpsrlvd %ymm3, %ymm4, %ymm3 -; X64-NEXT: vpsrld $16, %ymm3, %ymm3 ; X64-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[8],ymm2[8],ymm1[9],ymm2[9],ymm1[10],ymm2[10],ymm1[11],ymm2[11] ; X64-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[8],ymm0[8],ymm2[9],ymm0[9],ymm2[10],ymm0[10],ymm2[11],ymm0[11] ; X64-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0 +; X64-NEXT: vpsrld $16, %ymm3, %ymm1 ; X64-NEXT: vpsrld $16, %ymm0, %ymm0 -; X64-NEXT: vpackusdw %ymm3, %ymm0, %ymm0 +; X64-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 ; X64-NEXT: retq +; X64-NEXT: ## -- End function %lshr = lshr <16 x i16> %r, %a ret <16 x i16> %lshr } @@ -685,6 +703,7 @@ define <32 x i8> @lshr_32i8(<32 x i8> %r, <32 x i8> %a) nounwind { ; X32-NEXT: vpaddb %ymm1, %ymm1, %ymm1 ; X32-NEXT: vpblendvb %ymm1, %ymm2, %ymm0, %ymm0 ; X32-NEXT: retl +; X32-NEXT: ## -- End function ; ; X64-LABEL: lshr_32i8: ; X64: ## BB#0: @@ -701,6 +720,7 @@ define <32 x i8> @lshr_32i8(<32 x i8> %r, <32 x i8> %a) nounwind { ; X64-NEXT: vpaddb %ymm1, %ymm1, %ymm1 ; X64-NEXT: vpblendvb %ymm1, %ymm2, %ymm0, %ymm0 ; X64-NEXT: retq +; X64-NEXT: ## -- End function %lshr = lshr <32 x i8> %r, %a ret <32 x i8> %lshr } diff --git a/llvm/test/CodeGen/X86/avx512-cmp.ll b/llvm/test/CodeGen/X86/avx512-cmp.ll index eae7b94f513..d2e95f692e4 100644 --- a/llvm/test/CodeGen/X86/avx512-cmp.ll +++ b/llvm/test/CodeGen/X86/avx512-cmp.ll @@ -14,6 +14,7 @@ define double @test1(double %a, double %b) nounwind { ; ALL-NEXT: LBB0_2: ## %l2 ; ALL-NEXT: vaddsd %xmm1, %xmm0, %xmm0 ; ALL-NEXT: retq +; ALL-NEXT: ## -- End function %tobool = fcmp une double %a, %b br i1 %tobool, label %l1, label %l2 @@ -36,6 +37,7 @@ define float @test2(float %a, float %b) nounwind { ; ALL-NEXT: LBB1_2: ## %l2 ; ALL-NEXT: vaddss %xmm1, %xmm0, %xmm0 ; ALL-NEXT: retq +; ALL-NEXT: ## -- End function %tobool = fcmp olt float %a, %b br i1 %tobool, label %l1, label %l2 @@ -124,11 +126,11 @@ entry: define i32 @test8(i32 %a1, i32 %a2, i32 %a3) { ; ALL-LABEL: test8: ; ALL: ## BB#0: -; ALL-NEXT: notl %edi ; ALL-NEXT: xorl $-2147483648, %esi ## imm = 0x80000000 ; ALL-NEXT: testl %edx, %edx ; ALL-NEXT: movl $1, %eax ; ALL-NEXT: cmovel %eax, %edx +; ALL-NEXT: notl %edi ; ALL-NEXT: orl %edi, %esi ; ALL-NEXT: cmovnel %edx, %eax ; ALL-NEXT: retq diff --git a/llvm/test/CodeGen/X86/avx512-cvt.ll b/llvm/test/CodeGen/X86/avx512-cvt.ll index 140299f5495..295f98ce61b 100644 --- a/llvm/test/CodeGen/X86/avx512-cvt.ll +++ b/llvm/test/CodeGen/X86/avx512-cvt.ll @@ -1545,19 +1545,19 @@ define <4 x double> @uitofp_4i1_double(<4 x i32> %a) { } define <2 x float> @uitofp_2i1_float(<2 x i32> %a) { -; NOVL-LABEL: uitofp_2i1_float: -; NOVL: # BB#0: -; NOVL-NEXT: vpxor %xmm1, %xmm1, %xmm1 -; NOVL-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3] -; NOVL-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm0 -; NOVL-NEXT: vpextrb $8, %xmm0, %eax -; NOVL-NEXT: andl $1, %eax -; NOVL-NEXT: vcvtsi2ssl %eax, %xmm2, %xmm1 -; NOVL-NEXT: vpextrb $0, %xmm0, %eax -; NOVL-NEXT: andl $1, %eax -; NOVL-NEXT: vcvtsi2ssl %eax, %xmm2, %xmm0 -; NOVL-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[2,3] -; NOVL-NEXT: retq +; KNL-LABEL: uitofp_2i1_float: +; KNL: # BB#0: +; KNL-NEXT: vpxor %xmm1, %xmm1, %xmm1 +; KNL-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3] +; KNL-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm0 +; KNL-NEXT: vpextrb $8, %xmm0, %eax +; KNL-NEXT: andl $1, %eax +; KNL-NEXT: vpextrb $0, %xmm0, %ecx +; KNL-NEXT: vcvtsi2ssl %eax, %xmm2, %xmm0 +; KNL-NEXT: andl $1, %ecx +; KNL-NEXT: vcvtsi2ssl %ecx, %xmm2, %xmm1 +; KNL-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3] +; KNL-NEXT: retq ; ; VL-LABEL: uitofp_2i1_float: ; VL: # BB#0: @@ -1567,6 +1567,34 @@ define <2 x float> @uitofp_2i1_float(<2 x i32> %a) { ; VL-NEXT: vpbroadcastd {{.*}}(%rip), %xmm0 {%k1} {z} ; VL-NEXT: vcvtudq2ps %xmm0, %xmm0 ; VL-NEXT: retq +; +; AVX512DQ-LABEL: uitofp_2i1_float: +; AVX512DQ: # BB#0: +; AVX512DQ-NEXT: vpxor %xmm1, %xmm1, %xmm1 +; AVX512DQ-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3] +; AVX512DQ-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm0 +; AVX512DQ-NEXT: vpextrb $8, %xmm0, %eax +; AVX512DQ-NEXT: andl $1, %eax +; AVX512DQ-NEXT: vcvtsi2ssl %eax, %xmm2, %xmm1 +; AVX512DQ-NEXT: vpextrb $0, %xmm0, %eax +; AVX512DQ-NEXT: andl $1, %eax +; AVX512DQ-NEXT: vcvtsi2ssl %eax, %xmm2, %xmm0 +; AVX512DQ-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[2,3] +; AVX512DQ-NEXT: retq +; +; AVX512BW-LABEL: uitofp_2i1_float: +; AVX512BW: # BB#0: +; AVX512BW-NEXT: vpxor %xmm1, %xmm1, %xmm1 +; AVX512BW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3] +; AVX512BW-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm0 +; AVX512BW-NEXT: vpextrb $8, %xmm0, %eax +; AVX512BW-NEXT: andl $1, %eax +; AVX512BW-NEXT: vcvtsi2ssl %eax, %xmm2, %xmm1 +; AVX512BW-NEXT: vpextrb $0, %xmm0, %eax +; AVX512BW-NEXT: andl $1, %eax +; AVX512BW-NEXT: vcvtsi2ssl %eax, %xmm2, %xmm0 +; AVX512BW-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[2,3] +; AVX512BW-NEXT: retq %mask = icmp ult <2 x i32> %a, zeroinitializer %1 = uitofp <2 x i1> %mask to <2 x float> ret <2 x float> %1 diff --git a/llvm/test/CodeGen/X86/avx512-insert-extract.ll b/llvm/test/CodeGen/X86/avx512-insert-extract.ll index 29a5325a0ae..459bd3dabfc 100644 --- a/llvm/test/CodeGen/X86/avx512-insert-extract.ll +++ b/llvm/test/CodeGen/X86/avx512-insert-extract.ll @@ -12,6 +12,7 @@ define <16 x float> @test1(<16 x float> %x, float* %br, float %y) nounwind { ; KNL-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0],xmm0[3] ; KNL-NEXT: vinsertf32x4 $3, %xmm0, %zmm2, %zmm0 ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test1: ; SKX: ## BB#0: @@ -21,6 +22,7 @@ define <16 x float> @test1(<16 x float> %x, float* %br, float %y) nounwind { ; SKX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0],xmm0[3] ; SKX-NEXT: vinsertf32x4 $3, %xmm0, %zmm2, %zmm0 ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %rrr = load float, float* %br %rrr2 = insertelement <16 x float> %x, float %rrr, i32 1 %rrr3 = insertelement <16 x float> %rrr2, float %y, i32 14 @@ -36,6 +38,7 @@ define <8 x double> @test2(<8 x double> %x, double* %br, double %y) nounwind { ; KNL-NEXT: vmovsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; KNL-NEXT: vinsertf32x4 $3, %xmm0, %zmm2, %zmm0 ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test2: ; SKX: ## BB#0: @@ -45,6 +48,7 @@ define <8 x double> @test2(<8 x double> %x, double* %br, double %y) nounwind { ; SKX-NEXT: vmovsd {{.*#+}} xmm0 = xmm1[0],xmm0[1] ; SKX-NEXT: vinsertf64x2 $3, %xmm0, %zmm2, %zmm0 ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %rrr = load double, double* %br %rrr2 = insertelement <8 x double> %x, double %rrr, i32 1 %rrr3 = insertelement <8 x double> %rrr2, double %y, i32 6 @@ -58,6 +62,7 @@ define <16 x float> @test3(<16 x float> %x) nounwind { ; KNL-NEXT: vinsertps {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[2,3] ; KNL-NEXT: vinsertf32x4 $0, %xmm1, %zmm0, %zmm0 ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test3: ; SKX: ## BB#0: @@ -65,6 +70,7 @@ define <16 x float> @test3(<16 x float> %x) nounwind { ; SKX-NEXT: vinsertps {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[2,3] ; SKX-NEXT: vinsertf32x4 $0, %xmm1, %zmm0, %zmm0 ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %eee = extractelement <16 x float> %x, i32 4 %rrr2 = insertelement <16 x float> %x, float %eee, i32 1 ret <16 x float> %rrr2 @@ -78,6 +84,7 @@ define <8 x i64> @test4(<8 x i64> %x) nounwind { ; KNL-NEXT: vpinsrq $1, %rax, %xmm0, %xmm1 ; KNL-NEXT: vinserti32x4 $0, %xmm1, %zmm0, %zmm0 ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test4: ; SKX: ## BB#0: @@ -86,6 +93,7 @@ define <8 x i64> @test4(<8 x i64> %x) nounwind { ; SKX-NEXT: vpinsrq $1, %rax, %xmm0, %xmm1 ; SKX-NEXT: vinserti64x2 $0, %xmm1, %zmm0, %zmm0 ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %eee = extractelement <8 x i64> %x, i32 4 %rrr2 = insertelement <8 x i64> %x, i64 %eee, i32 1 ret <8 x i64> %rrr2 @@ -96,11 +104,13 @@ define i32 @test5(<4 x float> %x) nounwind { ; KNL: ## BB#0: ; KNL-NEXT: vextractps $3, %xmm0, %eax ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test5: ; SKX: ## BB#0: ; SKX-NEXT: vextractps $3, %xmm0, %eax ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %ef = extractelement <4 x float> %x, i32 3 %ei = bitcast float %ef to i32 ret i32 %ei @@ -111,11 +121,13 @@ define void @test6(<4 x float> %x, float* %out) nounwind { ; KNL: ## BB#0: ; KNL-NEXT: vextractps $3, %xmm0, (%rdi) ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test6: ; SKX: ## BB#0: ; SKX-NEXT: vextractps $3, %xmm0, (%rdi) ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %ef = extractelement <4 x float> %x, i32 3 store float %ef, float* %out, align 4 ret void @@ -135,6 +147,7 @@ define float @test7(<16 x float> %x, i32 %ind) nounwind { ; KNL-NEXT: movq %rbp, %rsp ; KNL-NEXT: popq %rbp ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test7: ; SKX: ## BB#0: @@ -150,6 +163,7 @@ define float @test7(<16 x float> %x, i32 %ind) nounwind { ; SKX-NEXT: popq %rbp ; SKX-NEXT: vzeroupper ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %e = extractelement <16 x float> %x, i32 %ind ret float %e } @@ -168,6 +182,7 @@ define double @test8(<8 x double> %x, i32 %ind) nounwind { ; KNL-NEXT: movq %rbp, %rsp ; KNL-NEXT: popq %rbp ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test8: ; SKX: ## BB#0: @@ -183,6 +198,7 @@ define double @test8(<8 x double> %x, i32 %ind) nounwind { ; SKX-NEXT: popq %rbp ; SKX-NEXT: vzeroupper ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %e = extractelement <8 x double> %x, i32 %ind ret double %e } @@ -201,6 +217,7 @@ define float @test9(<8 x float> %x, i32 %ind) nounwind { ; KNL-NEXT: movq %rbp, %rsp ; KNL-NEXT: popq %rbp ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test9: ; SKX: ## BB#0: @@ -216,6 +233,7 @@ define float @test9(<8 x float> %x, i32 %ind) nounwind { ; SKX-NEXT: popq %rbp ; SKX-NEXT: vzeroupper ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %e = extractelement <8 x float> %x, i32 %ind ret float %e } @@ -234,6 +252,7 @@ define i32 @test10(<16 x i32> %x, i32 %ind) nounwind { ; KNL-NEXT: movq %rbp, %rsp ; KNL-NEXT: popq %rbp ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test10: ; SKX: ## BB#0: @@ -249,6 +268,7 @@ define i32 @test10(<16 x i32> %x, i32 %ind) nounwind { ; SKX-NEXT: popq %rbp ; SKX-NEXT: vzeroupper ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %e = extractelement <16 x i32> %x, i32 %ind ret i32 %e } @@ -1114,137 +1134,137 @@ define i32 @test_insertelement_v32i1(i32 %a, i32 %b, <32 x i32> %x , <32 x i32> ; KNL-NEXT: .cfi_def_cfa_register %rbp ; KNL-NEXT: andq $-32, %rsp ; KNL-NEXT: subq $32, %rsp -; KNL-NEXT: xorl %eax, %eax -; KNL-NEXT: cmpl %esi, %edi -; KNL-NEXT: setb %al ; KNL-NEXT: vpcmpltud %zmm3, %zmm1, %k0 ; KNL-NEXT: kshiftlw $14, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx +; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $15, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %edx -; KNL-NEXT: vmovd %edx, %xmm1 -; KNL-NEXT: vpinsrb $1, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %ecx +; KNL-NEXT: vmovd %ecx, %xmm1 +; KNL-NEXT: vpinsrb $1, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $13, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $2, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $2, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $12, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $3, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $3, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $11, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $4, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $4, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $10, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $5, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $5, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $9, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $6, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $6, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $8, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $7, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $7, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $7, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $8, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $8, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $6, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $9, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $9, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $5, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $10, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $10, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $4, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $11, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $11, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $3, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $12, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $12, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $2, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $13, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $13, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $1, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $14, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $14, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %ecx -; KNL-NEXT: vpinsrb $15, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vpinsrb $15, %eax, %xmm1, %xmm1 ; KNL-NEXT: vpcmpltud %zmm2, %zmm0, %k0 -; KNL-NEXT: kshiftlw $14, %k0, %k1 -; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx ; KNL-NEXT: kshiftlw $15, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %edx -; KNL-NEXT: vmovd %edx, %xmm0 -; KNL-NEXT: vpinsrb $1, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: kshiftlw $14, %k0, %k1 +; KNL-NEXT: kshiftrw $15, %k1, %k1 +; KNL-NEXT: vmovd %eax, %xmm0 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 ; KNL-NEXT: kshiftlw $13, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $2, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 ; KNL-NEXT: kshiftlw $12, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $3, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $3, %eax, %xmm0, %xmm0 ; KNL-NEXT: kshiftlw $11, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $4, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $4, %eax, %xmm0, %xmm0 ; KNL-NEXT: kshiftlw $10, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $5, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 ; KNL-NEXT: kshiftlw $9, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $6, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 ; KNL-NEXT: kshiftlw $8, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $7, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 ; KNL-NEXT: kshiftlw $7, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $8, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $8, %eax, %xmm0, %xmm0 ; KNL-NEXT: kshiftlw $6, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $9, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $9, %eax, %xmm0, %xmm0 ; KNL-NEXT: kshiftlw $5, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $10, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $10, %eax, %xmm0, %xmm0 ; KNL-NEXT: kshiftlw $4, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $11, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $11, %eax, %xmm0, %xmm0 ; KNL-NEXT: kshiftlw $3, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $12, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $12, %eax, %xmm0, %xmm0 ; KNL-NEXT: kshiftlw $2, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $13, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $13, %eax, %xmm0, %xmm0 ; KNL-NEXT: kshiftlw $1, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vpinsrb $14, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $14, %eax, %xmm0, %xmm0 +; KNL-NEXT: xorl %eax, %eax +; KNL-NEXT: cmpl %esi, %edi ; KNL-NEXT: kshiftrw $15, %k0, %k0 ; KNL-NEXT: kmovw %k0, %ecx ; KNL-NEXT: vpinsrb $15, %ecx, %xmm0, %xmm0 +; KNL-NEXT: setb %al ; KNL-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 ; KNL-NEXT: vpsllw $7, %ymm0, %ymm0 ; KNL-NEXT: vpand {{.*}}(%rip), %ymm0, %ymm0 @@ -1299,8 +1319,8 @@ define i8 @test_iinsertelement_v4i1(i32 %a, i32 %b, <4 x i32> %x , <4 x i32> %y) ; KNL-NEXT: vpcmpgtd %xmm0, %xmm1, %xmm0 ; KNL-NEXT: vpextrb $4, %xmm0, %ecx ; KNL-NEXT: kmovw %ecx, %k1 -; KNL-NEXT: vpternlogq $255, %zmm1, %zmm1, %zmm1 {%k1} {z} ; KNL-NEXT: vpextrb $0, %xmm0, %ecx +; KNL-NEXT: vpternlogq $255, %zmm1, %zmm1, %zmm1 {%k1} {z} ; KNL-NEXT: kmovw %ecx, %k1 ; KNL-NEXT: vpternlogq $255, %zmm2, %zmm2, %zmm2 {%k1} {z} ; KNL-NEXT: vmovdqa64 {{.*#+}} zmm3 = [0,8,2,3,4,5,6,7] @@ -2124,8 +2144,8 @@ define i16 @test_extractelement_variable_v32i16(<32 x i16> %t1, i32 %index) { define i8 @test_extractelement_variable_v16i8(<16 x i8> %t1, i32 %index) { ; KNL-LABEL: test_extractelement_variable_v16i8: ; KNL: ## BB#0: -; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def> ; KNL-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp) +; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def> ; KNL-NEXT: andl $15, %edi ; KNL-NEXT: leaq -{{[0-9]+}}(%rsp), %rax ; KNL-NEXT: movb (%rdi,%rax), %al @@ -2156,8 +2176,8 @@ define i8 @test_extractelement_variable_v32i8(<32 x i8> %t1, i32 %index) { ; KNL-NEXT: .cfi_def_cfa_register %rbp ; KNL-NEXT: andq $-32, %rsp ; KNL-NEXT: subq $64, %rsp -; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def> ; KNL-NEXT: vmovaps %ymm0, (%rsp) +; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def> ; KNL-NEXT: andl $31, %edi ; KNL-NEXT: movq %rsp, %rax ; KNL-NEXT: movb (%rdi,%rax), %al @@ -2204,9 +2224,9 @@ define i8 @test_extractelement_variable_v64i8(<64 x i8> %t1, i32 %index) { ; KNL-NEXT: .cfi_def_cfa_register %rbp ; KNL-NEXT: andq $-64, %rsp ; KNL-NEXT: subq $128, %rsp -; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def> ; KNL-NEXT: vmovaps %ymm1, {{[0-9]+}}(%rsp) ; KNL-NEXT: vmovaps %ymm0, (%rsp) +; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def> ; KNL-NEXT: andl $63, %edi ; KNL-NEXT: movq %rsp, %rax ; KNL-NEXT: movb (%rdi,%rax), %al @@ -2295,12 +2315,12 @@ define i8 @test_extractelement_variable_v64i8_indexi8(<64 x i8> %t1, i8 %index) define zeroext i8 @test_extractelement_varible_v2i1(<2 x i64> %a, <2 x i64> %b, i32 %index) { ; KNL-LABEL: test_extractelement_varible_v2i1: ; KNL: ## BB#0: -; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def> ; KNL-NEXT: vmovdqa {{.*#+}} xmm2 = [9223372036854775808,9223372036854775808] ; KNL-NEXT: vpxor %xmm2, %xmm1, %xmm1 ; KNL-NEXT: vpxor %xmm2, %xmm0, %xmm0 ; KNL-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm0 ; KNL-NEXT: vmovdqa %xmm0, -{{[0-9]+}}(%rsp) +; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def> ; KNL-NEXT: andl $1, %edi ; KNL-NEXT: movl -24(%rsp,%rdi,8), %eax ; KNL-NEXT: andl $1, %eax @@ -2325,12 +2345,12 @@ define zeroext i8 @test_extractelement_varible_v2i1(<2 x i64> %a, <2 x i64> %b, define zeroext i8 @test_extractelement_varible_v4i1(<4 x i32> %a, <4 x i32> %b, i32 %index) { ; KNL-LABEL: test_extractelement_varible_v4i1: ; KNL: ## BB#0: -; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def> ; KNL-NEXT: vpbroadcastd {{.*}}(%rip), %xmm2 ; KNL-NEXT: vpxor %xmm2, %xmm1, %xmm1 ; KNL-NEXT: vpxor %xmm2, %xmm0, %xmm0 ; KNL-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm0 ; KNL-NEXT: vmovdqa %xmm0, -{{[0-9]+}}(%rsp) +; KNL-NEXT: ## kill: %EDI<def> %EDI<kill> %RDI<def> ; KNL-NEXT: andl $3, %edi ; KNL-NEXT: movl -24(%rsp,%rdi,4), %eax ; KNL-NEXT: andl $1, %eax diff --git a/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll b/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll index 86902ac926a..3c2e6afd225 100644 --- a/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll +++ b/llvm/test/CodeGen/X86/avx512-intrinsics-upgrade.ll @@ -2880,7 +2880,6 @@ declare <8 x i64> @llvm.x86.avx512.mask.pmulu.dq.512(<16 x i32>, <16 x i32>, <8 define <4 x float> @test_mask_vextractf32x4(<4 x float> %b, <16 x float> %a, i8 %mask) { ; CHECK-LABEL: test_mask_vextractf32x4: ; CHECK: ## BB#0: -; CHECK-NEXT: vextractf32x4 $2, %zmm1, %xmm1 ; CHECK-NEXT: kmovw %edi, %k0 ; CHECK-NEXT: kshiftlw $12, %k0, %k1 ; CHECK-NEXT: kshiftrw $15, %k1, %k1 @@ -2898,6 +2897,7 @@ define <4 x float> @test_mask_vextractf32x4(<4 x float> %b, <16 x float> %a, i8 ; CHECK-NEXT: vpinsrb $8, %eax, %xmm2, %xmm2 ; CHECK-NEXT: kmovw %k1, %eax ; CHECK-NEXT: vpinsrb $12, %eax, %xmm2, %xmm2 +; CHECK-NEXT: vextractf32x4 $2, %zmm1, %xmm1 ; CHECK-NEXT: vpslld $31, %xmm2, %xmm2 ; CHECK-NEXT: vblendvps %xmm2, %xmm1, %xmm0, %xmm0 ; CHECK-NEXT: retq @@ -2941,7 +2941,6 @@ declare <4 x i64> @llvm.x86.avx512.mask.vextracti64x4.512(<8 x i64>, i32, <4 x i define <4 x i32> @test_maskz_vextracti32x4(<16 x i32> %a, i8 %mask) { ; CHECK-LABEL: test_maskz_vextracti32x4: ; CHECK: ## BB#0: -; CHECK-NEXT: vextracti32x4 $2, %zmm0, %xmm0 ; CHECK-NEXT: kmovw %edi, %k0 ; CHECK-NEXT: kshiftlw $12, %k0, %k1 ; CHECK-NEXT: kshiftrw $15, %k1, %k1 @@ -2959,6 +2958,7 @@ define <4 x i32> @test_maskz_vextracti32x4(<16 x i32> %a, i8 %mask) { ; CHECK-NEXT: vpinsrb $8, %eax, %xmm1, %xmm1 ; CHECK-NEXT: kmovw %k1, %eax ; CHECK-NEXT: vpinsrb $12, %eax, %xmm1, %xmm1 +; CHECK-NEXT: vextracti32x4 $2, %zmm0, %xmm0 ; CHECK-NEXT: vpslld $31, %xmm1, %xmm1 ; CHECK-NEXT: vpsrad $31, %xmm1, %xmm1 ; CHECK-NEXT: vpand %xmm0, %xmm1, %xmm0 diff --git a/llvm/test/CodeGen/X86/avx512-mask-op.ll b/llvm/test/CodeGen/X86/avx512-mask-op.ll index e1a92c60d18..0ee86d5fb45 100644 --- a/llvm/test/CodeGen/X86/avx512-mask-op.ll +++ b/llvm/test/CodeGen/X86/avx512-mask-op.ll @@ -1837,73 +1837,8 @@ define void @ktest_2(<32 x float> %in, float * %base) { ; KNL-NEXT: .cfi_def_cfa_register %rbp ; KNL-NEXT: andq $-32, %rsp ; KNL-NEXT: subq $32, %rsp -; KNL-NEXT: vmovups (%rdi), %zmm2 -; KNL-NEXT: vmovups 64(%rdi), %zmm3 -; KNL-NEXT: vcmpltps %zmm1, %zmm3, %k1 -; KNL-NEXT: kshiftlw $14, %k1, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: kshiftlw $15, %k1, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %ecx -; KNL-NEXT: vmovd %ecx, %xmm3 -; KNL-NEXT: vpinsrb $1, %eax, %xmm3, %xmm3 -; KNL-NEXT: kshiftlw $13, %k1, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $2, %eax, %xmm3, %xmm3 -; KNL-NEXT: kshiftlw $12, %k1, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $3, %eax, %xmm3, %xmm3 -; KNL-NEXT: kshiftlw $11, %k1, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $4, %eax, %xmm3, %xmm3 -; KNL-NEXT: kshiftlw $10, %k1, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $5, %eax, %xmm3, %xmm3 -; KNL-NEXT: kshiftlw $9, %k1, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $6, %eax, %xmm3, %xmm3 -; KNL-NEXT: kshiftlw $8, %k1, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $7, %eax, %xmm3, %xmm3 -; KNL-NEXT: kshiftlw $7, %k1, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $8, %eax, %xmm3, %xmm3 -; KNL-NEXT: kshiftlw $6, %k1, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $9, %eax, %xmm3, %xmm3 -; KNL-NEXT: kshiftlw $5, %k1, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $10, %eax, %xmm3, %xmm3 -; KNL-NEXT: kshiftlw $4, %k1, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $11, %eax, %xmm3, %xmm3 -; KNL-NEXT: kshiftlw $3, %k1, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $12, %eax, %xmm3, %xmm3 -; KNL-NEXT: kshiftlw $2, %k1, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $13, %eax, %xmm3, %xmm3 -; KNL-NEXT: kshiftlw $1, %k1, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $14, %eax, %xmm3, %xmm3 -; KNL-NEXT: kshiftrw $15, %k1, %k0 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $15, %eax, %xmm3, %xmm3 -; KNL-NEXT: vcmpltps %zmm0, %zmm2, %k2 +; KNL-NEXT: vmovups 64(%rdi), %zmm2 +; KNL-NEXT: vcmpltps %zmm1, %zmm2, %k2 ; KNL-NEXT: kshiftlw $14, %k2, %k0 ; KNL-NEXT: kshiftrw $15, %k0, %k0 ; KNL-NEXT: kmovw %k0, %eax @@ -1967,138 +1902,203 @@ define void @ktest_2(<32 x float> %in, float * %base) { ; KNL-NEXT: kshiftrw $15, %k2, %k0 ; KNL-NEXT: kmovw %k0, %eax ; KNL-NEXT: vpinsrb $15, %eax, %xmm2, %xmm2 -; KNL-NEXT: vinserti128 $1, %xmm3, %ymm2, %ymm2 -; KNL-NEXT: vmovups 4(%rdi), %zmm3 {%k2} {z} -; KNL-NEXT: vmovups 68(%rdi), %zmm4 {%k1} {z} +; KNL-NEXT: vmovups (%rdi), %zmm3 +; KNL-NEXT: vcmpltps %zmm0, %zmm3, %k1 +; KNL-NEXT: kshiftlw $14, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: kshiftlw $15, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %ecx +; KNL-NEXT: vmovd %ecx, %xmm3 +; KNL-NEXT: vpinsrb $1, %eax, %xmm3, %xmm3 +; KNL-NEXT: kshiftlw $13, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vpinsrb $2, %eax, %xmm3, %xmm3 +; KNL-NEXT: kshiftlw $12, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vpinsrb $3, %eax, %xmm3, %xmm3 +; KNL-NEXT: kshiftlw $11, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vpinsrb $4, %eax, %xmm3, %xmm3 +; KNL-NEXT: kshiftlw $10, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vpinsrb $5, %eax, %xmm3, %xmm3 +; KNL-NEXT: kshiftlw $9, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vpinsrb $6, %eax, %xmm3, %xmm3 +; KNL-NEXT: kshiftlw $8, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vpinsrb $7, %eax, %xmm3, %xmm3 +; KNL-NEXT: kshiftlw $7, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vpinsrb $8, %eax, %xmm3, %xmm3 +; KNL-NEXT: kshiftlw $6, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vpinsrb $9, %eax, %xmm3, %xmm3 +; KNL-NEXT: kshiftlw $5, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vpinsrb $10, %eax, %xmm3, %xmm3 +; KNL-NEXT: kshiftlw $4, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vpinsrb $11, %eax, %xmm3, %xmm3 +; KNL-NEXT: kshiftlw $3, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vpinsrb $12, %eax, %xmm3, %xmm3 +; KNL-NEXT: kshiftlw $2, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vpinsrb $13, %eax, %xmm3, %xmm3 +; KNL-NEXT: kshiftlw $1, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vpinsrb $14, %eax, %xmm3, %xmm3 +; KNL-NEXT: kshiftrw $15, %k1, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vpinsrb $15, %eax, %xmm3, %xmm3 +; KNL-NEXT: vmovups 68(%rdi), %zmm4 {%k2} {z} ; KNL-NEXT: vcmpltps %zmm4, %zmm1, %k0 -; KNL-NEXT: kshiftlw $14, %k0, %k1 -; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: kshiftlw $15, %k0, %k1 -; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx +; KNL-NEXT: kshiftlw $14, %k0, %k2 +; KNL-NEXT: kshiftrw $15, %k2, %k2 +; KNL-NEXT: kmovw %k2, %eax +; KNL-NEXT: kshiftlw $15, %k0, %k2 +; KNL-NEXT: kshiftrw $15, %k2, %k2 +; KNL-NEXT: kmovw %k2, %ecx ; KNL-NEXT: vmovd %ecx, %xmm4 ; KNL-NEXT: vpinsrb $1, %eax, %xmm4, %xmm4 -; KNL-NEXT: kshiftlw $13, %k0, %k1 -; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: kshiftlw $13, %k0, %k2 +; KNL-NEXT: kshiftrw $15, %k2, %k2 +; KNL-NEXT: kmovw %k2, %eax ; KNL-NEXT: vpinsrb $2, %eax, %xmm4, %xmm4 -; KNL-NEXT: kshiftlw $12, %k0, %k1 -; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: kshiftlw $12, %k0, %k2 +; KNL-NEXT: kshiftrw $15, %k2, %k2 +; KNL-NEXT: kmovw %k2, %eax ; KNL-NEXT: vpinsrb $3, %eax, %xmm4, %xmm4 -; KNL-NEXT: kshiftlw $11, %k0, %k1 -; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: kshiftlw $11, %k0, %k2 +; KNL-NEXT: kshiftrw $15, %k2, %k2 +; KNL-NEXT: kmovw %k2, %eax ; KNL-NEXT: vpinsrb $4, %eax, %xmm4, %xmm4 -; KNL-NEXT: kshiftlw $10, %k0, %k1 -; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: kshiftlw $10, %k0, %k2 +; KNL-NEXT: kshiftrw $15, %k2, %k2 +; KNL-NEXT: kmovw %k2, %eax ; KNL-NEXT: vpinsrb $5, %eax, %xmm4, %xmm4 -; KNL-NEXT: kshiftlw $9, %k0, %k1 -; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: kshiftlw $9, %k0, %k2 +; KNL-NEXT: kshiftrw $15, %k2, %k2 +; KNL-NEXT: kmovw %k2, %eax ; KNL-NEXT: vpinsrb $6, %eax, %xmm4, %xmm4 -; KNL-NEXT: kshiftlw $8, %k0, %k1 -; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: kshiftlw $8, %k0, %k2 +; KNL-NEXT: kshiftrw $15, %k2, %k2 +; KNL-NEXT: kmovw %k2, %eax ; KNL-NEXT: vpinsrb $7, %eax, %xmm4, %xmm4 -; KNL-NEXT: kshiftlw $7, %k0, %k1 -; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: kshiftlw $7, %k0, %k2 +; KNL-NEXT: kshiftrw $15, %k2, %k2 +; KNL-NEXT: kmovw %k2, %eax ; KNL-NEXT: vpinsrb $8, %eax, %xmm4, %xmm4 -; KNL-NEXT: kshiftlw $6, %k0, %k1 -; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: kshiftlw $6, %k0, %k2 +; KNL-NEXT: kshiftrw $15, %k2, %k2 +; KNL-NEXT: kmovw %k2, %eax ; KNL-NEXT: vpinsrb $9, %eax, %xmm4, %xmm4 -; KNL-NEXT: kshiftlw $5, %k0, %k1 -; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: kshiftlw $5, %k0, %k2 +; KNL-NEXT: kshiftrw $15, %k2, %k2 +; KNL-NEXT: kmovw %k2, %eax ; KNL-NEXT: vpinsrb $10, %eax, %xmm4, %xmm4 -; KNL-NEXT: kshiftlw $4, %k0, %k1 -; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: kshiftlw $4, %k0, %k2 +; KNL-NEXT: kshiftrw $15, %k2, %k2 +; KNL-NEXT: kmovw %k2, %eax ; KNL-NEXT: vpinsrb $11, %eax, %xmm4, %xmm4 -; KNL-NEXT: kshiftlw $3, %k0, %k1 -; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: kshiftlw $3, %k0, %k2 +; KNL-NEXT: kshiftrw $15, %k2, %k2 +; KNL-NEXT: kmovw %k2, %eax ; KNL-NEXT: vpinsrb $12, %eax, %xmm4, %xmm4 -; KNL-NEXT: kshiftlw $2, %k0, %k1 -; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: kshiftlw $2, %k0, %k2 +; KNL-NEXT: kshiftrw $15, %k2, %k2 +; KNL-NEXT: kmovw %k2, %eax ; KNL-NEXT: vpinsrb $13, %eax, %xmm4, %xmm4 -; KNL-NEXT: kshiftlw $1, %k0, %k1 -; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: kshiftlw $1, %k0, %k2 +; KNL-NEXT: kshiftrw $15, %k2, %k2 +; KNL-NEXT: kmovw %k2, %eax ; KNL-NEXT: vpinsrb $14, %eax, %xmm4, %xmm4 +; KNL-NEXT: vinserti128 $1, %xmm2, %ymm3, %ymm2 ; KNL-NEXT: kshiftrw $15, %k0, %k0 ; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $15, %eax, %xmm4, %xmm4 -; KNL-NEXT: vcmpltps %zmm3, %zmm0, %k0 +; KNL-NEXT: vpinsrb $15, %eax, %xmm4, %xmm3 +; KNL-NEXT: vmovups 4(%rdi), %zmm4 {%k1} {z} +; KNL-NEXT: vcmpltps %zmm4, %zmm0, %k0 ; KNL-NEXT: kshiftlw $14, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $15, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vmovd %ecx, %xmm3 -; KNL-NEXT: vpinsrb $1, %eax, %xmm3, %xmm3 +; KNL-NEXT: vmovd %ecx, %xmm4 +; KNL-NEXT: vpinsrb $1, %eax, %xmm4, %xmm4 ; KNL-NEXT: kshiftlw $13, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $2, %eax, %xmm3, %xmm3 +; KNL-NEXT: vpinsrb $2, %eax, %xmm4, %xmm4 ; KNL-NEXT: kshiftlw $12, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $3, %eax, %xmm3, %xmm3 +; KNL-NEXT: vpinsrb $3, %eax, %xmm4, %xmm4 ; KNL-NEXT: kshiftlw $11, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $4, %eax, %xmm3, %xmm3 +; KNL-NEXT: vpinsrb $4, %eax, %xmm4, %xmm4 ; KNL-NEXT: kshiftlw $10, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $5, %eax, %xmm3, %xmm3 +; KNL-NEXT: vpinsrb $5, %eax, %xmm4, %xmm4 ; KNL-NEXT: kshiftlw $9, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $6, %eax, %xmm3, %xmm3 +; KNL-NEXT: vpinsrb $6, %eax, %xmm4, %xmm4 ; KNL-NEXT: kshiftlw $8, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $7, %eax, %xmm3, %xmm3 +; KNL-NEXT: vpinsrb $7, %eax, %xmm4, %xmm4 ; KNL-NEXT: kshiftlw $7, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $8, %eax, %xmm3, %xmm3 +; KNL-NEXT: vpinsrb $8, %eax, %xmm4, %xmm4 ; KNL-NEXT: kshiftlw $6, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $9, %eax, %xmm3, %xmm3 +; KNL-NEXT: vpinsrb $9, %eax, %xmm4, %xmm4 ; KNL-NEXT: kshiftlw $5, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $10, %eax, %xmm3, %xmm3 +; KNL-NEXT: vpinsrb $10, %eax, %xmm4, %xmm4 ; KNL-NEXT: kshiftlw $4, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $11, %eax, %xmm3, %xmm3 +; KNL-NEXT: vpinsrb $11, %eax, %xmm4, %xmm4 ; KNL-NEXT: kshiftlw $3, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $12, %eax, %xmm3, %xmm3 +; KNL-NEXT: vpinsrb $12, %eax, %xmm4, %xmm4 ; KNL-NEXT: kshiftlw $2, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $13, %eax, %xmm3, %xmm3 +; KNL-NEXT: vpinsrb $13, %eax, %xmm4, %xmm4 ; KNL-NEXT: kshiftlw $1, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $14, %eax, %xmm3, %xmm3 +; KNL-NEXT: vpinsrb $14, %eax, %xmm4, %xmm4 ; KNL-NEXT: kshiftrw $15, %k0, %k0 ; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $15, %eax, %xmm3, %xmm3 -; KNL-NEXT: vinserti128 $1, %xmm4, %ymm3, %ymm3 +; KNL-NEXT: vpinsrb $15, %eax, %xmm4, %xmm4 +; KNL-NEXT: vinserti128 $1, %xmm3, %ymm4, %ymm3 ; KNL-NEXT: vpor %ymm3, %ymm2, %ymm2 ; KNL-NEXT: vextracti128 $1, %ymm2, %xmm3 ; KNL-NEXT: vpmovsxbd %xmm3, %zmm3 @@ -2943,36 +2943,6 @@ define void @store_64i1(<64 x i1>* %a, <64 x i1> %v) { ; ; KNL-LABEL: store_64i1: ; KNL: ## BB#0: -; KNL-NEXT: pushq %rbp -; KNL-NEXT: Lcfi9: -; KNL-NEXT: .cfi_def_cfa_offset 16 -; KNL-NEXT: pushq %r15 -; KNL-NEXT: Lcfi10: -; KNL-NEXT: .cfi_def_cfa_offset 24 -; KNL-NEXT: pushq %r14 -; KNL-NEXT: Lcfi11: -; KNL-NEXT: .cfi_def_cfa_offset 32 -; KNL-NEXT: pushq %r13 -; KNL-NEXT: Lcfi12: -; KNL-NEXT: .cfi_def_cfa_offset 40 -; KNL-NEXT: pushq %r12 -; KNL-NEXT: Lcfi13: -; KNL-NEXT: .cfi_def_cfa_offset 48 -; KNL-NEXT: pushq %rbx -; KNL-NEXT: Lcfi14: -; KNL-NEXT: .cfi_def_cfa_offset 56 -; KNL-NEXT: Lcfi15: -; KNL-NEXT: .cfi_offset %rbx, -56 -; KNL-NEXT: Lcfi16: -; KNL-NEXT: .cfi_offset %r12, -48 -; KNL-NEXT: Lcfi17: -; KNL-NEXT: .cfi_offset %r13, -40 -; KNL-NEXT: Lcfi18: -; KNL-NEXT: .cfi_offset %r14, -32 -; KNL-NEXT: Lcfi19: -; KNL-NEXT: .cfi_offset %r15, -24 -; KNL-NEXT: Lcfi20: -; KNL-NEXT: .cfi_offset %rbp, -16 ; KNL-NEXT: vpmovsxbd %xmm0, %zmm0 ; KNL-NEXT: vpslld $31, %zmm0, %zmm0 ; KNL-NEXT: vpmovsxbd %xmm1, %zmm1 @@ -2984,281 +2954,275 @@ define void @store_64i1(<64 x i1>* %a, <64 x i1> %v) { ; KNL-NEXT: vptestmd %zmm3, %zmm3, %k0 ; KNL-NEXT: kshiftlw $14, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %r8d +; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $15, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %r9d +; KNL-NEXT: kmovw %k1, %ecx ; KNL-NEXT: kshiftlw $13, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %r10d +; KNL-NEXT: vmovd %ecx, %xmm3 +; KNL-NEXT: kmovw %k1, %ecx ; KNL-NEXT: kshiftlw $12, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %r11d +; KNL-NEXT: vpinsrb $1, %eax, %xmm3, %xmm3 +; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $11, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %r14d +; KNL-NEXT: vpinsrb $2, %ecx, %xmm3, %xmm3 +; KNL-NEXT: kmovw %k1, %ecx ; KNL-NEXT: kshiftlw $10, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %r15d +; KNL-NEXT: vpinsrb $3, %eax, %xmm3, %xmm3 +; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $9, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %r12d +; KNL-NEXT: vpinsrb $4, %ecx, %xmm3, %xmm3 +; KNL-NEXT: kmovw %k1, %ecx ; KNL-NEXT: kshiftlw $8, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %r13d +; KNL-NEXT: vpinsrb $5, %eax, %xmm3, %xmm3 +; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $7, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ebx +; KNL-NEXT: vpinsrb $6, %ecx, %xmm3, %xmm3 +; KNL-NEXT: kmovw %k1, %ecx ; KNL-NEXT: kshiftlw $6, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ebp +; KNL-NEXT: vpinsrb $7, %eax, %xmm3, %xmm3 +; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $5, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $8, %ecx, %xmm3, %xmm3 +; KNL-NEXT: kmovw %k1, %ecx ; KNL-NEXT: kshiftlw $4, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx +; KNL-NEXT: vpinsrb $9, %eax, %xmm3, %xmm3 +; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $3, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %edx +; KNL-NEXT: vpinsrb $10, %ecx, %xmm3, %xmm3 +; KNL-NEXT: kmovw %k1, %ecx ; KNL-NEXT: kshiftlw $2, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %esi +; KNL-NEXT: vpinsrb $11, %eax, %xmm3, %xmm3 +; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $1, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: vmovd %r9d, %xmm3 -; KNL-NEXT: kmovw %k1, %r9d -; KNL-NEXT: vptestmd %zmm2, %zmm2, %k2 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: vpinsrb $1, %r8d, %xmm3, %xmm2 -; KNL-NEXT: vpinsrb $2, %r10d, %xmm2, %xmm2 -; KNL-NEXT: vpinsrb $3, %r11d, %xmm2, %xmm2 -; KNL-NEXT: vpinsrb $4, %r14d, %xmm2, %xmm2 -; KNL-NEXT: vpinsrb $5, %r15d, %xmm2, %xmm2 -; KNL-NEXT: vpinsrb $6, %r12d, %xmm2, %xmm2 -; KNL-NEXT: vpinsrb $7, %r13d, %xmm2, %xmm2 -; KNL-NEXT: vpinsrb $8, %ebx, %xmm2, %xmm2 -; KNL-NEXT: vpinsrb $9, %ebp, %xmm2, %xmm2 -; KNL-NEXT: vpinsrb $10, %eax, %xmm2, %xmm2 -; KNL-NEXT: vpinsrb $11, %ecx, %xmm2, %xmm2 -; KNL-NEXT: vpinsrb $12, %edx, %xmm2, %xmm2 -; KNL-NEXT: vpinsrb $13, %esi, %xmm2, %xmm2 -; KNL-NEXT: vpinsrb $14, %r9d, %xmm2, %xmm2 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $15, %eax, %xmm2, %xmm2 -; KNL-NEXT: vpmovsxbd %xmm2, %zmm2 -; KNL-NEXT: vpslld $31, %zmm2, %zmm2 -; KNL-NEXT: vptestmd %zmm2, %zmm2, %k0 -; KNL-NEXT: kmovw %k0, 6(%rdi) -; KNL-NEXT: kshiftlw $14, %k2, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %r8d -; KNL-NEXT: kshiftlw $15, %k2, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %r10d -; KNL-NEXT: kshiftlw $13, %k2, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %r9d -; KNL-NEXT: kshiftlw $12, %k2, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %r11d -; KNL-NEXT: kshiftlw $11, %k2, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %r14d -; KNL-NEXT: kshiftlw $10, %k2, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %r15d -; KNL-NEXT: kshiftlw $9, %k2, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %r12d -; KNL-NEXT: kshiftlw $8, %k2, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %r13d -; KNL-NEXT: kshiftlw $7, %k2, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %ecx -; KNL-NEXT: kshiftlw $6, %k2, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %esi -; KNL-NEXT: kshiftlw $5, %k2, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %ebp -; KNL-NEXT: kshiftlw $4, %k2, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %ebx -; KNL-NEXT: kshiftlw $3, %k2, %k0 +; KNL-NEXT: vpinsrb $12, %ecx, %xmm3, %xmm3 +; KNL-NEXT: kmovw %k1, %ecx +; KNL-NEXT: vptestmd %zmm2, %zmm2, %k1 ; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: vpinsrb $13, %eax, %xmm3, %xmm2 ; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: kshiftlw $2, %k2, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %edx -; KNL-NEXT: kshiftlw $1, %k2, %k0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: vmovd %r10d, %xmm2 -; KNL-NEXT: kmovw %k0, %r10d -; KNL-NEXT: vptestmd %zmm1, %zmm1, %k1 -; KNL-NEXT: kshiftrw $15, %k2, %k0 -; KNL-NEXT: vpinsrb $1, %r8d, %xmm2, %xmm1 -; KNL-NEXT: vpinsrb $2, %r9d, %xmm1, %xmm1 -; KNL-NEXT: vpinsrb $3, %r11d, %xmm1, %xmm1 -; KNL-NEXT: vpinsrb $4, %r14d, %xmm1, %xmm1 -; KNL-NEXT: vpinsrb $5, %r15d, %xmm1, %xmm1 -; KNL-NEXT: vpinsrb $6, %r12d, %xmm1, %xmm1 -; KNL-NEXT: vpinsrb $7, %r13d, %xmm1, %xmm1 -; KNL-NEXT: vpinsrb $8, %ecx, %xmm1, %xmm1 -; KNL-NEXT: vpinsrb $9, %esi, %xmm1, %xmm1 -; KNL-NEXT: vpinsrb $10, %ebp, %xmm1, %xmm1 -; KNL-NEXT: vpinsrb $11, %ebx, %xmm1, %xmm1 -; KNL-NEXT: vpinsrb $12, %eax, %xmm1, %xmm1 -; KNL-NEXT: vpinsrb $13, %edx, %xmm1, %xmm1 -; KNL-NEXT: vpinsrb $14, %r10d, %xmm1, %xmm1 -; KNL-NEXT: kmovw %k0, %eax -; KNL-NEXT: vpinsrb $15, %eax, %xmm1, %xmm1 -; KNL-NEXT: vpmovsxbd %xmm1, %zmm1 -; KNL-NEXT: vpslld $31, %zmm1, %zmm1 -; KNL-NEXT: vptestmd %zmm1, %zmm1, %k0 -; KNL-NEXT: kmovw %k0, 4(%rdi) ; KNL-NEXT: kshiftlw $14, %k1, %k0 ; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %r8d +; KNL-NEXT: vpinsrb $14, %ecx, %xmm2, %xmm2 +; KNL-NEXT: kmovw %k0, %ecx ; KNL-NEXT: kshiftlw $15, %k1, %k0 ; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %r10d +; KNL-NEXT: vpinsrb $15, %eax, %xmm2, %xmm2 +; KNL-NEXT: kmovw %k0, %eax ; KNL-NEXT: kshiftlw $13, %k1, %k0 ; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %r9d +; KNL-NEXT: vmovd %eax, %xmm3 +; KNL-NEXT: kmovw %k0, %eax ; KNL-NEXT: kshiftlw $12, %k1, %k0 +; KNL-NEXT: vpinsrb $1, %ecx, %xmm3, %xmm3 ; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %r11d +; KNL-NEXT: kmovw %k0, %ecx ; KNL-NEXT: kshiftlw $11, %k1, %k0 +; KNL-NEXT: vpinsrb $2, %eax, %xmm3, %xmm3 ; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %r14d +; KNL-NEXT: kmovw %k0, %eax ; KNL-NEXT: kshiftlw $10, %k1, %k0 +; KNL-NEXT: vpinsrb $3, %ecx, %xmm3, %xmm3 ; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %r15d +; KNL-NEXT: kmovw %k0, %ecx ; KNL-NEXT: kshiftlw $9, %k1, %k0 +; KNL-NEXT: vpinsrb $4, %eax, %xmm3, %xmm3 ; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %r12d +; KNL-NEXT: kmovw %k0, %eax ; KNL-NEXT: kshiftlw $8, %k1, %k0 +; KNL-NEXT: vpinsrb $5, %ecx, %xmm3, %xmm3 ; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %r13d +; KNL-NEXT: kmovw %k0, %ecx ; KNL-NEXT: kshiftlw $7, %k1, %k0 +; KNL-NEXT: vpinsrb $6, %eax, %xmm3, %xmm3 ; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %ecx +; KNL-NEXT: kmovw %k0, %eax ; KNL-NEXT: kshiftlw $6, %k1, %k0 +; KNL-NEXT: vpinsrb $7, %ecx, %xmm3, %xmm3 ; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %esi +; KNL-NEXT: kmovw %k0, %ecx ; KNL-NEXT: kshiftlw $5, %k1, %k0 +; KNL-NEXT: vpinsrb $8, %eax, %xmm3, %xmm3 ; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %ebp +; KNL-NEXT: kmovw %k0, %eax ; KNL-NEXT: kshiftlw $4, %k1, %k0 +; KNL-NEXT: vpinsrb $9, %ecx, %xmm3, %xmm3 ; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %ebx +; KNL-NEXT: kmovw %k0, %ecx ; KNL-NEXT: kshiftlw $3, %k1, %k0 +; KNL-NEXT: vpinsrb $10, %eax, %xmm3, %xmm3 ; KNL-NEXT: kshiftrw $15, %k0, %k0 ; KNL-NEXT: kmovw %k0, %eax ; KNL-NEXT: kshiftlw $2, %k1, %k0 +; KNL-NEXT: vpinsrb $11, %ecx, %xmm3, %xmm3 ; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: kmovw %k0, %edx +; KNL-NEXT: kmovw %k0, %ecx ; KNL-NEXT: kshiftlw $1, %k1, %k0 +; KNL-NEXT: vpinsrb $12, %eax, %xmm3, %xmm3 ; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: vmovd %r10d, %xmm1 -; KNL-NEXT: kmovw %k0, %r10d -; KNL-NEXT: vptestmd %zmm0, %zmm0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vptestmd %zmm1, %zmm1, %k0 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: vpinsrb $1, %r8d, %xmm1, %xmm0 -; KNL-NEXT: vpinsrb $2, %r9d, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $3, %r11d, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $4, %r14d, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $5, %r15d, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $6, %r12d, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $7, %r13d, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $8, %ecx, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $9, %esi, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $10, %ebp, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $11, %ebx, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $12, %eax, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $13, %edx, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $14, %r10d, %xmm0, %xmm0 -; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $15, %eax, %xmm0, %xmm0 -; KNL-NEXT: vpmovsxbd %xmm0, %zmm0 -; KNL-NEXT: vpslld $31, %zmm0, %zmm0 -; KNL-NEXT: vptestmd %zmm0, %zmm0, %k1 -; KNL-NEXT: kmovw %k1, 2(%rdi) +; KNL-NEXT: vpmovsxbd %xmm2, %zmm1 +; KNL-NEXT: vpslld $31, %zmm1, %zmm1 +; KNL-NEXT: vpinsrb $13, %ecx, %xmm3, %xmm2 +; KNL-NEXT: kmovw %k1, %ecx +; KNL-NEXT: vptestmd %zmm1, %zmm1, %k1 +; KNL-NEXT: vpinsrb $14, %eax, %xmm2, %xmm1 +; KNL-NEXT: vpinsrb $15, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, 6(%rdi) +; KNL-NEXT: vpmovsxbd %xmm1, %zmm1 +; KNL-NEXT: vpslld $31, %zmm1, %zmm1 +; KNL-NEXT: vptestmd %zmm1, %zmm1, %k1 +; KNL-NEXT: kmovw %k1, 4(%rdi) ; KNL-NEXT: kshiftlw $14, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %r8d +; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $15, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %r9d +; KNL-NEXT: kmovw %k1, %ecx ; KNL-NEXT: kshiftlw $13, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %r10d +; KNL-NEXT: vmovd %ecx, %xmm1 +; KNL-NEXT: kmovw %k1, %ecx ; KNL-NEXT: kshiftlw $12, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %r11d +; KNL-NEXT: vpinsrb $1, %eax, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $11, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %r14d +; KNL-NEXT: vpinsrb $2, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %ecx ; KNL-NEXT: kshiftlw $10, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %r15d +; KNL-NEXT: vpinsrb $3, %eax, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $9, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %r12d +; KNL-NEXT: vpinsrb $4, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %ecx ; KNL-NEXT: kshiftlw $8, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %r13d +; KNL-NEXT: vpinsrb $5, %eax, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $7, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %edx +; KNL-NEXT: vpinsrb $6, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %ecx ; KNL-NEXT: kshiftlw $6, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %esi +; KNL-NEXT: vpinsrb $7, %eax, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $5, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ebp +; KNL-NEXT: vpinsrb $8, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %ecx ; KNL-NEXT: kshiftlw $4, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ebx +; KNL-NEXT: vpinsrb $9, %eax, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $3, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax +; KNL-NEXT: vpinsrb $10, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %ecx ; KNL-NEXT: kshiftlw $2, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %ecx +; KNL-NEXT: vpinsrb $11, %eax, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $1, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: vmovd %r9d, %xmm0 -; KNL-NEXT: kmovw %k1, %r9d -; KNL-NEXT: vpinsrb $1, %r8d, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $2, %r10d, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $3, %r11d, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $4, %r14d, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $5, %r15d, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $6, %r12d, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $7, %r13d, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $8, %edx, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $9, %esi, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $10, %ebp, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $11, %ebx, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $12, %eax, %xmm0, %xmm0 -; KNL-NEXT: kshiftrw $15, %k0, %k0 -; KNL-NEXT: vpinsrb $13, %ecx, %xmm0, %xmm0 -; KNL-NEXT: vpinsrb $14, %r9d, %xmm0, %xmm0 +; KNL-NEXT: vpinsrb $12, %ecx, %xmm1, %xmm1 +; KNL-NEXT: kmovw %k1, %ecx +; KNL-NEXT: vptestmd %zmm0, %zmm0, %k1 +; KNL-NEXT: vpinsrb $13, %eax, %xmm1, %xmm0 +; KNL-NEXT: vpinsrb $14, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: vpinsrb $15, %eax, %xmm0, %xmm0 +; KNL-NEXT: vpmovsxbd %xmm0, %zmm0 +; KNL-NEXT: vpslld $31, %zmm0, %zmm0 +; KNL-NEXT: vptestmd %zmm0, %zmm0, %k0 +; KNL-NEXT: kmovw %k0, 2(%rdi) +; KNL-NEXT: kshiftlw $14, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: kshiftlw $15, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %ecx +; KNL-NEXT: kshiftlw $13, %k1, %k0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: vmovd %ecx, %xmm0 +; KNL-NEXT: kmovw %k0, %ecx +; KNL-NEXT: kshiftlw $12, %k1, %k0 +; KNL-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: kshiftlw $11, %k1, %k0 +; KNL-NEXT: vpinsrb $2, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %ecx +; KNL-NEXT: kshiftlw $10, %k1, %k0 +; KNL-NEXT: vpinsrb $3, %eax, %xmm0, %xmm0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: kshiftlw $9, %k1, %k0 +; KNL-NEXT: vpinsrb $4, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %ecx +; KNL-NEXT: kshiftlw $8, %k1, %k0 +; KNL-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: kshiftlw $7, %k1, %k0 +; KNL-NEXT: vpinsrb $6, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %ecx +; KNL-NEXT: kshiftlw $6, %k1, %k0 +; KNL-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: kshiftlw $5, %k1, %k0 +; KNL-NEXT: vpinsrb $8, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %ecx +; KNL-NEXT: kshiftlw $4, %k1, %k0 +; KNL-NEXT: vpinsrb $9, %eax, %xmm0, %xmm0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: kshiftlw $3, %k1, %k0 +; KNL-NEXT: vpinsrb $10, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %ecx +; KNL-NEXT: kshiftlw $2, %k1, %k0 +; KNL-NEXT: vpinsrb $11, %eax, %xmm0, %xmm0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %eax +; KNL-NEXT: kshiftlw $1, %k1, %k0 +; KNL-NEXT: vpinsrb $12, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kshiftrw $15, %k0, %k0 +; KNL-NEXT: kmovw %k0, %ecx +; KNL-NEXT: kshiftrw $15, %k1, %k0 +; KNL-NEXT: vpinsrb $13, %eax, %xmm0, %xmm0 +; KNL-NEXT: vpinsrb $14, %ecx, %xmm0, %xmm0 ; KNL-NEXT: kmovw %k0, %eax ; KNL-NEXT: vpinsrb $15, %eax, %xmm0, %xmm0 ; KNL-NEXT: vpmovsxbd %xmm0, %zmm0 ; KNL-NEXT: vpslld $31, %zmm0, %zmm0 ; KNL-NEXT: vptestmd %zmm0, %zmm0, %k0 ; KNL-NEXT: kmovw %k0, (%rdi) -; KNL-NEXT: popq %rbx -; KNL-NEXT: popq %r12 -; KNL-NEXT: popq %r13 -; KNL-NEXT: popq %r14 -; KNL-NEXT: popq %r15 -; KNL-NEXT: popq %rbp ; KNL-NEXT: retq ; ; SKX-LABEL: store_64i1: diff --git a/llvm/test/CodeGen/X86/avx512-vec-cmp.ll b/llvm/test/CodeGen/X86/avx512-vec-cmp.ll index 2b04b9229b3..02ee3adeb7a 100644 --- a/llvm/test/CodeGen/X86/avx512-vec-cmp.ll +++ b/llvm/test/CodeGen/X86/avx512-vec-cmp.ll @@ -8,6 +8,7 @@ define <16 x float> @test1(<16 x float> %x, <16 x float> %y) nounwind { ; CHECK-NEXT: vcmpleps %zmm1, %zmm0, %k1 ; CHECK-NEXT: vblendmps %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %mask = fcmp ole <16 x float> %x, %y %max = select <16 x i1> %mask, <16 x float> %x, <16 x float> %y ret <16 x float> %max @@ -19,6 +20,7 @@ define <8 x double> @test2(<8 x double> %x, <8 x double> %y) nounwind { ; CHECK-NEXT: vcmplepd %zmm1, %zmm0, %k1 ; CHECK-NEXT: vblendmpd %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %mask = fcmp ole <8 x double> %x, %y %max = select <8 x i1> %mask, <8 x double> %x, <8 x double> %y ret <8 x double> %max @@ -30,6 +32,7 @@ define <16 x i32> @test3(<16 x i32> %x, <16 x i32> %x1, <16 x i32>* %yp) nounwin ; CHECK-NEXT: vpcmpeqd (%rdi), %zmm0, %k1 ; CHECK-NEXT: vpblendmd %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %y = load <16 x i32>, <16 x i32>* %yp, align 4 %mask = icmp eq <16 x i32> %x, %y %max = select <16 x i1> %mask, <16 x i32> %x, <16 x i32> %x1 @@ -42,6 +45,7 @@ define <16 x i32> @test4_unsigned(<16 x i32> %x, <16 x i32> %y, <16 x i32> %x1) ; CHECK-NEXT: vpcmpnltud %zmm1, %zmm0, %k1 ; CHECK-NEXT: vpblendmd %zmm2, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %mask = icmp uge <16 x i32> %x, %y %max = select <16 x i1> %mask, <16 x i32> %x1, <16 x i32> %y ret <16 x i32> %max @@ -53,6 +57,7 @@ define <8 x i64> @test5(<8 x i64> %x, <8 x i64> %y) nounwind { ; CHECK-NEXT: vpcmpeqq %zmm1, %zmm0, %k1 ; CHECK-NEXT: vpblendmq %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %mask = icmp eq <8 x i64> %x, %y %max = select <8 x i1> %mask, <8 x i64> %x, <8 x i64> %y ret <8 x i64> %max @@ -64,6 +69,7 @@ define <8 x i64> @test6_unsigned(<8 x i64> %x, <8 x i64> %y, <8 x i64> %x1) noun ; CHECK-NEXT: vpcmpnleuq %zmm1, %zmm0, %k1 ; CHECK-NEXT: vpblendmq %zmm2, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %mask = icmp ugt <8 x i64> %x, %y %max = select <8 x i1> %mask, <8 x i64> %x1, <8 x i64> %y ret <8 x i64> %max @@ -117,12 +123,14 @@ define <8 x i32> @test9(<8 x i32> %x, <8 x i32> %y) nounwind { ; KNL-NEXT: vpblendmd %zmm0, %zmm1, %zmm0 {%k1} ; KNL-NEXT: ## kill: %YMM0<def> %YMM0<kill> %ZMM0<kill> ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test9: ; SKX: ## BB#0: ; SKX-NEXT: vpcmpeqd %ymm1, %ymm0, %k1 ; SKX-NEXT: vpblendmd %ymm0, %ymm1, %ymm0 {%k1} ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %mask = icmp eq <8 x i32> %x, %y %max = select <8 x i1> %mask, <8 x i32> %x, <8 x i32> %y ret <8 x i32> %max @@ -137,12 +145,14 @@ define <8 x float> @test10(<8 x float> %x, <8 x float> %y) nounwind { ; KNL-NEXT: vblendmps %zmm0, %zmm1, %zmm0 {%k1} ; KNL-NEXT: ## kill: %YMM0<def> %YMM0<kill> %ZMM0<kill> ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test10: ; SKX: ## BB#0: ; SKX-NEXT: vcmpeqps %ymm1, %ymm0, %k1 ; SKX-NEXT: vblendmps %ymm0, %ymm1, %ymm0 {%k1} ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %mask = fcmp oeq <8 x float> %x, %y %max = select <8 x i1> %mask, <8 x float> %x, <8 x float> %y @@ -154,6 +164,7 @@ define <8 x i32> @test11_unsigned(<8 x i32> %x, <8 x i32> %y) nounwind { ; CHECK: ## BB#0: ; CHECK-NEXT: vpmaxud %ymm1, %ymm0, %ymm0 ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %mask = icmp ugt <8 x i32> %x, %y %max = select <8 x i1> %mask, <8 x i32> %x, <8 x i32> %y ret <8 x i32> %max @@ -168,6 +179,7 @@ define i16 @test12(<16 x i64> %a, <16 x i64> %b) nounwind { ; KNL-NEXT: kmovw %k0, %eax ; KNL-NEXT: ## kill: %AX<def> %AX<kill> %EAX<kill> ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test12: ; SKX: ## BB#0: @@ -178,6 +190,7 @@ define i16 @test12(<16 x i64> %a, <16 x i64> %b) nounwind { ; SKX-NEXT: ## kill: %AX<def> %AX<kill> %EAX<kill> ; SKX-NEXT: vzeroupper ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %res = icmp eq <16 x i64> %a, %b %res1 = bitcast <16 x i1> %res to i16 ret i16 %res1 @@ -330,6 +343,7 @@ define i32 @test12_v32i32(<32 x i32> %a, <32 x i32> %b) nounwind { ; KNL-NEXT: movq %rbp, %rsp ; KNL-NEXT: popq %rbp ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test12_v32i32: ; SKX: ## BB#0: @@ -339,6 +353,7 @@ define i32 @test12_v32i32(<32 x i32> %a, <32 x i32> %b) nounwind { ; SKX-NEXT: kmovd %k0, %eax ; SKX-NEXT: vzeroupper ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %res = icmp eq <32 x i32> %a, %b %res1 = bitcast <32 x i1> %res to i32 ret i32 %res1 @@ -562,72 +577,72 @@ define i64 @test12_v64i16(<64 x i16> %a, <64 x i16> %b) nounwind { ; KNL-NEXT: vpinsrb $15, %eax, %xmm0, %xmm0 ; KNL-NEXT: vpmovsxbd %xmm0, %zmm0 ; KNL-NEXT: vpslld $31, %zmm0, %zmm0 -; KNL-NEXT: vptestmd %zmm0, %zmm0, %k0 -; KNL-NEXT: kmovw %k0, {{[0-9]+}}(%rsp) -; KNL-NEXT: vpcmpeqw %ymm6, %ymm2, %ymm0 -; KNL-NEXT: vpmovsxwd %ymm0, %zmm0 -; KNL-NEXT: vpslld $31, %zmm0, %zmm0 -; KNL-NEXT: vptestmd %zmm0, %zmm0, %k0 +; KNL-NEXT: vpcmpeqw %ymm6, %ymm2, %ymm1 +; KNL-NEXT: vpmovsxwd %ymm1, %zmm1 +; KNL-NEXT: vpslld $31, %zmm1, %zmm1 +; KNL-NEXT: vptestmd %zmm1, %zmm1, %k0 ; KNL-NEXT: kshiftlw $14, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax ; KNL-NEXT: kshiftlw $15, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %ecx -; KNL-NEXT: vmovd %ecx, %xmm0 -; KNL-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 +; KNL-NEXT: vmovd %ecx, %xmm1 +; KNL-NEXT: vpinsrb $1, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $13, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 +; KNL-NEXT: vpinsrb $2, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $12, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $3, %eax, %xmm0, %xmm0 +; KNL-NEXT: vpinsrb $3, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $11, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $4, %eax, %xmm0, %xmm0 +; KNL-NEXT: vpinsrb $4, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $10, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 +; KNL-NEXT: vpinsrb $5, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $9, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 +; KNL-NEXT: vpinsrb $6, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $8, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 +; KNL-NEXT: vpinsrb $7, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $7, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $8, %eax, %xmm0, %xmm0 +; KNL-NEXT: vpinsrb $8, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $6, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $9, %eax, %xmm0, %xmm0 +; KNL-NEXT: vpinsrb $9, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $5, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $10, %eax, %xmm0, %xmm0 +; KNL-NEXT: vpinsrb $10, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $4, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $11, %eax, %xmm0, %xmm0 +; KNL-NEXT: vpinsrb $11, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $3, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $12, %eax, %xmm0, %xmm0 +; KNL-NEXT: vpinsrb $12, %eax, %xmm1, %xmm1 ; KNL-NEXT: kshiftlw $2, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 ; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $13, %eax, %xmm0, %xmm0 ; KNL-NEXT: kshiftlw $1, %k0, %k1 ; KNL-NEXT: kshiftrw $15, %k1, %k1 -; KNL-NEXT: kmovw %k1, %eax -; KNL-NEXT: vpinsrb $14, %eax, %xmm0, %xmm0 +; KNL-NEXT: kmovw %k1, %ecx +; KNL-NEXT: vptestmd %zmm0, %zmm0, %k1 +; KNL-NEXT: vpinsrb $13, %eax, %xmm1, %xmm0 +; KNL-NEXT: vpinsrb $14, %ecx, %xmm0, %xmm0 +; KNL-NEXT: kmovw %k1, {{[0-9]+}}(%rsp) ; KNL-NEXT: kshiftrw $15, %k0, %k0 ; KNL-NEXT: kmovw %k0, %eax ; KNL-NEXT: vpinsrb $15, %eax, %xmm0, %xmm0 @@ -642,6 +657,7 @@ define i64 @test12_v64i16(<64 x i16> %a, <64 x i16> %b) nounwind { ; KNL-NEXT: movq %rbp, %rsp ; KNL-NEXT: popq %rbp ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test12_v64i16: ; SKX: ## BB#0: @@ -651,6 +667,7 @@ define i64 @test12_v64i16(<64 x i16> %a, <64 x i16> %b) nounwind { ; SKX-NEXT: kmovq %k0, %rax ; SKX-NEXT: vzeroupper ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %res = icmp eq <64 x i16> %a, %b %res1 = bitcast <64 x i1> %res to i64 ret i64 %res1 @@ -704,6 +721,7 @@ define <16 x i32> @test16(<16 x i32> %x, <16 x i32> %y, <16 x i32> %x1) nounwind ; CHECK-NEXT: vpcmpled %zmm0, %zmm1, %k1 ; CHECK-NEXT: vpblendmd %zmm2, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %mask = icmp sge <16 x i32> %x, %y %max = select <16 x i1> %mask, <16 x i32> %x1, <16 x i32> %y ret <16 x i32> %max @@ -715,6 +733,7 @@ define <16 x i32> @test17(<16 x i32> %x, <16 x i32> %x1, <16 x i32>* %y.ptr) nou ; CHECK-NEXT: vpcmpgtd (%rdi), %zmm0, %k1 ; CHECK-NEXT: vpblendmd %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %y = load <16 x i32>, <16 x i32>* %y.ptr, align 4 %mask = icmp sgt <16 x i32> %x, %y %max = select <16 x i1> %mask, <16 x i32> %x, <16 x i32> %x1 @@ -727,6 +746,7 @@ define <16 x i32> @test18(<16 x i32> %x, <16 x i32> %x1, <16 x i32>* %y.ptr) nou ; CHECK-NEXT: vpcmpled (%rdi), %zmm0, %k1 ; CHECK-NEXT: vpblendmd %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %y = load <16 x i32>, <16 x i32>* %y.ptr, align 4 %mask = icmp sle <16 x i32> %x, %y %max = select <16 x i1> %mask, <16 x i32> %x, <16 x i32> %x1 @@ -739,6 +759,7 @@ define <16 x i32> @test19(<16 x i32> %x, <16 x i32> %x1, <16 x i32>* %y.ptr) nou ; CHECK-NEXT: vpcmpleud (%rdi), %zmm0, %k1 ; CHECK-NEXT: vpblendmd %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %y = load <16 x i32>, <16 x i32>* %y.ptr, align 4 %mask = icmp ule <16 x i32> %x, %y %max = select <16 x i1> %mask, <16 x i32> %x, <16 x i32> %x1 @@ -752,6 +773,7 @@ define <16 x i32> @test20(<16 x i32> %x, <16 x i32> %y, <16 x i32> %x1, <16 x i3 ; CHECK-NEXT: vpcmpeqd %zmm3, %zmm2, %k1 {%k1} ; CHECK-NEXT: vpblendmd %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %mask1 = icmp eq <16 x i32> %x1, %y1 %mask0 = icmp eq <16 x i32> %x, %y %mask = select <16 x i1> %mask0, <16 x i1> %mask1, <16 x i1> zeroinitializer @@ -766,6 +788,7 @@ define <8 x i64> @test21(<8 x i64> %x, <8 x i64> %y, <8 x i64> %x1, <8 x i64> %y ; CHECK-NEXT: vpcmpleq %zmm2, %zmm3, %k1 {%k1} ; CHECK-NEXT: vpblendmq %zmm0, %zmm2, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %mask1 = icmp sge <8 x i64> %x1, %y1 %mask0 = icmp sle <8 x i64> %x, %y %mask = select <8 x i1> %mask0, <8 x i1> %mask1, <8 x i1> zeroinitializer @@ -780,6 +803,7 @@ define <8 x i64> @test22(<8 x i64> %x, <8 x i64>* %y.ptr, <8 x i64> %x1, <8 x i6 ; CHECK-NEXT: vpcmpgtq (%rdi), %zmm0, %k1 {%k1} ; CHECK-NEXT: vpblendmq %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %mask1 = icmp sgt <8 x i64> %x1, %y1 %y = load <8 x i64>, <8 x i64>* %y.ptr, align 4 %mask0 = icmp sgt <8 x i64> %x, %y @@ -795,6 +819,7 @@ define <16 x i32> @test23(<16 x i32> %x, <16 x i32>* %y.ptr, <16 x i32> %x1, <16 ; CHECK-NEXT: vpcmpleud (%rdi), %zmm0, %k1 {%k1} ; CHECK-NEXT: vpblendmd %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %mask1 = icmp sge <16 x i32> %x1, %y1 %y = load <16 x i32>, <16 x i32>* %y.ptr, align 4 %mask0 = icmp ule <16 x i32> %x, %y @@ -809,6 +834,7 @@ define <8 x i64> @test24(<8 x i64> %x, <8 x i64> %x1, i64* %yb.ptr) nounwind { ; CHECK-NEXT: vpcmpeqq (%rdi){1to8}, %zmm0, %k1 ; CHECK-NEXT: vpblendmq %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %yb = load i64, i64* %yb.ptr, align 4 %y.0 = insertelement <8 x i64> undef, i64 %yb, i32 0 %y = shufflevector <8 x i64> %y.0, <8 x i64> undef, <8 x i32> zeroinitializer @@ -823,6 +849,7 @@ define <16 x i32> @test25(<16 x i32> %x, i32* %yb.ptr, <16 x i32> %x1) nounwind ; CHECK-NEXT: vpcmpled (%rdi){1to16}, %zmm0, %k1 ; CHECK-NEXT: vpblendmd %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %yb = load i32, i32* %yb.ptr, align 4 %y.0 = insertelement <16 x i32> undef, i32 %yb, i32 0 %y = shufflevector <16 x i32> %y.0, <16 x i32> undef, <16 x i32> zeroinitializer @@ -838,6 +865,7 @@ define <16 x i32> @test26(<16 x i32> %x, i32* %yb.ptr, <16 x i32> %x1, <16 x i32 ; CHECK-NEXT: vpcmpgtd (%rdi){1to16}, %zmm0, %k1 {%k1} ; CHECK-NEXT: vpblendmd %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %mask1 = icmp sge <16 x i32> %x1, %y1 %yb = load i32, i32* %yb.ptr, align 4 %y.0 = insertelement <16 x i32> undef, i32 %yb, i32 0 @@ -855,6 +883,7 @@ define <8 x i64> @test27(<8 x i64> %x, i64* %yb.ptr, <8 x i64> %x1, <8 x i64> %y ; CHECK-NEXT: vpcmpleq (%rdi){1to8}, %zmm0, %k1 {%k1} ; CHECK-NEXT: vpblendmq %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %mask1 = icmp sge <8 x i64> %x1, %y1 %yb = load i64, i64* %yb.ptr, align 4 %y.0 = insertelement <8 x i64> undef, i64 %yb, i32 0 @@ -920,12 +949,14 @@ define <4 x double> @test30(<4 x double> %x, <4 x double> %y) nounwind { ; KNL-NEXT: vcmpeqpd %ymm1, %ymm0, %ymm2 ; KNL-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0 ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test30: ; SKX: ## BB#0: ; SKX-NEXT: vcmpeqpd %ymm1, %ymm0, %k1 ; SKX-NEXT: vblendmpd %ymm0, %ymm1, %ymm0 {%k1} ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %mask = fcmp oeq <4 x double> %x, %y %max = select <4 x i1> %mask, <4 x double> %x, <4 x double> %y @@ -938,12 +969,14 @@ define <2 x double> @test31(<2 x double> %x, <2 x double> %x1, <2 x double>* %yp ; KNL-NEXT: vcmpltpd (%rdi), %xmm0, %xmm2 ; KNL-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0 ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test31: ; SKX: ## BB#0: ; SKX-NEXT: vcmpltpd (%rdi), %xmm0, %k1 ; SKX-NEXT: vblendmpd %xmm0, %xmm1, %xmm0 {%k1} ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %y = load <2 x double>, <2 x double>* %yp, align 4 %mask = fcmp olt <2 x double> %x, %y @@ -957,12 +990,14 @@ define <4 x double> @test32(<4 x double> %x, <4 x double> %x1, <4 x double>* %yp ; KNL-NEXT: vcmpltpd (%rdi), %ymm0, %ymm2 ; KNL-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0 ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test32: ; SKX: ## BB#0: ; SKX-NEXT: vcmpltpd (%rdi), %ymm0, %k1 ; SKX-NEXT: vblendmpd %ymm0, %ymm1, %ymm0 {%k1} ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %y = load <4 x double>, <4 x double>* %yp, align 4 %mask = fcmp ogt <4 x double> %y, %x @@ -976,6 +1011,7 @@ define <8 x double> @test33(<8 x double> %x, <8 x double> %x1, <8 x double>* %yp ; CHECK-NEXT: vcmpltpd (%rdi), %zmm0, %k1 ; CHECK-NEXT: vblendmpd %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %y = load <8 x double>, <8 x double>* %yp, align 4 %mask = fcmp olt <8 x double> %x, %y %max = select <8 x i1> %mask, <8 x double> %x, <8 x double> %x1 @@ -988,12 +1024,14 @@ define <4 x float> @test34(<4 x float> %x, <4 x float> %x1, <4 x float>* %yp) no ; KNL-NEXT: vcmpltps (%rdi), %xmm0, %xmm2 ; KNL-NEXT: vblendvps %xmm2, %xmm0, %xmm1, %xmm0 ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test34: ; SKX: ## BB#0: ; SKX-NEXT: vcmpltps (%rdi), %xmm0, %k1 ; SKX-NEXT: vblendmps %xmm0, %xmm1, %xmm0 {%k1} ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %y = load <4 x float>, <4 x float>* %yp, align 4 %mask = fcmp olt <4 x float> %x, %y %max = select <4 x i1> %mask, <4 x float> %x, <4 x float> %x1 @@ -1010,12 +1048,14 @@ define <8 x float> @test35(<8 x float> %x, <8 x float> %x1, <8 x float>* %yp) no ; KNL-NEXT: vblendmps %zmm0, %zmm1, %zmm0 {%k1} ; KNL-NEXT: ## kill: %YMM0<def> %YMM0<kill> %ZMM0<kill> ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test35: ; SKX: ## BB#0: ; SKX-NEXT: vcmpltps (%rdi), %ymm0, %k1 ; SKX-NEXT: vblendmps %ymm0, %ymm1, %ymm0 {%k1} ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %y = load <8 x float>, <8 x float>* %yp, align 4 %mask = fcmp ogt <8 x float> %y, %x @@ -1029,6 +1069,7 @@ define <16 x float> @test36(<16 x float> %x, <16 x float> %x1, <16 x float>* %yp ; CHECK-NEXT: vcmpltps (%rdi), %zmm0, %k1 ; CHECK-NEXT: vblendmps %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %y = load <16 x float>, <16 x float>* %yp, align 4 %mask = fcmp olt <16 x float> %x, %y %max = select <16 x i1> %mask, <16 x float> %x, <16 x float> %x1 @@ -1041,6 +1082,7 @@ define <8 x double> @test37(<8 x double> %x, <8 x double> %x1, double* %ptr) nou ; CHECK-NEXT: vcmpltpd (%rdi){1to8}, %zmm0, %k1 ; CHECK-NEXT: vblendmpd %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %a = load double, double* %ptr %v = insertelement <8 x double> undef, double %a, i32 0 @@ -1058,12 +1100,14 @@ define <4 x double> @test38(<4 x double> %x, <4 x double> %x1, double* %ptr) nou ; KNL-NEXT: vcmpltpd %ymm2, %ymm0, %ymm2 ; KNL-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0 ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test38: ; SKX: ## BB#0: ; SKX-NEXT: vcmpltpd (%rdi){1to4}, %ymm0, %k1 ; SKX-NEXT: vblendmpd %ymm0, %ymm1, %ymm0 {%k1} ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %a = load double, double* %ptr %v = insertelement <4 x double> undef, double %a, i32 0 @@ -1081,12 +1125,14 @@ define <2 x double> @test39(<2 x double> %x, <2 x double> %x1, double* %ptr) nou ; KNL-NEXT: vcmpltpd %xmm2, %xmm0, %xmm2 ; KNL-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0 ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test39: ; SKX: ## BB#0: ; SKX-NEXT: vcmpltpd (%rdi){1to2}, %xmm0, %k1 ; SKX-NEXT: vblendmpd %xmm0, %xmm1, %xmm0 {%k1} ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %a = load double, double* %ptr %v = insertelement <2 x double> undef, double %a, i32 0 @@ -1104,6 +1150,7 @@ define <16 x float> @test40(<16 x float> %x, <16 x float> %x1, float* %ptr) n ; CHECK-NEXT: vcmpltps (%rdi){1to16}, %zmm0, %k1 ; CHECK-NEXT: vblendmps %zmm0, %zmm1, %zmm0 {%k1} ; CHECK-NEXT: retq +; CHECK-NEXT: ## -- End function %a = load float, float* %ptr %v = insertelement <16 x float> undef, float %a, i32 0 @@ -1124,12 +1171,14 @@ define <8 x float> @test41(<8 x float> %x, <8 x float> %x1, float* %ptr) noun ; KNL-NEXT: vblendmps %zmm0, %zmm1, %zmm0 {%k1} ; KNL-NEXT: ## kill: %YMM0<def> %YMM0<kill> %ZMM0<kill> ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test41: ; SKX: ## BB#0: ; SKX-NEXT: vcmpltps (%rdi){1to8}, %ymm0, %k1 ; SKX-NEXT: vblendmps %ymm0, %ymm1, %ymm0 {%k1} ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %a = load float, float* %ptr %v = insertelement <8 x float> undef, float %a, i32 0 @@ -1147,12 +1196,14 @@ define <4 x float> @test42(<4 x float> %x, <4 x float> %x1, float* %ptr) noun ; KNL-NEXT: vcmpltps %xmm2, %xmm0, %xmm2 ; KNL-NEXT: vblendvps %xmm2, %xmm0, %xmm1, %xmm0 ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test42: ; SKX: ## BB#0: ; SKX-NEXT: vcmpltps (%rdi){1to4}, %xmm0, %k1 ; SKX-NEXT: vblendmps %xmm0, %xmm1, %xmm0 {%k1} ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %a = load float, float* %ptr %v = insertelement <4 x float> undef, float %a, i32 0 @@ -1172,6 +1223,7 @@ define <8 x double> @test43(<8 x double> %x, <8 x double> %x1, double* %ptr,<8 x ; KNL-NEXT: vcmpltpd (%rdi){1to8}, %zmm0, %k1 {%k1} ; KNL-NEXT: vblendmpd %zmm0, %zmm1, %zmm0 {%k1} ; KNL-NEXT: retq +; KNL-NEXT: ## -- End function ; ; SKX-LABEL: test43: ; SKX: ## BB#0: @@ -1180,6 +1232,7 @@ define <8 x double> @test43(<8 x double> %x, <8 x double> %x1, double* %ptr,<8 x ; SKX-NEXT: vcmpltpd (%rdi){1to8}, %zmm0, %k1 {%k1} ; SKX-NEXT: vblendmpd %zmm0, %zmm1, %zmm0 {%k1} ; SKX-NEXT: retq +; SKX-NEXT: ## -- End function %a = load double, double* %ptr %v = insertelement <8 x double> undef, double %a, i32 0 diff --git a/llvm/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll b/llvm/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll index 2b89373ceb0..d56c4675b73 100644 --- a/llvm/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll +++ b/llvm/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll @@ -1685,8 +1685,6 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: .cfi_offset %esi, -12 ; AVX512F-32-NEXT: .Lcfi9: ; AVX512F-32-NEXT: .cfi_offset %ebx, -8 -; AVX512F-32-NEXT: vmovdqa64 %zmm1, %zmm6 -; AVX512F-32-NEXT: vmovdqa64 %zmm0, %zmm5 ; AVX512F-32-NEXT: movl {{[0-9]+}}(%esp), %ecx ; AVX512F-32-NEXT: movb %cl, %al ; AVX512F-32-NEXT: shrb $5, %al @@ -1707,39 +1705,39 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpsllw $8, %xmm2, %xmm2 ; AVX512F-32-NEXT: kmovd %ecx, %k0 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm3 ; AVX512F-32-NEXT: vpbroadcastw %xmm3, %xmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm2, %ymm3, %ymm3 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm2, %ymm3, %ymm3 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 ; AVX512F-32-NEXT: kmovd %edx, %k0 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 ; AVX512F-32-NEXT: vpslld $24, %xmm3, %xmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm2, %ymm3, %ymm3 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm2, %ymm3, %ymm3 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 ; AVX512F-32-NEXT: kmovd %ebx, %k0 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 ; AVX512F-32-NEXT: vpbroadcastd %xmm3, %xmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm2, %ymm3, %ymm3 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm2, %ymm3, %ymm3 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 ; AVX512F-32-NEXT: kmovd %eax, %k0 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 ; AVX512F-32-NEXT: vpsllq $40, %xmm3, %xmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm2, %ymm3, %ymm3 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm2, %ymm3, %ymm3 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[4,5,6,7] ; AVX512F-32-NEXT: movb %cl, %al ; AVX512F-32-NEXT: shrb $6, %al @@ -1748,8 +1746,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %cl, %al @@ -1758,8 +1756,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpsllq $56, %xmm2, %xmm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %ch, %al @@ -1767,8 +1765,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpbroadcastq %xmm2, %xmm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: andb $2, %al @@ -1777,8 +1775,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5,6] ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %ch, %dl @@ -1789,8 +1787,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: shrb $3, %al @@ -1798,8 +1796,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4] ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax @@ -1809,8 +1807,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpbroadcastd %xmm2, %xmm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax @@ -1820,8 +1818,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2] ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax @@ -1831,8 +1829,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax @@ -1842,8 +1840,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0] ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax @@ -1852,8 +1850,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %al, %dl @@ -1864,8 +1862,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpsllw $8, %xmm2, %xmm2 ; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %al, %bl @@ -1877,8 +1875,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 ; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: shrb $3, %dl @@ -1887,8 +1885,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpslld $24, %xmm2, %xmm2 ; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %al, %dl @@ -1898,8 +1896,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpbroadcastd %xmm2, %xmm2 ; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %al, %dl @@ -1910,8 +1908,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpsllq $40, %xmm2, %xmm2 ; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %al, %dl @@ -1921,8 +1919,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 ; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: # kill: %AL<def> %AL<kill> %EAX<kill> %EAX<def> @@ -1932,8 +1930,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpsllq $56, %xmm2, %xmm2 ; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax @@ -1942,8 +1940,8 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpbroadcastq %xmm2, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %al, %dl @@ -1952,444 +1950,444 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: kmovd %edx, %k1 ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5,6] -; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 -; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] -; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm3 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm2, %ymm4, %ymm3, %ymm3 +; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm4[4,5,6,7] +; AVX512F-32-NEXT: vpmovb2m %zmm3, %k0 ; AVX512F-32-NEXT: movb %al, %dl ; AVX512F-32-NEXT: andb $15, %dl ; AVX512F-32-NEXT: movb %dl, %al ; AVX512F-32-NEXT: shrb $2, %dl ; AVX512F-32-NEXT: kmovd %edx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 -; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 -; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 -; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] -; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm3 +; AVX512F-32-NEXT: vpbroadcastw %xmm3, %xmm3 +; AVX512F-32-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm4 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm5 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm2, %ymm5, %ymm4, %ymm4 +; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[4,5,6,7] +; AVX512F-32-NEXT: vpmovb2m %zmm4, %k0 ; AVX512F-32-NEXT: shrb $3, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4] -; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm3, %ymm4, %ymm2, %ymm2 -; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm4[4,5,6,7] -; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm4 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm4[0,1,2,3,4] +; AVX512F-32-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm5 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm6 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm2, %ymm6, %ymm5, %ymm5 +; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm5 = zmm5[0,1,2,3],zmm6[4,5,6,7] +; AVX512F-32-NEXT: vpmovb2m %zmm5, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: shrl $28, %eax ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 -; AVX512F-32-NEXT: vpbroadcastd %xmm2, %xmm2 -; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm4, %ymm2, %ymm2 -; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm4[4,5,6,7] -; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm5 +; AVX512F-32-NEXT: vpbroadcastd %xmm5, %xmm5 +; AVX512F-32-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm6 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm7 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm2, %ymm7, %ymm6, %ymm6 +; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm7[4,5,6,7] +; AVX512F-32-NEXT: vpmovb2m %zmm6, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: movl %ecx, %esi ; AVX512F-32-NEXT: shrl $29, %eax ; AVX512F-32-NEXT: andb $1, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2] -; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm0 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm1 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm1, %ymm0, %ymm2, %ymm2 -; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm2[0,1,2,3],zmm0[4,5,6,7] -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm6 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm6 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm6[0,1,2] +; AVX512F-32-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm7 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm6 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm6, %ymm2, %ymm7, %ymm7 +; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm7[0,1,2,3],zmm2[4,5,6,7] +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %esi, %eax ; AVX512F-32-NEXT: shrl $30, %eax ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastw %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm1 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm0 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255] -; AVX512F-32-NEXT: vpblendvb %ymm2, %ymm0, %ymm1, %ymm1 -; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm1[0,1,2,3],zmm0[4,5,6,7] -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255] +; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %esi, %eax ; AVX512F-32-NEXT: shrl $31, %eax ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0] -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm1, %ymm0, %ymm0 -; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm1[4,5,6,7] -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0] +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl {{[0-9]+}}(%esp), %ecx ; AVX512F-32-NEXT: kmovd %ecx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm0 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm0, %ymm1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm7 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm1, %ymm7, %ymm1 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm2, %ymm3 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm3, %ymm4, %ymm3 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm3, %zmm2, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %cl, %al ; AVX512F-32-NEXT: andb $2, %al ; AVX512F-32-NEXT: shrb %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpsllw $8, %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpsllw $8, %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %cl, %dl ; AVX512F-32-NEXT: andb $15, %dl ; AVX512F-32-NEXT: movb %dl, %al ; AVX512F-32-NEXT: shrb $2, %dl ; AVX512F-32-NEXT: kmovd %edx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastw %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: shrb $3, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpslld $24, %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpslld $24, %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %cl, %al ; AVX512F-32-NEXT: shrb $4, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastd %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastd %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %cl, %al ; AVX512F-32-NEXT: shrb $5, %al ; AVX512F-32-NEXT: andb $1, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpsllq $40, %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpsllq $40, %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %cl, %al ; AVX512F-32-NEXT: shrb $6, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastw %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %cl, %al ; AVX512F-32-NEXT: shrb $7, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpsllq $56, %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpsllq $56, %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %ch, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastq %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastq %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: andb $2, %al ; AVX512F-32-NEXT: shrb %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6] -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5,6] +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %ch, %dl ; AVX512F-32-NEXT: andb $15, %dl ; AVX512F-32-NEXT: movb %dl, %al ; AVX512F-32-NEXT: shrb $2, %dl ; AVX512F-32-NEXT: kmovd %edx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastw %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: shrb $3, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4] -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4] +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: andl $61440, %eax # imm = 0xF000 ; AVX512F-32-NEXT: shrl $12, %eax ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastd %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastd %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: shrl $13, %eax ; AVX512F-32-NEXT: andb $1, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2] -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2] +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: andl $49152, %eax # imm = 0xC000 ; AVX512F-32-NEXT: shrl $14, %eax ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastw %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: andl $32768, %eax # imm = 0x8000 ; AVX512F-32-NEXT: shrl $15, %eax ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0] -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0] +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %ebx ; AVX512F-32-NEXT: shrl $16, %ebx ; AVX512F-32-NEXT: kmovd %ebx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %bl, %dl ; AVX512F-32-NEXT: andb $2, %dl ; AVX512F-32-NEXT: shrb %dl ; AVX512F-32-NEXT: kmovd %edx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpsllw $8, %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpsllw $8, %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 ; AVX512F-32-NEXT: movb %bl, %al ; AVX512F-32-NEXT: andb $15, %al ; AVX512F-32-NEXT: movb %al, %dl ; AVX512F-32-NEXT: shrb $2, %al ; AVX512F-32-NEXT: kmovd %eax, %k0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k1 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm0 -; AVX512F-32-NEXT: vpbroadcastw %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k1 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 +; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: shrb $3, %dl ; AVX512F-32-NEXT: kmovd %edx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpslld $24, %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpslld $24, %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %bl, %al ; AVX512F-32-NEXT: shrb $4, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastd %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastd %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %bl, %al ; AVX512F-32-NEXT: shrb $5, %al ; AVX512F-32-NEXT: andb $1, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpsllq $40, %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpsllq $40, %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %bl, %al ; AVX512F-32-NEXT: shrb $6, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastw %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: # kill: %BL<def> %BL<kill> %EBX<kill> %EBX<def> ; AVX512F-32-NEXT: shrb $7, %bl ; AVX512F-32-NEXT: kmovd %ebx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpsllq $56, %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpsllq $56, %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: shrl $24, %eax ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastq %xmm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastq %xmm2, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %al, %dl ; AVX512F-32-NEXT: andb $2, %dl ; AVX512F-32-NEXT: shrb %dl ; AVX512F-32-NEXT: kmovd %edx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6] -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5,6] +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 ; AVX512F-32-NEXT: movb %al, %dl ; AVX512F-32-NEXT: andb $15, %dl ; AVX512F-32-NEXT: movb %dl, %al ; AVX512F-32-NEXT: shrb $2, %dl ; AVX512F-32-NEXT: kmovd %edx, %k0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k1 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm0 -; AVX512F-32-NEXT: vpbroadcastw %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k1 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 +; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 ; AVX512F-32-NEXT: shrb $3, %al -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm0 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 +; AVX512F-32-NEXT: kmovd %eax, %k0 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm3[0,1,2,3,4] +; AVX512F-32-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm2, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm3, %ymm3 +; AVX512F-32-NEXT: movl %ecx, %eax +; AVX512F-32-NEXT: shrl $28, %eax +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm3, %zmm2, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm2, %ymm3 ; AVX512F-32-NEXT: kmovd %eax, %k0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4] -; AVX512F-32-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm0, %ymm4 -; AVX512F-32-NEXT: vpblendvb %ymm3, %ymm4, %ymm1, %ymm1 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm4 +; AVX512F-32-NEXT: vpbroadcastd %xmm4, %xmm4 +; AVX512F-32-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm3, %ymm4, %ymm3 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: shrl $29, %eax ; AVX512F-32-NEXT: andb $1, %al ; AVX512F-32-NEXT: kmovd %eax, %k0 -; AVX512F-32-NEXT: movl %ecx, %eax -; AVX512F-32-NEXT: shrl $28, %eax -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm0, %ymm1 -; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm3 -; AVX512F-32-NEXT: vpbroadcastd %xmm3, %xmm3 -; AVX512F-32-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm1, %ymm3, %ymm1 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm0, %ymm1 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm3[0,1,2] -; AVX512F-32-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm1, %ymm3, %ymm1 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm3, %zmm2, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k1 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm2, %ymm3 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm4 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm4[0,1,2] +; AVX512F-32-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4 +; AVX512F-32-NEXT: vpblendvb %ymm6, %ymm3, %ymm4, %ymm3 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: shrl $30, %eax -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm0 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm0, %ymm1 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm3, %zmm2, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm2, %ymm3 ; AVX512F-32-NEXT: kmovd %eax, %k0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vpbroadcastw %xmm3, %xmm3 -; AVX512F-32-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3 -; AVX512F-32-NEXT: vpblendvb %ymm2, %ymm1, %ymm3, %ymm1 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm4 +; AVX512F-32-NEXT: vpbroadcastw %xmm4, %xmm4 +; AVX512F-32-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4 +; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm3, %ymm4, %ymm3 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm3, %zmm2, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: shrl $31, %eax ; AVX512F-32-NEXT: kshiftlq $1, %k0, %k0 @@ -2397,12 +2395,12 @@ define i64 @test_mask_cmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) { ; AVX512F-32-NEXT: kmovd %eax, %k1 ; AVX512F-32-NEXT: kshiftlq $63, %k1, %k1 ; AVX512F-32-NEXT: korq %k1, %k0, %k1 -; AVX512F-32-NEXT: vpcmpeqb %zmm6, %zmm5, %k0 {%k1} -; AVX512F-32-NEXT: vpcmpgtb %zmm5, %zmm6, %k2 {%k1} -; AVX512F-32-NEXT: vpcmpleb %zmm6, %zmm5, %k3 {%k1} -; AVX512F-32-NEXT: vpcmpneqb %zmm6, %zmm5, %k4 {%k1} -; AVX512F-32-NEXT: vpcmpleb %zmm5, %zmm6, %k5 {%k1} -; AVX512F-32-NEXT: vpcmpgtb %zmm6, %zmm5, %k1 {%k1} +; AVX512F-32-NEXT: vpcmpeqb %zmm1, %zmm0, %k0 {%k1} +; AVX512F-32-NEXT: vpcmpgtb %zmm0, %zmm1, %k2 {%k1} +; AVX512F-32-NEXT: vpcmpleb %zmm1, %zmm0, %k3 {%k1} +; AVX512F-32-NEXT: vpcmpneqb %zmm1, %zmm0, %k4 {%k1} +; AVX512F-32-NEXT: vpcmpleb %zmm0, %zmm1, %k5 {%k1} +; AVX512F-32-NEXT: vpcmpgtb %zmm1, %zmm0, %k1 {%k1} ; AVX512F-32-NEXT: kmovq %k0, (%esp) ; AVX512F-32-NEXT: movl (%esp), %eax ; AVX512F-32-NEXT: movl {{[0-9]+}}(%esp), %edx @@ -2571,8 +2569,6 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: .cfi_offset %esi, -12 ; AVX512F-32-NEXT: .Lcfi15: ; AVX512F-32-NEXT: .cfi_offset %ebx, -8 -; AVX512F-32-NEXT: vmovdqa64 %zmm1, %zmm6 -; AVX512F-32-NEXT: vmovdqa64 %zmm0, %zmm5 ; AVX512F-32-NEXT: movl {{[0-9]+}}(%esp), %ecx ; AVX512F-32-NEXT: movb %cl, %al ; AVX512F-32-NEXT: shrb $5, %al @@ -2593,39 +2589,39 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpsllw $8, %xmm2, %xmm2 ; AVX512F-32-NEXT: kmovd %ecx, %k0 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm3 ; AVX512F-32-NEXT: vpbroadcastw %xmm3, %xmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm2, %ymm3, %ymm3 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm2, %ymm3, %ymm3 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 ; AVX512F-32-NEXT: kmovd %edx, %k0 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 ; AVX512F-32-NEXT: vpslld $24, %xmm3, %xmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm2, %ymm3, %ymm3 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm2, %ymm3, %ymm3 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 ; AVX512F-32-NEXT: kmovd %ebx, %k0 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 ; AVX512F-32-NEXT: vpbroadcastd %xmm3, %xmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm2, %ymm3, %ymm3 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm2, %ymm3, %ymm3 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 ; AVX512F-32-NEXT: kmovd %eax, %k0 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 ; AVX512F-32-NEXT: vpsllq $40, %xmm3, %xmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm2, %ymm3, %ymm3 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm2, %ymm3, %ymm3 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm3[0,1,2,3],zmm2[4,5,6,7] ; AVX512F-32-NEXT: movb %cl, %al ; AVX512F-32-NEXT: shrb $6, %al @@ -2634,8 +2630,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %cl, %al @@ -2644,8 +2640,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpsllq $56, %xmm2, %xmm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %ch, %al @@ -2653,8 +2649,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpbroadcastq %xmm2, %xmm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: andb $2, %al @@ -2663,8 +2659,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5,6] ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %ch, %dl @@ -2675,8 +2671,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: shrb $3, %al @@ -2684,8 +2680,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4] ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax @@ -2695,8 +2691,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpbroadcastd %xmm2, %xmm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax @@ -2706,8 +2702,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2] ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax @@ -2717,8 +2713,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax @@ -2728,8 +2724,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0] ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax @@ -2738,8 +2734,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %al, %dl @@ -2750,8 +2746,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpsllw $8, %xmm2, %xmm2 ; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %al, %bl @@ -2763,8 +2759,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 ; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: shrb $3, %dl @@ -2773,8 +2769,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpslld $24, %xmm2, %xmm2 ; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %al, %dl @@ -2784,8 +2780,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpbroadcastd %xmm2, %xmm2 ; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %al, %dl @@ -2796,8 +2792,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpsllq $40, %xmm2, %xmm2 ; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %al, %dl @@ -2807,8 +2803,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 ; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: # kill: %AL<def> %AL<kill> %EAX<kill> %EAX<def> @@ -2818,8 +2814,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpsllq $56, %xmm2, %xmm2 ; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax @@ -2828,8 +2824,8 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpbroadcastq %xmm2, %ymm2 ; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 ; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] ; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %al, %dl @@ -2838,444 +2834,444 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: kmovd %edx, %k1 ; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 ; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5,6] -; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 -; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] -; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm3 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm2, %ymm4, %ymm3, %ymm3 +; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm3 = zmm3[0,1,2,3],zmm4[4,5,6,7] +; AVX512F-32-NEXT: vpmovb2m %zmm3, %k0 ; AVX512F-32-NEXT: movb %al, %dl ; AVX512F-32-NEXT: andb $15, %dl ; AVX512F-32-NEXT: movb %dl, %al ; AVX512F-32-NEXT: shrb $2, %dl ; AVX512F-32-NEXT: kmovd %edx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 -; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 -; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm3, %ymm2, %ymm2 -; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] -; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm3 +; AVX512F-32-NEXT: vpbroadcastw %xmm3, %xmm3 +; AVX512F-32-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm4 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm5 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm2, %ymm5, %ymm4, %ymm4 +; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm4 = zmm4[0,1,2,3],zmm5[4,5,6,7] +; AVX512F-32-NEXT: vpmovb2m %zmm4, %k0 ; AVX512F-32-NEXT: shrb $3, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4] -; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm3, %ymm4, %ymm2, %ymm2 -; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm4[4,5,6,7] -; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm4 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm4[0,1,2,3,4] +; AVX512F-32-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm5 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm6 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm2, %ymm6, %ymm5, %ymm5 +; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm5 = zmm5[0,1,2,3],zmm6[4,5,6,7] +; AVX512F-32-NEXT: vpmovb2m %zmm5, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: shrl $28, %eax ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 -; AVX512F-32-NEXT: vpbroadcastd %xmm2, %xmm2 -; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm0 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm0, %ymm4, %ymm2, %ymm2 -; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm4[4,5,6,7] -; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm5 +; AVX512F-32-NEXT: vpbroadcastd %xmm5, %xmm5 +; AVX512F-32-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm6 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm7 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm2, %ymm7, %ymm6, %ymm6 +; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm6 = zmm6[0,1,2,3],zmm7[4,5,6,7] +; AVX512F-32-NEXT: vpmovb2m %zmm6, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: movl %ecx, %esi ; AVX512F-32-NEXT: shrl $29, %eax ; AVX512F-32-NEXT: andb $1, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2] -; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm0 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm1 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm1, %ymm0, %ymm2, %ymm2 -; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm2[0,1,2,3],zmm0[4,5,6,7] -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm6 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm6 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm6[0,1,2] +; AVX512F-32-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm7 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm6 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm6, %ymm2, %ymm7, %ymm7 +; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm7[0,1,2,3],zmm2[4,5,6,7] +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %esi, %eax ; AVX512F-32-NEXT: shrl $30, %eax ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastw %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm1 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm0 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255] -; AVX512F-32-NEXT: vpblendvb %ymm2, %ymm0, %ymm1, %ymm1 -; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm1[0,1,2,3],zmm0[4,5,6,7] -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255] +; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %esi, %eax ; AVX512F-32-NEXT: shrl $31, %eax ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0] -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm1, %ymm0, %ymm0 -; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,2,3],zmm1[4,5,6,7] -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0] +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0] +; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm3, %ymm2, %ymm2 +; AVX512F-32-NEXT: vshufi64x2 {{.*#+}} zmm2 = zmm2[0,1,2,3],zmm3[4,5,6,7] +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl {{[0-9]+}}(%esp), %ecx ; AVX512F-32-NEXT: kmovd %ecx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm0 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm0, %ymm1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm7 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm1, %ymm7, %ymm1 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm2, %ymm3 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm3, %ymm4, %ymm3 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm3, %zmm2, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %cl, %al ; AVX512F-32-NEXT: andb $2, %al ; AVX512F-32-NEXT: shrb %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpsllw $8, %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpsllw $8, %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %cl, %dl ; AVX512F-32-NEXT: andb $15, %dl ; AVX512F-32-NEXT: movb %dl, %al ; AVX512F-32-NEXT: shrb $2, %dl ; AVX512F-32-NEXT: kmovd %edx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastw %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: shrb $3, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpslld $24, %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpslld $24, %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %cl, %al ; AVX512F-32-NEXT: shrb $4, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastd %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastd %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %cl, %al ; AVX512F-32-NEXT: shrb $5, %al ; AVX512F-32-NEXT: andb $1, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpsllq $40, %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpsllq $40, %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %cl, %al ; AVX512F-32-NEXT: shrb $6, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastw %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %cl, %al ; AVX512F-32-NEXT: shrb $7, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpsllq $56, %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpsllq $56, %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %ch, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastq %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastq %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: andb $2, %al ; AVX512F-32-NEXT: shrb %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6] -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5,6] +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %ch, %dl ; AVX512F-32-NEXT: andb $15, %dl ; AVX512F-32-NEXT: movb %dl, %al ; AVX512F-32-NEXT: shrb $2, %dl ; AVX512F-32-NEXT: kmovd %edx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastw %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: shrb $3, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4] -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4] +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: andl $61440, %eax # imm = 0xF000 ; AVX512F-32-NEXT: shrl $12, %eax ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastd %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastd %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: shrl $13, %eax ; AVX512F-32-NEXT: andb $1, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2] -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2] +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: andl $49152, %eax # imm = 0xC000 ; AVX512F-32-NEXT: shrl $14, %eax ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastw %xmm0, %xmm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: andl $32768, %eax # imm = 0x8000 ; AVX512F-32-NEXT: shrl $15, %eax ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0] -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0] +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %ebx ; AVX512F-32-NEXT: shrl $16, %ebx ; AVX512F-32-NEXT: kmovd %ebx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %bl, %dl ; AVX512F-32-NEXT: andb $2, %dl ; AVX512F-32-NEXT: shrb %dl ; AVX512F-32-NEXT: kmovd %edx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpsllw $8, %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpsllw $8, %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 ; AVX512F-32-NEXT: movb %bl, %al ; AVX512F-32-NEXT: andb $15, %al ; AVX512F-32-NEXT: movb %al, %dl ; AVX512F-32-NEXT: shrb $2, %al ; AVX512F-32-NEXT: kmovd %eax, %k0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k1 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm0 -; AVX512F-32-NEXT: vpbroadcastw %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k1 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 +; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: shrb $3, %dl ; AVX512F-32-NEXT: kmovd %edx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpslld $24, %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpslld $24, %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %bl, %al ; AVX512F-32-NEXT: shrb $4, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastd %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastd %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %bl, %al ; AVX512F-32-NEXT: shrb $5, %al ; AVX512F-32-NEXT: andb $1, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpsllq $40, %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpsllq $40, %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %bl, %al ; AVX512F-32-NEXT: shrb $6, %al ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastw %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: # kill: %BL<def> %BL<kill> %EBX<kill> %EBX<def> ; AVX512F-32-NEXT: shrb $7, %bl ; AVX512F-32-NEXT: kmovd %ebx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpsllq $56, %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpsllq $56, %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: shrl $24, %eax ; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpbroadcastq %xmm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpbroadcastq %xmm2, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movb %al, %dl ; AVX512F-32-NEXT: andb $2, %dl ; AVX512F-32-NEXT: shrb %dl ; AVX512F-32-NEXT: kmovd %edx, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6] -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5,6] +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 ; AVX512F-32-NEXT: movb %al, %dl ; AVX512F-32-NEXT: andb $15, %dl ; AVX512F-32-NEXT: movb %dl, %al ; AVX512F-32-NEXT: shrb $2, %dl ; AVX512F-32-NEXT: kmovd %edx, %k0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k1 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm0 -; AVX512F-32-NEXT: vpbroadcastw %xmm0, %xmm0 -; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm1, %ymm4 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm0 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k1 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 +; AVX512F-32-NEXT: vpbroadcastw %xmm2, %xmm2 +; AVX512F-32-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm3, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm2, %ymm2 ; AVX512F-32-NEXT: shrb $3, %al -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm0 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm2, %zmm3, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 +; AVX512F-32-NEXT: kmovd %eax, %k0 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm3[0,1,2,3,4] +; AVX512F-32-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm2, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm4, %ymm3, %ymm3 +; AVX512F-32-NEXT: movl %ecx, %eax +; AVX512F-32-NEXT: shrl $28, %eax +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm3, %zmm2, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm2, %ymm3 ; AVX512F-32-NEXT: kmovd %eax, %k0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm1 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4] -; AVX512F-32-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm0, %ymm4 -; AVX512F-32-NEXT: vpblendvb %ymm3, %ymm4, %ymm1, %ymm1 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm4 +; AVX512F-32-NEXT: vpbroadcastd %xmm4, %xmm4 +; AVX512F-32-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4 +; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255] +; AVX512F-32-NEXT: vpblendvb %ymm5, %ymm3, %ymm4, %ymm3 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: shrl $29, %eax ; AVX512F-32-NEXT: andb $1, %al ; AVX512F-32-NEXT: kmovd %eax, %k0 -; AVX512F-32-NEXT: movl %ecx, %eax -; AVX512F-32-NEXT: shrl $28, %eax -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm0, %ymm1 -; AVX512F-32-NEXT: kmovd %eax, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm3 -; AVX512F-32-NEXT: vpbroadcastd %xmm3, %xmm3 -; AVX512F-32-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm1, %ymm3, %ymm1 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k1 -; AVX512F-32-NEXT: vpmovm2b %k1, %zmm0 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm0, %ymm1 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm3[0,1,2] -; AVX512F-32-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3 -; AVX512F-32-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,255,255] -; AVX512F-32-NEXT: vpblendvb %ymm4, %ymm1, %ymm3, %ymm1 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm3, %zmm2, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k1 +; AVX512F-32-NEXT: vpmovm2b %k1, %zmm2 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm2, %ymm3 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm4 +; AVX512F-32-NEXT: vpslldq {{.*#+}} xmm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm4[0,1,2] +; AVX512F-32-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4 +; AVX512F-32-NEXT: vpblendvb %ymm6, %ymm3, %ymm4, %ymm3 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: shrl $30, %eax -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm0 -; AVX512F-32-NEXT: vextracti64x4 $1, %zmm0, %ymm1 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm3, %zmm2, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm2 +; AVX512F-32-NEXT: vextracti64x4 $1, %zmm2, %ymm3 ; AVX512F-32-NEXT: kmovd %eax, %k0 -; AVX512F-32-NEXT: vpmovm2b %k0, %zmm3 -; AVX512F-32-NEXT: vpbroadcastw %xmm3, %xmm3 -; AVX512F-32-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3 -; AVX512F-32-NEXT: vpblendvb %ymm2, %ymm1, %ymm3, %ymm1 -; AVX512F-32-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0 -; AVX512F-32-NEXT: vpmovb2m %zmm0, %k0 +; AVX512F-32-NEXT: vpmovm2b %k0, %zmm4 +; AVX512F-32-NEXT: vpbroadcastw %xmm4, %xmm4 +; AVX512F-32-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4 +; AVX512F-32-NEXT: vpblendvb %ymm7, %ymm3, %ymm4, %ymm3 +; AVX512F-32-NEXT: vinserti64x4 $1, %ymm3, %zmm2, %zmm2 +; AVX512F-32-NEXT: vpmovb2m %zmm2, %k0 ; AVX512F-32-NEXT: movl %ecx, %eax ; AVX512F-32-NEXT: shrl $31, %eax ; AVX512F-32-NEXT: kshiftlq $1, %k0, %k0 @@ -3283,12 +3279,12 @@ define i64 @test_mask_x86_avx512_ucmp_b_512(<64 x i8> %a0, <64 x i8> %a1, i64 %m ; AVX512F-32-NEXT: kmovd %eax, %k1 ; AVX512F-32-NEXT: kshiftlq $63, %k1, %k1 ; AVX512F-32-NEXT: korq %k1, %k0, %k1 -; AVX512F-32-NEXT: vpcmpeqb %zmm6, %zmm5, %k0 {%k1} -; AVX512F-32-NEXT: vpcmpltub %zmm6, %zmm5, %k2 {%k1} -; AVX512F-32-NEXT: vpcmpleub %zmm6, %zmm5, %k3 {%k1} -; AVX512F-32-NEXT: vpcmpneqb %zmm6, %zmm5, %k4 {%k1} -; AVX512F-32-NEXT: vpcmpnltub %zmm6, %zmm5, %k5 {%k1} -; AVX512F-32-NEXT: vpcmpnleub %zmm6, %zmm5, %k1 {%k1} +; AVX512F-32-NEXT: vpcmpeqb %zmm1, %zmm0, %k0 {%k1} +; AVX512F-32-NEXT: vpcmpltub %zmm1, %zmm0, %k2 {%k1} +; AVX512F-32-NEXT: vpcmpleub %zmm1, %zmm0, %k3 {%k1} +; AVX512F-32-NEXT: vpcmpneqb %zmm1, %zmm0, %k4 {%k1} +; AVX512F-32-NEXT: vpcmpnltub %zmm1, %zmm0, %k5 {%k1} +; AVX512F-32-NEXT: vpcmpnleub %zmm1, %zmm0, %k1 {%k1} ; AVX512F-32-NEXT: kmovq %k0, (%esp) ; AVX512F-32-NEXT: movl (%esp), %eax ; AVX512F-32-NEXT: movl {{[0-9]+}}(%esp), %edx diff --git a/llvm/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll b/llvm/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll index f4504ed07fc..9f344c82f6b 100644 --- a/llvm/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll +++ b/llvm/test/CodeGen/X86/avx512bwvl-intrinsics-upgrade.ll @@ -2695,32 +2695,32 @@ declare <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16>, <16 x i16>, <32 define <8 x i32> @test_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1) { ; CHECK-LABEL: test_cmp_b_256: ; CHECK: ## BB#0: -; CHECK-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1] -; CHECK-NEXT: kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0] -; CHECK-NEXT: vpcmpgtb %ymm0, %ymm1, %k0 ## encoding: [0x62,0xf1,0x75,0x28,0x64,0xc0] -; CHECK-NEXT: kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8] -; CHECK-NEXT: vpcmpleb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x02] -; CHECK-NEXT: kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0] ; CHECK-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04] -; CHECK-NEXT: kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0] +; CHECK-NEXT: kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0] ; CHECK-NEXT: vpcmpleb %ymm0, %ymm1, %k0 ## encoding: [0x62,0xf3,0x75,0x28,0x3f,0xc0,0x02] -; CHECK-NEXT: kmovd %k0, %edi ## encoding: [0xc5,0xfb,0x93,0xf8] +; CHECK-NEXT: kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8] ; CHECK-NEXT: vpcmpgtb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1] -; CHECK-NEXT: kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0] -; CHECK-NEXT: vmovd %esi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6] -; CHECK-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01] -; CHECK-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x02] +; CHECK-NEXT: kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0] +; CHECK-NEXT: vmovd %eax, %xmm2 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd0] +; CHECK-NEXT: vpinsrd $1, %ecx, %xmm2, %xmm2 ## encoding: [0xc4,0xe3,0x69,0x22,0xd1,0x01] +; CHECK-NEXT: vpinsrd $2, %edx, %xmm2, %xmm2 ## encoding: [0xc4,0xe3,0x69,0x22,0xd2,0x02] ; CHECK-NEXT: kxnord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x46,0xc0] ; CHECK-NEXT: kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0] -; CHECK-NEXT: vpinsrd $3, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x03] -; CHECK-NEXT: vmovd %ecx, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9] -; CHECK-NEXT: vmovd %r8d, %xmm2 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0] -; CHECK-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9] -; CHECK-NEXT: ## xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] -; CHECK-NEXT: vmovd %edx, %xmm2 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2] -; CHECK-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca] -; CHECK-NEXT: ## xmm1 = xmm1[0],xmm2[0] -; CHECK-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01] +; CHECK-NEXT: vpinsrd $3, %eax, %xmm2, %xmm2 ## encoding: [0xc4,0xe3,0x69,0x22,0xd0,0x03] +; CHECK-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1] +; CHECK-NEXT: kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0] +; CHECK-NEXT: vpcmpgtb %ymm0, %ymm1, %k0 ## encoding: [0x62,0xf1,0x75,0x28,0x64,0xc0] +; CHECK-NEXT: kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8] +; CHECK-NEXT: vpcmpleb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x02] +; CHECK-NEXT: kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0] +; CHECK-NEXT: vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1] +; CHECK-NEXT: vmovd %eax, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc8] +; CHECK-NEXT: vpunpckldq %xmm0, %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x62,0xc0] +; CHECK-NEXT: ## xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] +; CHECK-NEXT: vmovd %edx, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xca] +; CHECK-NEXT: vpunpcklqdq %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6c,0xc1] +; CHECK-NEXT: ## xmm0 = xmm0[0],xmm1[0] +; CHECK-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x38,0xc2,0x01] ; CHECK-NEXT: retq ## encoding: [0xc3] %res0 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 -1) %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0 @@ -2750,23 +2750,23 @@ define <8 x i32> @test_mask_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1, i32 %mask) { ; CHECK-NEXT: vpcmpgtb %ymm0, %ymm1, %k0 {%k1} ## encoding: [0x62,0xf1,0x75,0x29,0x64,0xc0] ; CHECK-NEXT: kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8] ; CHECK-NEXT: vpcmpleb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x02] -; CHECK-NEXT: kmovd %k0, %r10d ## encoding: [0xc5,0x7b,0x93,0xd0] -; CHECK-NEXT: kxord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x47,0xc0] -; CHECK-NEXT: kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0] +; CHECK-NEXT: kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0] ; CHECK-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04] -; CHECK-NEXT: kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0] +; CHECK-NEXT: kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0] ; CHECK-NEXT: vpcmpleb %ymm0, %ymm1, %k0 {%k1} ## encoding: [0x62,0xf3,0x75,0x29,0x3f,0xc0,0x02] -; CHECK-NEXT: kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8] +; CHECK-NEXT: kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0] ; CHECK-NEXT: vpcmpgtb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x64,0xc1] -; CHECK-NEXT: kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0] -; CHECK-NEXT: vmovd %eax, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0] -; CHECK-NEXT: vpinsrd $1, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x01] -; CHECK-NEXT: vpinsrd $2, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x02] +; CHECK-NEXT: kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8] +; CHECK-NEXT: vmovd %esi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6] +; CHECK-NEXT: vpinsrd $1, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01] +; CHECK-NEXT: vpinsrd $2, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x02] ; CHECK-NEXT: vpinsrd $3, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x03] ; CHECK-NEXT: vmovd %r8d, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xc8] ; CHECK-NEXT: vpinsrd $1, %r9d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xc9,0x01] -; CHECK-NEXT: vpinsrd $2, %r10d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xca,0x02] -; CHECK-NEXT: vpinsrd $3, %esi, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xce,0x03] +; CHECK-NEXT: vpinsrd $2, %edx, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xca,0x02] +; CHECK-NEXT: kxord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x47,0xc0] +; CHECK-NEXT: kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0] +; CHECK-NEXT: vpinsrd $3, %eax, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xc8,0x03] ; CHECK-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01] ; CHECK-NEXT: retq ## encoding: [0xc3] %res0 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 %mask) @@ -2793,32 +2793,32 @@ declare i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8>, <32 x i8>, i32, i32) noun define <8 x i32> @test_ucmp_b_256(<32 x i8> %a0, <32 x i8> %a1) { ; CHECK-LABEL: test_ucmp_b_256: ; CHECK: ## BB#0: -; CHECK-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1] -; CHECK-NEXT: kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0] -; CHECK-NEXT: vpcmpltub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x01] -; CHECK-NEXT: kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8] -; CHECK-NEXT: vpcmpleub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x02] -; CHECK-NEXT: kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0] ; CHECK-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04] -; CHECK-NEXT: kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0] +; CHECK-NEXT: kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0] ; CHECK-NEXT: vpcmpnltub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x05] -; CHECK-NEXT: kmovd %k0, %edi ## encoding: [0xc5,0xfb,0x93,0xf8] +; CHECK-NEXT: kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8] ; CHECK-NEXT: vpcmpnleub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x06] -; CHECK-NEXT: kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0] -; CHECK-NEXT: vmovd %esi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6] -; CHECK-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01] -; CHECK-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x02] +; CHECK-NEXT: kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0] +; CHECK-NEXT: vmovd %eax, %xmm2 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd0] +; CHECK-NEXT: vpinsrd $1, %ecx, %xmm2, %xmm2 ## encoding: [0xc4,0xe3,0x69,0x22,0xd1,0x01] +; CHECK-NEXT: vpinsrd $2, %edx, %xmm2, %xmm2 ## encoding: [0xc4,0xe3,0x69,0x22,0xd2,0x02] ; CHECK-NEXT: kxnord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x46,0xc0] ; CHECK-NEXT: kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0] -; CHECK-NEXT: vpinsrd $3, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x03] -; CHECK-NEXT: vmovd %ecx, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9] -; CHECK-NEXT: vmovd %r8d, %xmm2 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0] -; CHECK-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9] -; CHECK-NEXT: ## xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] -; CHECK-NEXT: vmovd %edx, %xmm2 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2] -; CHECK-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca] -; CHECK-NEXT: ## xmm1 = xmm1[0],xmm2[0] -; CHECK-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01] +; CHECK-NEXT: vpinsrd $3, %eax, %xmm2, %xmm2 ## encoding: [0xc4,0xe3,0x69,0x22,0xd0,0x03] +; CHECK-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1] +; CHECK-NEXT: kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0] +; CHECK-NEXT: vpcmpltub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x01] +; CHECK-NEXT: kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8] +; CHECK-NEXT: vpcmpleub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x02] +; CHECK-NEXT: kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0] +; CHECK-NEXT: vmovd %ecx, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc1] +; CHECK-NEXT: vmovd %eax, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc8] +; CHECK-NEXT: vpunpckldq %xmm0, %xmm1, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x62,0xc0] +; CHECK-NEXT: ## xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] +; CHECK-NEXT: vmovd %edx, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xca] +; CHECK-NEXT: vpunpcklqdq %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6c,0xc1] +; CHECK-NEXT: ## xmm0 = xmm0[0],xmm1[0] +; CHECK-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x38,0xc2,0x01] ; CHECK-NEXT: retq ## encoding: [0xc3] %res0 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 -1) %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0 @@ -2848,23 +2848,23 @@ define <8 x i32> @test_mask_ucmp_b_256(<32 x i8> %a0, <32 x i8> %a1, i32 %mask) ; CHECK-NEXT: vpcmpltub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x01] ; CHECK-NEXT: kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8] ; CHECK-NEXT: vpcmpleub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x02] -; CHECK-NEXT: kmovd %k0, %r10d ## encoding: [0xc5,0x7b,0x93,0xd0] -; CHECK-NEXT: kxord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x47,0xc0] -; CHECK-NEXT: kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0] +; CHECK-NEXT: kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0] ; CHECK-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04] -; CHECK-NEXT: kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0] +; CHECK-NEXT: kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0] ; CHECK-NEXT: vpcmpnltub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x05] -; CHECK-NEXT: kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8] +; CHECK-NEXT: kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0] ; CHECK-NEXT: vpcmpnleub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x06] -; CHECK-NEXT: kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0] -; CHECK-NEXT: vmovd %eax, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc0] -; CHECK-NEXT: vpinsrd $1, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x01] -; CHECK-NEXT: vpinsrd $2, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x02] +; CHECK-NEXT: kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8] +; CHECK-NEXT: vmovd %esi, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6] +; CHECK-NEXT: vpinsrd $1, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01] +; CHECK-NEXT: vpinsrd $2, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x02] ; CHECK-NEXT: vpinsrd $3, %edi, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x03] ; CHECK-NEXT: vmovd %r8d, %xmm1 ## EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xc8] ; CHECK-NEXT: vpinsrd $1, %r9d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xc9,0x01] -; CHECK-NEXT: vpinsrd $2, %r10d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xca,0x02] -; CHECK-NEXT: vpinsrd $3, %esi, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xce,0x03] +; CHECK-NEXT: vpinsrd $2, %edx, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xca,0x02] +; CHECK-NEXT: kxord %k0, %k0, %k0 ## encoding: [0xc4,0xe1,0xfd,0x47,0xc0] +; CHECK-NEXT: kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0] +; CHECK-NEXT: vpinsrd $3, %eax, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xc8,0x03] ; CHECK-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01] ; CHECK-NEXT: retq ## encoding: [0xc3] %res0 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 %mask) diff --git a/llvm/test/CodeGen/X86/bitcast-and-setcc-256.ll b/llvm/test/CodeGen/X86/bitcast-and-setcc-256.ll index a6d6ca15530..21b06915182 100644 --- a/llvm/test/CodeGen/X86/bitcast-and-setcc-256.ll +++ b/llvm/test/CodeGen/X86/bitcast-and-setcc-256.ll @@ -453,10 +453,10 @@ define i32 @v32i8(<32 x i8> %a, <32 x i8> %b, <32 x i8> %c, <32 x i8> %d) { ; SSE2-SSSE3-NEXT: pcmpgtb %xmm2, %xmm0 ; SSE2-SSSE3-NEXT: pcmpgtb %xmm3, %xmm1 ; SSE2-SSSE3-NEXT: pcmpgtb %xmm6, %xmm4 -; SSE2-SSSE3-NEXT: pand %xmm0, %xmm4 ; SSE2-SSSE3-NEXT: pcmpgtb %xmm7, %xmm5 ; SSE2-SSSE3-NEXT: pand %xmm1, %xmm5 ; SSE2-SSSE3-NEXT: movdqa %xmm5, -{{[0-9]+}}(%rsp) +; SSE2-SSSE3-NEXT: pand %xmm0, %xmm4 ; SSE2-SSSE3-NEXT: movdqa %xmm4, -{{[0-9]+}}(%rsp) ; SSE2-SSSE3-NEXT: movb -{{[0-9]+}}(%rsp), %al ; SSE2-SSSE3-NEXT: andb $1, %al diff --git a/llvm/test/CodeGen/X86/extractelement-legalization-store-ordering.ll b/llvm/test/CodeGen/X86/extractelement-legalization-store-ordering.ll index 5d5cbc76f92..2ebd30adcdc 100644 --- a/llvm/test/CodeGen/X86/extractelement-legalization-store-ordering.ll +++ b/llvm/test/CodeGen/X86/extractelement-legalization-store-ordering.ll @@ -1,3 +1,4 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py ; RUN: llc < %s -mtriple i386-apple-darwin -mcpu=yonah | FileCheck %s target datalayout = "e-m:o-p:32:32-f64:32:64-f80:128-n8:16:32-S128" @@ -6,31 +7,32 @@ target datalayout = "e-m:o-p:32:32-f64:32:64-f80:128-n8:16:32-S128" ; into loads, off the stack or a previous store. ; Be very explicit about the ordering/stack offsets. -; CHECK-LABEL: test_extractelement_legalization_storereuse: -; CHECK: # BB#0 -; CHECK-NEXT: pushl %ebx -; CHECK-NEXT: pushl %edi -; CHECK-NEXT: pushl %esi -; CHECK-NEXT: movl 16(%esp), %eax -; CHECK-NEXT: movl 24(%esp), %ecx -; CHECK-NEXT: movl 20(%esp), %edx -; CHECK-NEXT: paddd (%edx), %xmm0 -; CHECK-NEXT: movdqa %xmm0, (%edx) -; CHECK-NEXT: movl (%edx), %esi -; CHECK-NEXT: movl 4(%edx), %edi -; CHECK-NEXT: shll $4, %ecx -; CHECK-NEXT: movl 8(%edx), %ebx -; CHECK-NEXT: movl 12(%edx), %edx -; CHECK-NEXT: movl %esi, 12(%eax,%ecx) -; CHECK-NEXT: movl %edi, (%eax,%ecx) -; CHECK-NEXT: movl %ebx, 8(%eax,%ecx) -; CHECK-NEXT: movl %edx, 4(%eax,%ecx) -; CHECK-NEXT: popl %esi -; CHECK-NEXT: popl %edi -; CHECK-NEXT: popl %ebx -; CHECK-NEXT: retl define void @test_extractelement_legalization_storereuse(<4 x i32> %a, i32* nocapture %x, i32* nocapture readonly %y, i32 %i) #0 { +; CHECK-LABEL: _test_extractelement_legalization_storereuse: ## @test_extractelement_legalization_storereuse +; CHECK: ## BB#0: ## %entry +; CHECK-NEXT: pushl %ebx +; CHECK-NEXT: pushl %edi +; CHECK-NEXT: pushl %esi +; CHECK-NEXT: movl {{[0-9]+}}(%esp), %eax +; CHECK-NEXT: movl {{[0-9]+}}(%esp), %ecx +; CHECK-NEXT: paddd (%ecx), %xmm0 +; CHECK-NEXT: movl {{[0-9]+}}(%esp), %edx +; CHECK-NEXT: movdqa %xmm0, (%ecx) +; CHECK-NEXT: movl (%ecx), %esi +; CHECK-NEXT: movl 4(%ecx), %edi +; CHECK-NEXT: shll $4, %edx +; CHECK-NEXT: movl 8(%ecx), %ebx +; CHECK-NEXT: movl 12(%ecx), %ecx +; CHECK-NEXT: movl %esi, 12(%eax,%edx) +; CHECK-NEXT: movl %edi, (%eax,%edx) +; CHECK-NEXT: movl %ebx, 8(%eax,%edx) +; CHECK-NEXT: movl %ecx, 4(%eax,%edx) +; CHECK-NEXT: popl %esi +; CHECK-NEXT: popl %edi +; CHECK-NEXT: popl %ebx +; CHECK-NEXT: retl +; CHECK-NEXT: ## -- End function entry: %0 = bitcast i32* %y to <4 x i32>* %1 = load <4 x i32>, <4 x i32>* %0, align 16 diff --git a/llvm/test/CodeGen/X86/fp128-i128.ll b/llvm/test/CodeGen/X86/fp128-i128.ll index 6c6bc8bdc1d..98082ec611d 100644 --- a/llvm/test/CodeGen/X86/fp128-i128.ll +++ b/llvm/test/CodeGen/X86/fp128-i128.ll @@ -50,8 +50,8 @@ define void @TestUnionLD1(fp128 %s, i64 %n) #0 { ; CHECK-NEXT: andq %rdi, %rcx ; CHECK-NEXT: movabsq $-281474976710656, %rdx # imm = 0xFFFF000000000000 ; CHECK-NEXT: andq -{{[0-9]+}}(%rsp), %rdx -; CHECK-NEXT: orq %rcx, %rdx ; CHECK-NEXT: movq %rax, -{{[0-9]+}}(%rsp) +; CHECK-NEXT: orq %rcx, %rdx ; CHECK-NEXT: movq %rdx, -{{[0-9]+}}(%rsp) ; CHECK-NEXT: movaps -{{[0-9]+}}(%rsp), %xmm0 ; CHECK-NEXT: jmp foo # TAILCALL diff --git a/llvm/test/CodeGen/X86/gather-addresses.ll b/llvm/test/CodeGen/X86/gather-addresses.ll index c3109673468..3383b6e2f4c 100644 --- a/llvm/test/CodeGen/X86/gather-addresses.ll +++ b/llvm/test/CodeGen/X86/gather-addresses.ll @@ -16,11 +16,10 @@ ; LIN: sarq $32, %r[[REG2]] ; LIN: movslq %e[[REG4]], %r[[REG3:.+]] ; LIN: sarq $32, %r[[REG4]] -; LIN: movsd (%rdi,%r[[REG1]],8), %xmm0 -; LIN: movhpd (%rdi,%r[[REG2]],8), %xmm0 -; LIN: movsd (%rdi,%r[[REG3]],8), %xmm1 -; LIN: movhpd (%rdi,%r[[REG4]],8), %xmm1 - +; LIN: movsd (%rdi,%rsi,8), %xmm1 +; LIN: movhpd (%rdi,%rax,8), %xmm1 +; LIN: movdqa (%rsi), %xmm0 +; LIN: movq %rdi, %xmm1 ; WIN: movdqa (%rdx), %xmm0 ; WIN: pand (%r8), %xmm0 ; WIN: pextrq $1, %xmm0, %r[[REG4:.+]] @@ -29,10 +28,10 @@ ; WIN: sarq $32, %r[[REG2]] ; WIN: movslq %e[[REG4]], %r[[REG3:.+]] ; WIN: sarq $32, %r[[REG4]] -; WIN: movsd (%rcx,%r[[REG1]],8), %xmm0 -; WIN: movhpd (%rcx,%r[[REG2]],8), %xmm0 -; WIN: movsd (%rcx,%r[[REG3]],8), %xmm1 -; WIN: movhpd (%rcx,%r[[REG4]],8), %xmm1 +; WIN: movsd (%rcx,%r9,8), %xmm1 +; WIN: movhpd (%rcx,%rax,8), %xmm1 +; WIN: movdqa (%rdx), %xmm0 +; WIN: movq %rdx, %xmm1 define <4 x double> @foo(double* %p, <4 x i32>* %i, <4 x i32>* %h) nounwind { %a = load <4 x i32>, <4 x i32>* %i diff --git a/llvm/test/CodeGen/X86/half.ll b/llvm/test/CodeGen/X86/half.ll index 4c8003f0c51..f72dfa1eef5 100644 --- a/llvm/test/CodeGen/X86/half.ll +++ b/llvm/test/CodeGen/X86/half.ll @@ -1,266 +1,834 @@ -; RUN: llc < %s -march=x86-64 -mtriple=x86_64-unknown-linux-gnu -mcpu=corei7 -mattr=-f16c -asm-verbose=false -fixup-byte-word-insts=1 \ -; RUN: | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-LIBCALL -check-prefix=BWON -; RUN: llc < %s -march=x86-64 -mtriple=x86_64-unknown-linux-gnu -mcpu=corei7 -mattr=-f16c -asm-verbose=false -fixup-byte-word-insts=0 \ -; RUN: | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-LIBCALL -check-prefix=BWOFF -; RUN: llc < %s -march=x86-64 -mtriple=x86_64-unknown-linux-gnu -mcpu=corei7 -mattr=+f16c -asm-verbose=false -fixup-byte-word-insts=1 \ -; RUN: | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-F16C -check-prefix=BWON -; RUN: llc < %s -mtriple=i686-unknown-linux-gnu -mattr +sse2 -asm-verbose=false -fixup-byte-word-insts=0 \ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: llc < %s -march=x86-64 -mtriple=x86_64-unknown-linux-gnu -mcpu=corei7 -mattr=-f16c -fixup-byte-word-insts=1 \ +; RUN: | FileCheck %s -check-prefixes=CHECK,CHECK-LIBCALL,BWON,NOF16-BWINSTS +; RUN: llc < %s -march=x86-64 -mtriple=x86_64-unknown-linux-gnu -mcpu=corei7 -mattr=-f16c -fixup-byte-word-insts=0 \ +; RUN: | FileCheck %s -check-prefixes=CHECK,CHECK-LIBCALL,BWOFF,NOF16-NOBWINSTS +; RUN: llc < %s -march=x86-64 -mtriple=x86_64-unknown-linux-gnu -mcpu=corei7 -mattr=+f16c -fixup-byte-word-insts=1 \ +; RUN: | FileCheck %s -check-prefixes=CHECK,BWON,CHECK-F16C +; RUN: llc < %s -mtriple=i686-unknown-linux-gnu -mattr +sse2 -fixup-byte-word-insts=0 \ ; RUN: | FileCheck %s -check-prefix=CHECK-I686 -define void @test_load_store(half* %in, half* %out) { -; CHECK-LABEL: test_load_store: -; BWON: movzwl (%rdi), %eax -; BWOFF: movw (%rdi), %ax -; CHECK: movw %ax, (%rsi) +define void @test_load_store(half* %in, half* %out) #0 { +; BWON-LABEL: test_load_store: +; BWON: # BB#0: +; BWON-NEXT: movzwl (%rdi), %eax +; BWON-NEXT: movw %ax, (%rsi) +; BWON-NEXT: retq +; +; BWOFF-LABEL: test_load_store: +; BWOFF: # BB#0: +; BWOFF-NEXT: movw (%rdi), %ax +; BWOFF-NEXT: movw %ax, (%rsi) +; BWOFF-NEXT: retq +; +; CHECK-I686-LABEL: test_load_store: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %eax +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %ecx +; CHECK-I686-NEXT: movw (%ecx), %cx +; CHECK-I686-NEXT: movw %cx, (%eax) +; CHECK-I686-NEXT: retl %val = load half, half* %in store half %val, half* %out ret void } -define i16 @test_bitcast_from_half(half* %addr) { -; CHECK-LABEL: test_bitcast_from_half: -; BWON: movzwl (%rdi), %eax -; BWOFF: movw (%rdi), %ax +define i16 @test_bitcast_from_half(half* %addr) #0 { +; BWON-LABEL: test_bitcast_from_half: +; BWON: # BB#0: +; BWON-NEXT: movzwl (%rdi), %eax +; BWON-NEXT: retq +; +; BWOFF-LABEL: test_bitcast_from_half: +; BWOFF: # BB#0: +; BWOFF-NEXT: movw (%rdi), %ax +; BWOFF-NEXT: retq +; +; CHECK-I686-LABEL: test_bitcast_from_half: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %eax +; CHECK-I686-NEXT: movw (%eax), %ax +; CHECK-I686-NEXT: retl %val = load half, half* %addr %val_int = bitcast half %val to i16 ret i16 %val_int } -define void @test_bitcast_to_half(half* %addr, i16 %in) { +define void @test_bitcast_to_half(half* %addr, i16 %in) #0 { ; CHECK-LABEL: test_bitcast_to_half: -; CHECK: movw %si, (%rdi) +; CHECK: # BB#0: +; CHECK-NEXT: movw %si, (%rdi) +; CHECK-NEXT: retq +; +; CHECK-I686-LABEL: test_bitcast_to_half: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: movw {{[0-9]+}}(%esp), %ax +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %ecx +; CHECK-I686-NEXT: movw %ax, (%ecx) +; CHECK-I686-NEXT: retl %val_fp = bitcast i16 %in to half store half %val_fp, half* %addr ret void } -define float @test_extend32(half* %addr) { -; CHECK-LABEL: test_extend32: - -; CHECK-LIBCALL: jmp __gnu_h2f_ieee -; CHECK-F16C: vcvtph2ps +define float @test_extend32(half* %addr) #0 { +; CHECK-LIBCALL-LABEL: test_extend32: +; CHECK-LIBCALL: # BB#0: +; CHECK-LIBCALL-NEXT: movzwl (%rdi), %edi +; CHECK-LIBCALL-NEXT: jmp __gnu_h2f_ieee # TAILCALL +; +; CHECK-F16C-LABEL: test_extend32: +; CHECK-F16C: # BB#0: +; CHECK-F16C-NEXT: movswl (%rdi), %eax +; CHECK-F16C-NEXT: vmovd %eax, %xmm0 +; CHECK-F16C-NEXT: vcvtph2ps %xmm0, %xmm0 +; CHECK-F16C-NEXT: retq +; +; CHECK-I686-LABEL: test_extend32: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: subl $12, %esp +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %eax +; CHECK-I686-NEXT: movzwl (%eax), %eax +; CHECK-I686-NEXT: movl %eax, (%esp) +; CHECK-I686-NEXT: calll __gnu_h2f_ieee +; CHECK-I686-NEXT: addl $12, %esp +; CHECK-I686-NEXT: retl %val16 = load half, half* %addr %val32 = fpext half %val16 to float ret float %val32 } -define double @test_extend64(half* %addr) { -; CHECK-LABEL: test_extend64: - -; CHECK-LIBCALL: callq __gnu_h2f_ieee -; CHECK-LIBCALL: cvtss2sd -; CHECK-F16C: vcvtph2ps -; CHECK-F16C: vcvtss2sd +define double @test_extend64(half* %addr) #0 { +; CHECK-LIBCALL-LABEL: test_extend64: +; CHECK-LIBCALL: # BB#0: +; CHECK-LIBCALL-NEXT: pushq %rax +; CHECK-LIBCALL-NEXT: movzwl (%rdi), %edi +; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee +; CHECK-LIBCALL-NEXT: cvtss2sd %xmm0, %xmm0 +; CHECK-LIBCALL-NEXT: popq %rax +; CHECK-LIBCALL-NEXT: retq +; +; CHECK-F16C-LABEL: test_extend64: +; CHECK-F16C: # BB#0: +; CHECK-F16C-NEXT: movswl (%rdi), %eax +; CHECK-F16C-NEXT: vmovd %eax, %xmm0 +; CHECK-F16C-NEXT: vcvtph2ps %xmm0, %xmm0 +; CHECK-F16C-NEXT: vcvtss2sd %xmm0, %xmm0, %xmm0 +; CHECK-F16C-NEXT: retq +; +; CHECK-I686-LABEL: test_extend64: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: subl $12, %esp +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %eax +; CHECK-I686-NEXT: movzwl (%eax), %eax +; CHECK-I686-NEXT: movl %eax, (%esp) +; CHECK-I686-NEXT: calll __gnu_h2f_ieee +; CHECK-I686-NEXT: addl $12, %esp +; CHECK-I686-NEXT: retl %val16 = load half, half* %addr %val32 = fpext half %val16 to double ret double %val32 } -define void @test_trunc32(float %in, half* %addr) { -; CHECK-LABEL: test_trunc32: - -; CHECK-LIBCALL: callq __gnu_f2h_ieee -; CHECK-F16C: vcvtps2ph +define void @test_trunc32(float %in, half* %addr) #0 { +; CHECK-LIBCALL-LABEL: test_trunc32: +; CHECK-LIBCALL: # BB#0: +; CHECK-LIBCALL-NEXT: pushq %rbx +; CHECK-LIBCALL-NEXT: movq %rdi, %rbx +; CHECK-LIBCALL-NEXT: callq __gnu_f2h_ieee +; CHECK-LIBCALL-NEXT: movw %ax, (%rbx) +; CHECK-LIBCALL-NEXT: popq %rbx +; CHECK-LIBCALL-NEXT: retq +; +; CHECK-F16C-LABEL: test_trunc32: +; CHECK-F16C: # BB#0: +; CHECK-F16C-NEXT: vcvtps2ph $4, %xmm0, %xmm0 +; CHECK-F16C-NEXT: vmovd %xmm0, %eax +; CHECK-F16C-NEXT: movw %ax, (%rdi) +; CHECK-F16C-NEXT: retq +; +; CHECK-I686-LABEL: test_trunc32: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: pushl %esi +; CHECK-I686-NEXT: subl $8, %esp +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %esi +; CHECK-I686-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero +; CHECK-I686-NEXT: movss %xmm0, (%esp) +; CHECK-I686-NEXT: calll __gnu_f2h_ieee +; CHECK-I686-NEXT: movw %ax, (%esi) +; CHECK-I686-NEXT: addl $8, %esp +; CHECK-I686-NEXT: popl %esi +; CHECK-I686-NEXT: retl %val16 = fptrunc float %in to half store half %val16, half* %addr ret void } -define void @test_trunc64(double %in, half* %addr) { +define void @test_trunc64(double %in, half* %addr) #0 { ; CHECK-LABEL: test_trunc64: - -; CHECK-LIBCALL: callq __truncdfhf2 -; CHECK-F16C: callq __truncdfhf2 +; CHECK: # BB#0: +; CHECK-NEXT: pushq %rbx +; CHECK-NEXT: movq %rdi, %rbx +; CHECK-NEXT: callq __truncdfhf2 +; CHECK-NEXT: movw %ax, (%rbx) +; CHECK-NEXT: popq %rbx +; CHECK-NEXT: retq +; +; CHECK-I686-LABEL: test_trunc64: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: pushl %esi +; CHECK-I686-NEXT: subl $8, %esp +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %esi +; CHECK-I686-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero +; CHECK-I686-NEXT: movsd %xmm0, (%esp) +; CHECK-I686-NEXT: calll __truncdfhf2 +; CHECK-I686-NEXT: movw %ax, (%esi) +; CHECK-I686-NEXT: addl $8, %esp +; CHECK-I686-NEXT: popl %esi +; CHECK-I686-NEXT: retl %val16 = fptrunc double %in to half store half %val16, half* %addr ret void } define i64 @test_fptosi_i64(half* %p) #0 { -; CHECK-LABEL: test_fptosi_i64: - -; CHECK-LIBCALL-NEXT: pushq %rax -; CHECK-LIBCALL-NEXT: movzwl (%rdi), %edi -; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee -; CHECK-LIBCALL-NEXT: cvttss2si %xmm0, %rax -; CHECK-LIBCALL-NEXT: popq %rcx -; CHECK-LIBCALL-NEXT: retq - -; CHECK-F16C-NEXT: movswl (%rdi), [[REG0:%[a-z0-9]+]] -; CHECK-F16C-NEXT: vmovd [[REG0]], [[REG1:%[a-z0-9]+]] -; CHECK-F16C-NEXT: vcvtph2ps [[REG1]], [[REG2:%[a-z0-9]+]] -; CHECK-F16C-NEXT: vcvttss2si [[REG2]], %rax -; CHECK-F16C-NEXT: retq +; CHECK-LIBCALL-LABEL: test_fptosi_i64: +; CHECK-LIBCALL: # BB#0: +; CHECK-LIBCALL-NEXT: pushq %rax +; CHECK-LIBCALL-NEXT: movzwl (%rdi), %edi +; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee +; CHECK-LIBCALL-NEXT: cvttss2si %xmm0, %rax +; CHECK-LIBCALL-NEXT: popq %rcx +; CHECK-LIBCALL-NEXT: retq +; +; CHECK-F16C-LABEL: test_fptosi_i64: +; CHECK-F16C: # BB#0: +; CHECK-F16C-NEXT: movswl (%rdi), %eax +; CHECK-F16C-NEXT: vmovd %eax, %xmm0 +; CHECK-F16C-NEXT: vcvtph2ps %xmm0, %xmm0 +; CHECK-F16C-NEXT: vcvttss2si %xmm0, %rax +; CHECK-F16C-NEXT: retq +; +; CHECK-I686-LABEL: test_fptosi_i64: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: subl $12, %esp +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %eax +; CHECK-I686-NEXT: movzwl (%eax), %eax +; CHECK-I686-NEXT: movl %eax, (%esp) +; CHECK-I686-NEXT: calll __gnu_h2f_ieee +; CHECK-I686-NEXT: fstps (%esp) +; CHECK-I686-NEXT: calll __fixsfdi +; CHECK-I686-NEXT: addl $12, %esp +; CHECK-I686-NEXT: retl %a = load half, half* %p, align 2 %r = fptosi half %a to i64 ret i64 %r } define void @test_sitofp_i64(i64 %a, half* %p) #0 { -; CHECK-LABEL: test_sitofp_i64: - -; CHECK-LIBCALL-NEXT: pushq [[ADDR:%[a-z]+]] -; CHECK-LIBCALL-NEXT: movq %rsi, [[ADDR]] -; CHECK-LIBCALL-NEXT: cvtsi2ssq %rdi, %xmm0 -; CHECK-LIBCALL-NEXT: callq __gnu_f2h_ieee -; CHECK-LIBCALL-NEXT: movw %ax, ([[ADDR]]) -; CHECK_LIBCALL-NEXT: popq [[ADDR]] -; CHECK_LIBCALL-NEXT: retq - -; CHECK-F16C-NEXT: vcvtsi2ssq %rdi, [[REG0:%[a-z0-9]+]], [[REG0]] -; CHECK-F16C-NEXT: vcvtps2ph $4, [[REG0]], [[REG0]] -; CHECK-F16C-NEXT: vmovd [[REG0]], %eax -; CHECK-F16C-NEXT: movw %ax, (%rsi) -; CHECK-F16C-NEXT: retq +; CHECK-LIBCALL-LABEL: test_sitofp_i64: +; CHECK-LIBCALL: # BB#0: +; CHECK-LIBCALL-NEXT: pushq %rbx +; CHECK-LIBCALL-NEXT: movq %rsi, %rbx +; CHECK-LIBCALL-NEXT: cvtsi2ssq %rdi, %xmm0 +; CHECK-LIBCALL-NEXT: callq __gnu_f2h_ieee +; CHECK-LIBCALL-NEXT: movw %ax, (%rbx) +; CHECK-LIBCALL-NEXT: popq %rbx +; CHECK-LIBCALL-NEXT: retq +; +; CHECK-F16C-LABEL: test_sitofp_i64: +; CHECK-F16C: # BB#0: +; CHECK-F16C-NEXT: vcvtsi2ssq %rdi, %xmm0, %xmm0 +; CHECK-F16C-NEXT: vcvtps2ph $4, %xmm0, %xmm0 +; CHECK-F16C-NEXT: vmovd %xmm0, %eax +; CHECK-F16C-NEXT: movw %ax, (%rsi) +; CHECK-F16C-NEXT: retq +; +; CHECK-I686-LABEL: test_sitofp_i64: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: pushl %esi +; CHECK-I686-NEXT: subl $24, %esp +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %esi +; CHECK-I686-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero +; CHECK-I686-NEXT: movlps %xmm0, {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: fildll {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: fstps {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero +; CHECK-I686-NEXT: movss %xmm0, (%esp) +; CHECK-I686-NEXT: calll __gnu_f2h_ieee +; CHECK-I686-NEXT: movw %ax, (%esi) +; CHECK-I686-NEXT: addl $24, %esp +; CHECK-I686-NEXT: popl %esi +; CHECK-I686-NEXT: retl %r = sitofp i64 %a to half store half %r, half* %p ret void } define i64 @test_fptoui_i64(half* %p) #0 { -; CHECK-LABEL: test_fptoui_i64: - -; FP_TO_UINT is expanded using FP_TO_SINT -; CHECK-LIBCALL-NEXT: pushq %rax -; CHECK-LIBCALL-NEXT: movzwl (%rdi), %edi -; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee -; CHECK-LIBCALL-NEXT: movss {{.[A-Z_0-9]+}}(%rip), [[REG1:%[a-z0-9]+]] -; CHECK-LIBCALL-NEXT: movaps %xmm0, [[REG2:%[a-z0-9]+]] -; CHECK-LIBCALL-NEXT: subss [[REG1]], [[REG2]] -; CHECK-LIBCALL-NEXT: cvttss2si [[REG2]], [[REG3:%[a-z0-9]+]] -; CHECK-LIBCALL-NEXT: movabsq $-9223372036854775808, [[REG4:%[a-z0-9]+]] -; CHECK-LIBCALL-NEXT: xorq [[REG3]], [[REG4]] -; CHECK-LIBCALL-NEXT: cvttss2si %xmm0, [[REG5:%[a-z0-9]+]] -; CHECK-LIBCALL-NEXT: ucomiss [[REG1]], %xmm0 -; CHECK-LIBCALL-NEXT: cmovaeq [[REG4]], [[REG5]] -; CHECK-LIBCALL-NEXT: popq %rcx -; CHECK-LIBCALL-NEXT: retq - -; CHECK-F16C-NEXT: movswl (%rdi), [[REG0:%[a-z0-9]+]] -; CHECK-F16C-NEXT: vmovd [[REG0]], [[REG1:%[a-z0-9]+]] -; CHECK-F16C-NEXT: vcvtph2ps [[REG1]], [[REG2:%[a-z0-9]+]] -; CHECK-F16C-NEXT: vmovss {{.[A-Z_0-9]+}}(%rip), [[REG3:%[a-z0-9]+]] -; CHECK-F16C-NEXT: vsubss [[REG3]], [[REG2]], [[REG4:%[a-z0-9]+]] -; CHECK-F16C-NEXT: vcvttss2si [[REG4]], [[REG5:%[a-z0-9]+]] -; CHECK-F16C-NEXT: movabsq $-9223372036854775808, [[REG6:%[a-z0-9]+]] -; CHECK-F16C-NEXT: xorq [[REG5]], [[REG6:%[a-z0-9]+]] -; CHECK-F16C-NEXT: vcvttss2si [[REG2]], [[REG7:%[a-z0-9]+]] -; CHECK-F16C-NEXT: vucomiss [[REG3]], [[REG2]] -; CHECK-F16C-NEXT: cmovaeq [[REG6]], %rax -; CHECK-F16C-NEXT: retq +; CHECK-LIBCALL-LABEL: test_fptoui_i64: +; CHECK-LIBCALL: # BB#0: +; CHECK-LIBCALL-NEXT: pushq %rax +; CHECK-LIBCALL-NEXT: movzwl (%rdi), %edi +; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee +; CHECK-LIBCALL-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero +; CHECK-LIBCALL-NEXT: movaps %xmm0, %xmm2 +; CHECK-LIBCALL-NEXT: subss %xmm1, %xmm2 +; CHECK-LIBCALL-NEXT: cvttss2si %xmm2, %rcx +; CHECK-LIBCALL-NEXT: movabsq $-9223372036854775808, %rdx # imm = 0x8000000000000000 +; CHECK-LIBCALL-NEXT: cvttss2si %xmm0, %rax +; CHECK-LIBCALL-NEXT: xorq %rcx, %rdx +; CHECK-LIBCALL-NEXT: ucomiss %xmm1, %xmm0 +; CHECK-LIBCALL-NEXT: cmovaeq %rdx, %rax +; CHECK-LIBCALL-NEXT: popq %rcx +; CHECK-LIBCALL-NEXT: retq +; +; CHECK-F16C-LABEL: test_fptoui_i64: +; CHECK-F16C: # BB#0: +; CHECK-F16C-NEXT: movswl (%rdi), %eax +; CHECK-F16C-NEXT: vmovd %eax, %xmm0 +; CHECK-F16C-NEXT: vcvtph2ps %xmm0, %xmm0 +; CHECK-F16C-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero +; CHECK-F16C-NEXT: vsubss %xmm1, %xmm0, %xmm2 +; CHECK-F16C-NEXT: vcvttss2si %xmm2, %rcx +; CHECK-F16C-NEXT: movabsq $-9223372036854775808, %rdx # imm = 0x8000000000000000 +; CHECK-F16C-NEXT: vcvttss2si %xmm0, %rax +; CHECK-F16C-NEXT: xorq %rcx, %rdx +; CHECK-F16C-NEXT: vucomiss %xmm1, %xmm0 +; CHECK-F16C-NEXT: cmovaeq %rdx, %rax +; CHECK-F16C-NEXT: retq +; +; CHECK-I686-LABEL: test_fptoui_i64: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: subl $12, %esp +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %eax +; CHECK-I686-NEXT: movzwl (%eax), %eax +; CHECK-I686-NEXT: movl %eax, (%esp) +; CHECK-I686-NEXT: calll __gnu_h2f_ieee +; CHECK-I686-NEXT: fstps (%esp) +; CHECK-I686-NEXT: calll __fixunssfdi +; CHECK-I686-NEXT: addl $12, %esp +; CHECK-I686-NEXT: retl %a = load half, half* %p, align 2 %r = fptoui half %a to i64 ret i64 %r } define void @test_uitofp_i64(i64 %a, half* %p) #0 { -; CHECK-LABEL: test_uitofp_i64: -; CHECK-LIBCALL-NEXT: pushq [[ADDR:%[a-z0-9]+]] -; CHECK-LIBCALL-NEXT: movq %rsi, [[ADDR]] -; CHECK-NEXT: testq %rdi, %rdi -; CHECK-NEXT: js [[LABEL1:.LBB[0-9_]+]] - -; simple conversion to float if non-negative -; CHECK-LIBCALL-NEXT: cvtsi2ssq %rdi, [[REG1:%[a-z0-9]+]] -; CHECK-F16C-NEXT: vcvtsi2ssq %rdi, [[REG1:%[a-z0-9]+]], [[REG1]] -; CHECK-NEXT: jmp [[LABEL2:.LBB[0-9_]+]] - -; convert using shift+or if negative -; CHECK-NEXT: [[LABEL1]]: -; CHECK-NEXT: movq %rdi, %rax -; CHECK-NEXT: shrq %rax -; CHECK-NEXT: andl $1, %edi -; CHECK-NEXT: orq %rax, [[REG2:%[a-z0-9]+]] -; CHECK-LIBCALL-NEXT: cvtsi2ssq [[REG2]], [[REG3:%[a-z0-9]+]] -; CHECK-LIBCALL-NEXT: addss [[REG3]], [[REG1]] -; CHECK-F16C-NEXT: vcvtsi2ssq [[REG2]], [[REG3:%[a-z0-9]+]], [[REG3]] -; CHECK-F16C-NEXT: vaddss [[REG3]], [[REG3]], [[REG1:[%a-z0-9]+]] - -; convert float to half -; CHECK-NEXT: [[LABEL2]]: -; CHECK-LIBCALL-NEXT: callq __gnu_f2h_ieee -; CHECK-LIBCALL-NEXT: movw %ax, ([[ADDR]]) -; CHECK-LIBCALL-NEXT: popq [[ADDR]] -; CHECK-F16C-NEXT: vcvtps2ph $4, [[REG1]], [[REG4:%[a-z0-9]+]] -; CHECK-F16C-NEXT: vmovd [[REG4]], %eax -; CHECK-F16C-NEXT: movw %ax, (%rsi) -; CHECK-NEXT: retq - +; CHECK-LIBCALL-LABEL: test_uitofp_i64: +; CHECK-LIBCALL: # BB#0: +; CHECK-LIBCALL-NEXT: pushq %rbx +; CHECK-LIBCALL-NEXT: movq %rsi, %rbx +; CHECK-LIBCALL-NEXT: testq %rdi, %rdi +; CHECK-LIBCALL-NEXT: js .LBB10_1 +; CHECK-LIBCALL-NEXT: # BB#2: +; CHECK-LIBCALL-NEXT: cvtsi2ssq %rdi, %xmm0 +; CHECK-LIBCALL-NEXT: jmp .LBB10_3 +; CHECK-LIBCALL-NEXT: .LBB10_1: +; CHECK-LIBCALL-NEXT: movq %rdi, %rax +; CHECK-LIBCALL-NEXT: shrq %rax +; CHECK-LIBCALL-NEXT: andl $1, %edi +; CHECK-LIBCALL-NEXT: orq %rax, %rdi +; CHECK-LIBCALL-NEXT: cvtsi2ssq %rdi, %xmm0 +; CHECK-LIBCALL-NEXT: addss %xmm0, %xmm0 +; CHECK-LIBCALL-NEXT: .LBB10_3: +; CHECK-LIBCALL-NEXT: callq __gnu_f2h_ieee +; CHECK-LIBCALL-NEXT: movw %ax, (%rbx) +; CHECK-LIBCALL-NEXT: popq %rbx +; CHECK-LIBCALL-NEXT: retq +; +; CHECK-F16C-LABEL: test_uitofp_i64: +; CHECK-F16C: # BB#0: +; CHECK-F16C-NEXT: testq %rdi, %rdi +; CHECK-F16C-NEXT: js .LBB10_1 +; CHECK-F16C-NEXT: # BB#2: +; CHECK-F16C-NEXT: vcvtsi2ssq %rdi, %xmm0, %xmm0 +; CHECK-F16C-NEXT: jmp .LBB10_3 +; CHECK-F16C-NEXT: .LBB10_1: +; CHECK-F16C-NEXT: movq %rdi, %rax +; CHECK-F16C-NEXT: shrq %rax +; CHECK-F16C-NEXT: andl $1, %edi +; CHECK-F16C-NEXT: orq %rax, %rdi +; CHECK-F16C-NEXT: vcvtsi2ssq %rdi, %xmm0, %xmm0 +; CHECK-F16C-NEXT: vaddss %xmm0, %xmm0, %xmm0 +; CHECK-F16C-NEXT: .LBB10_3: +; CHECK-F16C-NEXT: vcvtps2ph $4, %xmm0, %xmm0 +; CHECK-F16C-NEXT: vmovd %xmm0, %eax +; CHECK-F16C-NEXT: movw %ax, (%rsi) +; CHECK-F16C-NEXT: retq +; +; CHECK-I686-LABEL: test_uitofp_i64: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: pushl %esi +; CHECK-I686-NEXT: subl $24, %esp +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %esi +; CHECK-I686-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero +; CHECK-I686-NEXT: movlps %xmm0, {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: xorl %eax, %eax +; CHECK-I686-NEXT: cmpl $0, {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: setns %al +; CHECK-I686-NEXT: fildll {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: fadds {{\.LCPI.*}}(,%eax,4) +; CHECK-I686-NEXT: fstps (%esp) +; CHECK-I686-NEXT: calll __gnu_f2h_ieee +; CHECK-I686-NEXT: movw %ax, (%esi) +; CHECK-I686-NEXT: addl $24, %esp +; CHECK-I686-NEXT: popl %esi +; CHECK-I686-NEXT: retl %r = uitofp i64 %a to half store half %r, half* %p ret void } define <4 x float> @test_extend32_vec4(<4 x half>* %p) #0 { -; CHECK-LABEL: test_extend32_vec4: - -; CHECK-LIBCALL: callq __gnu_h2f_ieee -; CHECK-LIBCALL: callq __gnu_h2f_ieee -; CHECK-LIBCALL: callq __gnu_h2f_ieee -; CHECK-LIBCALL: callq __gnu_h2f_ieee -; CHECK-F16C: vcvtph2ps -; CHECK-F16C: vcvtph2ps -; CHECK-F16C: vcvtph2ps -; CHECK-F16C: vcvtph2ps +; CHECK-LIBCALL-LABEL: test_extend32_vec4: +; CHECK-LIBCALL: # BB#0: +; CHECK-LIBCALL-NEXT: pushq %rbx +; CHECK-LIBCALL-NEXT: subq $48, %rsp +; CHECK-LIBCALL-NEXT: movq %rdi, %rbx +; CHECK-LIBCALL-NEXT: movzwl 6(%rbx), %edi +; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee +; CHECK-LIBCALL-NEXT: movaps %xmm0, {{[0-9]+}}(%rsp) # 16-byte Spill +; CHECK-LIBCALL-NEXT: movzwl 4(%rbx), %edi +; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee +; CHECK-LIBCALL-NEXT: movaps %xmm0, {{[0-9]+}}(%rsp) # 16-byte Spill +; CHECK-LIBCALL-NEXT: movzwl (%rbx), %edi +; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee +; CHECK-LIBCALL-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-LIBCALL-NEXT: movzwl 2(%rbx), %edi +; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee +; CHECK-LIBCALL-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload +; CHECK-LIBCALL-NEXT: insertps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[2,3] +; CHECK-LIBCALL-NEXT: insertps $32, {{[0-9]+}}(%rsp), %xmm1 # 16-byte Folded Reload +; CHECK-LIBCALL-NEXT: # xmm1 = xmm1[0,1],mem[0],xmm1[3] +; CHECK-LIBCALL-NEXT: insertps $48, {{[0-9]+}}(%rsp), %xmm1 # 16-byte Folded Reload +; CHECK-LIBCALL-NEXT: # xmm1 = xmm1[0,1,2],mem[0] +; CHECK-LIBCALL-NEXT: movaps %xmm1, %xmm0 +; CHECK-LIBCALL-NEXT: addq $48, %rsp +; CHECK-LIBCALL-NEXT: popq %rbx +; CHECK-LIBCALL-NEXT: retq +; +; CHECK-F16C-LABEL: test_extend32_vec4: +; CHECK-F16C: # BB#0: +; CHECK-F16C-NEXT: movswl 6(%rdi), %eax +; CHECK-F16C-NEXT: vmovd %eax, %xmm0 +; CHECK-F16C-NEXT: vcvtph2ps %xmm0, %xmm0 +; CHECK-F16C-NEXT: movswl 4(%rdi), %eax +; CHECK-F16C-NEXT: vmovd %eax, %xmm1 +; CHECK-F16C-NEXT: vcvtph2ps %xmm1, %xmm1 +; CHECK-F16C-NEXT: movswl (%rdi), %eax +; CHECK-F16C-NEXT: vmovd %eax, %xmm2 +; CHECK-F16C-NEXT: vcvtph2ps %xmm2, %xmm2 +; CHECK-F16C-NEXT: movswl 2(%rdi), %eax +; CHECK-F16C-NEXT: vmovd %eax, %xmm3 +; CHECK-F16C-NEXT: vcvtph2ps %xmm3, %xmm3 +; CHECK-F16C-NEXT: vinsertps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[2,3] +; CHECK-F16C-NEXT: vinsertps {{.*#+}} xmm1 = xmm2[0,1],xmm1[0],xmm2[3] +; CHECK-F16C-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0] +; CHECK-F16C-NEXT: retq +; +; CHECK-I686-LABEL: test_extend32_vec4: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: pushl %esi +; CHECK-I686-NEXT: subl $56, %esp +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %esi +; CHECK-I686-NEXT: movzwl 2(%esi), %eax +; CHECK-I686-NEXT: movl %eax, (%esp) +; CHECK-I686-NEXT: calll __gnu_h2f_ieee +; CHECK-I686-NEXT: fstpt {{[0-9]+}}(%esp) # 10-byte Folded Spill +; CHECK-I686-NEXT: movzwl 4(%esi), %eax +; CHECK-I686-NEXT: movl %eax, (%esp) +; CHECK-I686-NEXT: calll __gnu_h2f_ieee +; CHECK-I686-NEXT: fstpt {{[0-9]+}}(%esp) # 10-byte Folded Spill +; CHECK-I686-NEXT: movzwl 6(%esi), %eax +; CHECK-I686-NEXT: movl %eax, (%esp) +; CHECK-I686-NEXT: calll __gnu_h2f_ieee +; CHECK-I686-NEXT: movzwl (%esi), %eax +; CHECK-I686-NEXT: movl %eax, (%esp) +; CHECK-I686-NEXT: fstps {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: fldt {{[0-9]+}}(%esp) # 10-byte Folded Reload +; CHECK-I686-NEXT: fstps {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: fldt {{[0-9]+}}(%esp) # 10-byte Folded Reload +; CHECK-I686-NEXT: fstps {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: calll __gnu_h2f_ieee +; CHECK-I686-NEXT: fstps {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero +; CHECK-I686-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero +; CHECK-I686-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1] +; CHECK-I686-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero +; CHECK-I686-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero +; CHECK-I686-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] +; CHECK-I686-NEXT: unpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0] +; CHECK-I686-NEXT: addl $56, %esp +; CHECK-I686-NEXT: popl %esi +; CHECK-I686-NEXT: retl %a = load <4 x half>, <4 x half>* %p, align 8 %b = fpext <4 x half> %a to <4 x float> ret <4 x float> %b } define <4 x double> @test_extend64_vec4(<4 x half>* %p) #0 { -; CHECK-LABEL: test_extend64_vec4 - -; CHECK-LIBCALL: callq __gnu_h2f_ieee -; CHECK-LIBCALL-DAG: callq __gnu_h2f_ieee -; CHECK-LIBCALL-DAG: callq __gnu_h2f_ieee -; CHECK-LIBCALL-DAG: callq __gnu_h2f_ieee -; CHECK-LIBCALL-DAG: cvtss2sd -; CHECK-LIBCALL-DAG: cvtss2sd -; CHECK-LIBCALL-DAG: cvtss2sd -; CHECK-LIBCALL: cvtss2sd -; CHECK-F16C: vcvtph2ps -; CHECK-F16C-DAG: vcvtph2ps -; CHECK-F16C-DAG: vcvtph2ps -; CHECK-F16C-DAG: vcvtph2ps -; CHECK-F16C-DAG: vcvtss2sd -; CHECK-F16C-DAG: vcvtss2sd -; CHECK-F16C-DAG: vcvtss2sd -; CHECK-F16C: vcvtss2sd +; CHECK-LIBCALL-LABEL: test_extend64_vec4: +; CHECK-LIBCALL: # BB#0: +; CHECK-LIBCALL-NEXT: pushq %rbx +; CHECK-LIBCALL-NEXT: subq $16, %rsp +; CHECK-LIBCALL-NEXT: movq %rdi, %rbx +; CHECK-LIBCALL-NEXT: movzwl 4(%rbx), %edi +; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee +; CHECK-LIBCALL-NEXT: movss %xmm0, {{[0-9]+}}(%rsp) # 4-byte Spill +; CHECK-LIBCALL-NEXT: movzwl 6(%rbx), %edi +; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee +; CHECK-LIBCALL-NEXT: movss %xmm0, {{[0-9]+}}(%rsp) # 4-byte Spill +; CHECK-LIBCALL-NEXT: movzwl (%rbx), %edi +; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee +; CHECK-LIBCALL-NEXT: movss %xmm0, {{[0-9]+}}(%rsp) # 4-byte Spill +; CHECK-LIBCALL-NEXT: movzwl 2(%rbx), %edi +; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee +; CHECK-LIBCALL-NEXT: cvtss2sd %xmm0, %xmm1 +; CHECK-LIBCALL-NEXT: movss {{[0-9]+}}(%rsp), %xmm0 # 4-byte Reload +; CHECK-LIBCALL-NEXT: # xmm0 = mem[0],zero,zero,zero +; CHECK-LIBCALL-NEXT: cvtss2sd %xmm0, %xmm0 +; CHECK-LIBCALL-NEXT: unpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0] +; CHECK-LIBCALL-NEXT: movss {{[0-9]+}}(%rsp), %xmm1 # 4-byte Reload +; CHECK-LIBCALL-NEXT: # xmm1 = mem[0],zero,zero,zero +; CHECK-LIBCALL-NEXT: cvtss2sd %xmm1, %xmm2 +; CHECK-LIBCALL-NEXT: movss {{[0-9]+}}(%rsp), %xmm1 # 4-byte Reload +; CHECK-LIBCALL-NEXT: # xmm1 = mem[0],zero,zero,zero +; CHECK-LIBCALL-NEXT: cvtss2sd %xmm1, %xmm1 +; CHECK-LIBCALL-NEXT: unpcklpd {{.*#+}} xmm1 = xmm1[0],xmm2[0] +; CHECK-LIBCALL-NEXT: addq $16, %rsp +; CHECK-LIBCALL-NEXT: popq %rbx +; CHECK-LIBCALL-NEXT: retq +; +; CHECK-F16C-LABEL: test_extend64_vec4: +; CHECK-F16C: # BB#0: +; CHECK-F16C-NEXT: movswl (%rdi), %eax +; CHECK-F16C-NEXT: vmovd %eax, %xmm0 +; CHECK-F16C-NEXT: vcvtph2ps %xmm0, %xmm0 +; CHECK-F16C-NEXT: movswl 2(%rdi), %eax +; CHECK-F16C-NEXT: vmovd %eax, %xmm1 +; CHECK-F16C-NEXT: vcvtph2ps %xmm1, %xmm1 +; CHECK-F16C-NEXT: movswl 4(%rdi), %eax +; CHECK-F16C-NEXT: vmovd %eax, %xmm2 +; CHECK-F16C-NEXT: vcvtph2ps %xmm2, %xmm2 +; CHECK-F16C-NEXT: movswl 6(%rdi), %eax +; CHECK-F16C-NEXT: vmovd %eax, %xmm3 +; CHECK-F16C-NEXT: vcvtph2ps %xmm3, %xmm3 +; CHECK-F16C-NEXT: vcvtss2sd %xmm3, %xmm3, %xmm3 +; CHECK-F16C-NEXT: vcvtss2sd %xmm2, %xmm2, %xmm2 +; CHECK-F16C-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm2[0],xmm3[0] +; CHECK-F16C-NEXT: vcvtss2sd %xmm1, %xmm1, %xmm1 +; CHECK-F16C-NEXT: vcvtss2sd %xmm0, %xmm0, %xmm0 +; CHECK-F16C-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0] +; CHECK-F16C-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 +; CHECK-F16C-NEXT: retq +; +; CHECK-I686-LABEL: test_extend64_vec4: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: pushl %esi +; CHECK-I686-NEXT: subl $88, %esp +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %esi +; CHECK-I686-NEXT: movzwl 6(%esi), %eax +; CHECK-I686-NEXT: movl %eax, (%esp) +; CHECK-I686-NEXT: calll __gnu_h2f_ieee +; CHECK-I686-NEXT: fstpt {{[0-9]+}}(%esp) # 10-byte Folded Spill +; CHECK-I686-NEXT: movzwl 4(%esi), %eax +; CHECK-I686-NEXT: movl %eax, (%esp) +; CHECK-I686-NEXT: calll __gnu_h2f_ieee +; CHECK-I686-NEXT: fstpt {{[0-9]+}}(%esp) # 10-byte Folded Spill +; CHECK-I686-NEXT: movzwl 2(%esi), %eax +; CHECK-I686-NEXT: movl %eax, (%esp) +; CHECK-I686-NEXT: calll __gnu_h2f_ieee +; CHECK-I686-NEXT: fstpt {{[0-9]+}}(%esp) # 10-byte Folded Spill +; CHECK-I686-NEXT: movzwl (%esi), %eax +; CHECK-I686-NEXT: movl %eax, (%esp) +; CHECK-I686-NEXT: calll __gnu_h2f_ieee +; CHECK-I686-NEXT: fstpl {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: fldt {{[0-9]+}}(%esp) # 10-byte Folded Reload +; CHECK-I686-NEXT: fstpl {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: fldt {{[0-9]+}}(%esp) # 10-byte Folded Reload +; CHECK-I686-NEXT: fstpl {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: fldt {{[0-9]+}}(%esp) # 10-byte Folded Reload +; CHECK-I686-NEXT: fstpl {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero +; CHECK-I686-NEXT: movhpd {{.*#+}} xmm0 = xmm0[0],mem[0] +; CHECK-I686-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero +; CHECK-I686-NEXT: movhpd {{.*#+}} xmm1 = xmm1[0],mem[0] +; CHECK-I686-NEXT: addl $88, %esp +; CHECK-I686-NEXT: popl %esi +; CHECK-I686-NEXT: retl %a = load <4 x half>, <4 x half>* %p, align 8 %b = fpext <4 x half> %a to <4 x double> ret <4 x double> %b } -define void @test_trunc32_vec4(<4 x float> %a, <4 x half>* %p) { -; CHECK-LABEL: test_trunc32_vec4: - -; CHECK-LIBCALL: callq __gnu_f2h_ieee -; CHECK-LIBCALL: callq __gnu_f2h_ieee -; CHECK-LIBCALL: callq __gnu_f2h_ieee -; CHECK-LIBCALL: callq __gnu_f2h_ieee -; CHECK-F16C: vcvtps2ph -; CHECK-F16C: vcvtps2ph -; CHECK-F16C: vcvtps2ph -; CHECK-F16C: vcvtps2ph -; CHECK: movw -; CHECK: movw -; CHECK: movw -; CHECK: movw +define void @test_trunc32_vec4(<4 x float> %a, <4 x half>* %p) #0 { +; NOF16-BWINSTS-LABEL: test_trunc32_vec4: +; NOF16-BWINSTS: # BB#0: +; NOF16-BWINSTS-NEXT: pushq %rbp +; NOF16-BWINSTS-NEXT: pushq %r15 +; NOF16-BWINSTS-NEXT: pushq %r14 +; NOF16-BWINSTS-NEXT: pushq %rbx +; NOF16-BWINSTS-NEXT: subq $24, %rsp +; NOF16-BWINSTS-NEXT: movq %rdi, %rbx +; NOF16-BWINSTS-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill +; NOF16-BWINSTS-NEXT: movshdup {{.*#+}} xmm0 = xmm0[1,1,3,3] +; NOF16-BWINSTS-NEXT: callq __gnu_f2h_ieee +; NOF16-BWINSTS-NEXT: movl %eax, %r14d +; NOF16-BWINSTS-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload +; NOF16-BWINSTS-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1] +; NOF16-BWINSTS-NEXT: callq __gnu_f2h_ieee +; NOF16-BWINSTS-NEXT: movl %eax, %r15d +; NOF16-BWINSTS-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload +; NOF16-BWINSTS-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1,2,3] +; NOF16-BWINSTS-NEXT: callq __gnu_f2h_ieee +; NOF16-BWINSTS-NEXT: movl %eax, %ebp +; NOF16-BWINSTS-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload +; NOF16-BWINSTS-NEXT: callq __gnu_f2h_ieee +; NOF16-BWINSTS-NEXT: movw %ax, (%rbx) +; NOF16-BWINSTS-NEXT: movw %bp, 6(%rbx) +; NOF16-BWINSTS-NEXT: movw %r15w, 4(%rbx) +; NOF16-BWINSTS-NEXT: movw %r14w, 2(%rbx) +; NOF16-BWINSTS-NEXT: addq $24, %rsp +; NOF16-BWINSTS-NEXT: popq %rbx +; NOF16-BWINSTS-NEXT: popq %r14 +; NOF16-BWINSTS-NEXT: popq %r15 +; NOF16-BWINSTS-NEXT: popq %rbp +; NOF16-BWINSTS-NEXT: retq +; +; BWOFF-LABEL: test_trunc32_vec4: +; BWOFF: # BB#0: +; BWOFF-NEXT: pushq %rbp +; BWOFF-NEXT: pushq %r15 +; BWOFF-NEXT: pushq %r14 +; BWOFF-NEXT: pushq %rbx +; BWOFF-NEXT: subq $24, %rsp +; BWOFF-NEXT: movq %rdi, %rbx +; BWOFF-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill +; BWOFF-NEXT: movshdup {{.*#+}} xmm0 = xmm0[1,1,3,3] +; BWOFF-NEXT: callq __gnu_f2h_ieee +; BWOFF-NEXT: movw %ax, %r14w +; BWOFF-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload +; BWOFF-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1] +; BWOFF-NEXT: callq __gnu_f2h_ieee +; BWOFF-NEXT: movw %ax, %r15w +; BWOFF-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload +; BWOFF-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1,2,3] +; BWOFF-NEXT: callq __gnu_f2h_ieee +; BWOFF-NEXT: movw %ax, %bp +; BWOFF-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload +; BWOFF-NEXT: callq __gnu_f2h_ieee +; BWOFF-NEXT: movw %ax, (%rbx) +; BWOFF-NEXT: movw %bp, 6(%rbx) +; BWOFF-NEXT: movw %r15w, 4(%rbx) +; BWOFF-NEXT: movw %r14w, 2(%rbx) +; BWOFF-NEXT: addq $24, %rsp +; BWOFF-NEXT: popq %rbx +; BWOFF-NEXT: popq %r14 +; BWOFF-NEXT: popq %r15 +; BWOFF-NEXT: popq %rbp +; BWOFF-NEXT: retq +; +; CHECK-F16C-LABEL: test_trunc32_vec4: +; CHECK-F16C: # BB#0: +; CHECK-F16C-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3] +; CHECK-F16C-NEXT: vcvtps2ph $4, %xmm1, %xmm1 +; CHECK-F16C-NEXT: vmovd %xmm1, %eax +; CHECK-F16C-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0] +; CHECK-F16C-NEXT: vcvtps2ph $4, %xmm1, %xmm1 +; CHECK-F16C-NEXT: vmovd %xmm1, %ecx +; CHECK-F16C-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[3,1,2,3] +; CHECK-F16C-NEXT: vcvtps2ph $4, %xmm1, %xmm1 +; CHECK-F16C-NEXT: vmovd %xmm1, %edx +; CHECK-F16C-NEXT: vcvtps2ph $4, %xmm0, %xmm0 +; CHECK-F16C-NEXT: vmovd %xmm0, %esi +; CHECK-F16C-NEXT: movw %si, (%rdi) +; CHECK-F16C-NEXT: movw %dx, 6(%rdi) +; CHECK-F16C-NEXT: movw %cx, 4(%rdi) +; CHECK-F16C-NEXT: movw %ax, 2(%rdi) +; CHECK-F16C-NEXT: retq +; +; CHECK-I686-LABEL: test_trunc32_vec4: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: pushl %ebp +; CHECK-I686-NEXT: pushl %ebx +; CHECK-I686-NEXT: pushl %edi +; CHECK-I686-NEXT: pushl %esi +; CHECK-I686-NEXT: subl $44, %esp +; CHECK-I686-NEXT: movaps %xmm0, {{[0-9]+}}(%esp) # 16-byte Spill +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %ebp +; CHECK-I686-NEXT: movaps %xmm0, %xmm1 +; CHECK-I686-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,1,2,3] +; CHECK-I686-NEXT: movss %xmm1, (%esp) +; CHECK-I686-NEXT: calll __gnu_f2h_ieee +; CHECK-I686-NEXT: movw %ax, %si +; CHECK-I686-NEXT: movaps {{[0-9]+}}(%esp), %xmm0 # 16-byte Reload +; CHECK-I686-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1] +; CHECK-I686-NEXT: movss %xmm0, (%esp) +; CHECK-I686-NEXT: calll __gnu_f2h_ieee +; CHECK-I686-NEXT: movw %ax, %di +; CHECK-I686-NEXT: movaps {{[0-9]+}}(%esp), %xmm0 # 16-byte Reload +; CHECK-I686-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1,2,3] +; CHECK-I686-NEXT: movss %xmm0, (%esp) +; CHECK-I686-NEXT: calll __gnu_f2h_ieee +; CHECK-I686-NEXT: movw %ax, %bx +; CHECK-I686-NEXT: movaps {{[0-9]+}}(%esp), %xmm0 # 16-byte Reload +; CHECK-I686-NEXT: movss %xmm0, (%esp) +; CHECK-I686-NEXT: calll __gnu_f2h_ieee +; CHECK-I686-NEXT: movw %ax, (%ebp) +; CHECK-I686-NEXT: movw %bx, 6(%ebp) +; CHECK-I686-NEXT: movw %di, 4(%ebp) +; CHECK-I686-NEXT: movw %si, 2(%ebp) +; CHECK-I686-NEXT: addl $44, %esp +; CHECK-I686-NEXT: popl %esi +; CHECK-I686-NEXT: popl %edi +; CHECK-I686-NEXT: popl %ebx +; CHECK-I686-NEXT: popl %ebp +; CHECK-I686-NEXT: retl %v = fptrunc <4 x float> %a to <4 x half> store <4 x half> %v, <4 x half>* %p ret void } -define void @test_trunc64_vec4(<4 x double> %a, <4 x half>* %p) { -; CHECK-LABEL: test_trunc64_vec4: -; CHECK: callq __truncdfhf2 -; CHECK: callq __truncdfhf2 -; CHECK: callq __truncdfhf2 -; CHECK: callq __truncdfhf2 -; CHECK: movw -; CHECK: movw -; CHECK: movw -; CHECK: movw +define void @test_trunc64_vec4(<4 x double> %a, <4 x half>* %p) #0 { +; NOF16-BWINSTS-LABEL: test_trunc64_vec4: +; NOF16-BWINSTS: # BB#0: +; NOF16-BWINSTS-NEXT: pushq %rbp +; NOF16-BWINSTS-NEXT: pushq %r15 +; NOF16-BWINSTS-NEXT: pushq %r14 +; NOF16-BWINSTS-NEXT: pushq %rbx +; NOF16-BWINSTS-NEXT: subq $40, %rsp +; NOF16-BWINSTS-NEXT: movq %rdi, %rbx +; NOF16-BWINSTS-NEXT: movaps %xmm1, (%rsp) # 16-byte Spill +; NOF16-BWINSTS-NEXT: movaps %xmm0, {{[0-9]+}}(%rsp) # 16-byte Spill +; NOF16-BWINSTS-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1] +; NOF16-BWINSTS-NEXT: callq __truncdfhf2 +; NOF16-BWINSTS-NEXT: movl %eax, %r14d +; NOF16-BWINSTS-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload +; NOF16-BWINSTS-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1] +; NOF16-BWINSTS-NEXT: callq __truncdfhf2 +; NOF16-BWINSTS-NEXT: movl %eax, %r15d +; NOF16-BWINSTS-NEXT: movaps {{[0-9]+}}(%rsp), %xmm0 # 16-byte Reload +; NOF16-BWINSTS-NEXT: callq __truncdfhf2 +; NOF16-BWINSTS-NEXT: movl %eax, %ebp +; NOF16-BWINSTS-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload +; NOF16-BWINSTS-NEXT: callq __truncdfhf2 +; NOF16-BWINSTS-NEXT: movw %ax, 4(%rbx) +; NOF16-BWINSTS-NEXT: movw %bp, (%rbx) +; NOF16-BWINSTS-NEXT: movw %r15w, 6(%rbx) +; NOF16-BWINSTS-NEXT: movw %r14w, 2(%rbx) +; NOF16-BWINSTS-NEXT: addq $40, %rsp +; NOF16-BWINSTS-NEXT: popq %rbx +; NOF16-BWINSTS-NEXT: popq %r14 +; NOF16-BWINSTS-NEXT: popq %r15 +; NOF16-BWINSTS-NEXT: popq %rbp +; NOF16-BWINSTS-NEXT: retq +; +; BWOFF-LABEL: test_trunc64_vec4: +; BWOFF: # BB#0: +; BWOFF-NEXT: pushq %rbp +; BWOFF-NEXT: pushq %r15 +; BWOFF-NEXT: pushq %r14 +; BWOFF-NEXT: pushq %rbx +; BWOFF-NEXT: subq $40, %rsp +; BWOFF-NEXT: movq %rdi, %rbx +; BWOFF-NEXT: movaps %xmm1, (%rsp) # 16-byte Spill +; BWOFF-NEXT: movaps %xmm0, {{[0-9]+}}(%rsp) # 16-byte Spill +; BWOFF-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1] +; BWOFF-NEXT: callq __truncdfhf2 +; BWOFF-NEXT: movw %ax, %r14w +; BWOFF-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload +; BWOFF-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1] +; BWOFF-NEXT: callq __truncdfhf2 +; BWOFF-NEXT: movw %ax, %r15w +; BWOFF-NEXT: movaps {{[0-9]+}}(%rsp), %xmm0 # 16-byte Reload +; BWOFF-NEXT: callq __truncdfhf2 +; BWOFF-NEXT: movw %ax, %bp +; BWOFF-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload +; BWOFF-NEXT: callq __truncdfhf2 +; BWOFF-NEXT: movw %ax, 4(%rbx) +; BWOFF-NEXT: movw %bp, (%rbx) +; BWOFF-NEXT: movw %r15w, 6(%rbx) +; BWOFF-NEXT: movw %r14w, 2(%rbx) +; BWOFF-NEXT: addq $40, %rsp +; BWOFF-NEXT: popq %rbx +; BWOFF-NEXT: popq %r14 +; BWOFF-NEXT: popq %r15 +; BWOFF-NEXT: popq %rbp +; BWOFF-NEXT: retq +; +; CHECK-F16C-LABEL: test_trunc64_vec4: +; CHECK-F16C: # BB#0: +; CHECK-F16C-NEXT: pushq %rbp +; CHECK-F16C-NEXT: pushq %r15 +; CHECK-F16C-NEXT: pushq %r14 +; CHECK-F16C-NEXT: pushq %rbx +; CHECK-F16C-NEXT: subq $88, %rsp +; CHECK-F16C-NEXT: movq %rdi, %rbx +; CHECK-F16C-NEXT: vmovupd %ymm0, {{[0-9]+}}(%rsp) # 32-byte Spill +; CHECK-F16C-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] +; CHECK-F16C-NEXT: vzeroupper +; CHECK-F16C-NEXT: callq __truncdfhf2 +; CHECK-F16C-NEXT: movl %eax, %r14d +; CHECK-F16C-NEXT: vmovups {{[0-9]+}}(%rsp), %ymm0 # 32-byte Reload +; CHECK-F16C-NEXT: vextractf128 $1, %ymm0, %xmm0 +; CHECK-F16C-NEXT: vmovapd %xmm0, {{[0-9]+}}(%rsp) # 16-byte Spill +; CHECK-F16C-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] +; CHECK-F16C-NEXT: vzeroupper +; CHECK-F16C-NEXT: callq __truncdfhf2 +; CHECK-F16C-NEXT: movl %eax, %r15d +; CHECK-F16C-NEXT: vmovups {{[0-9]+}}(%rsp), %ymm0 # 32-byte Reload +; CHECK-F16C-NEXT: # kill: %XMM0<def> %XMM0<kill> %YMM0<kill> +; CHECK-F16C-NEXT: vzeroupper +; CHECK-F16C-NEXT: callq __truncdfhf2 +; CHECK-F16C-NEXT: movl %eax, %ebp +; CHECK-F16C-NEXT: vmovaps {{[0-9]+}}(%rsp), %xmm0 # 16-byte Reload +; CHECK-F16C-NEXT: callq __truncdfhf2 +; CHECK-F16C-NEXT: movw %ax, 4(%rbx) +; CHECK-F16C-NEXT: movw %bp, (%rbx) +; CHECK-F16C-NEXT: movw %r15w, 6(%rbx) +; CHECK-F16C-NEXT: movw %r14w, 2(%rbx) +; CHECK-F16C-NEXT: addq $88, %rsp +; CHECK-F16C-NEXT: popq %rbx +; CHECK-F16C-NEXT: popq %r14 +; CHECK-F16C-NEXT: popq %r15 +; CHECK-F16C-NEXT: popq %rbp +; CHECK-F16C-NEXT: retq +; +; CHECK-I686-LABEL: test_trunc64_vec4: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: pushl %ebp +; CHECK-I686-NEXT: pushl %ebx +; CHECK-I686-NEXT: pushl %edi +; CHECK-I686-NEXT: pushl %esi +; CHECK-I686-NEXT: subl $60, %esp +; CHECK-I686-NEXT: movaps %xmm1, {{[0-9]+}}(%esp) # 16-byte Spill +; CHECK-I686-NEXT: movaps %xmm0, {{[0-9]+}}(%esp) # 16-byte Spill +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %ebp +; CHECK-I686-NEXT: movlps %xmm0, (%esp) +; CHECK-I686-NEXT: calll __truncdfhf2 +; CHECK-I686-NEXT: movw %ax, %si +; CHECK-I686-NEXT: movapd {{[0-9]+}}(%esp), %xmm0 # 16-byte Reload +; CHECK-I686-NEXT: movhpd %xmm0, (%esp) +; CHECK-I686-NEXT: calll __truncdfhf2 +; CHECK-I686-NEXT: movw %ax, %di +; CHECK-I686-NEXT: movaps {{[0-9]+}}(%esp), %xmm0 # 16-byte Reload +; CHECK-I686-NEXT: movlps %xmm0, (%esp) +; CHECK-I686-NEXT: calll __truncdfhf2 +; CHECK-I686-NEXT: movw %ax, %bx +; CHECK-I686-NEXT: movapd {{[0-9]+}}(%esp), %xmm0 # 16-byte Reload +; CHECK-I686-NEXT: movhpd %xmm0, (%esp) +; CHECK-I686-NEXT: calll __truncdfhf2 +; CHECK-I686-NEXT: movw %ax, 6(%ebp) +; CHECK-I686-NEXT: movw %bx, 4(%ebp) +; CHECK-I686-NEXT: movw %di, 2(%ebp) +; CHECK-I686-NEXT: movw %si, (%ebp) +; CHECK-I686-NEXT: addl $60, %esp +; CHECK-I686-NEXT: popl %esi +; CHECK-I686-NEXT: popl %edi +; CHECK-I686-NEXT: popl %ebx +; CHECK-I686-NEXT: popl %ebp +; CHECK-I686-NEXT: retl %v = fptrunc <4 x double> %a to <4 x half> store <4 x half> %v, <4 x half>* %p ret void @@ -268,44 +836,99 @@ define void @test_trunc64_vec4(<4 x double> %a, <4 x half>* %p) { declare float @test_floatret(); -; On i686, if SSE2 is available, the return value from test_floatret is loaded -; to f80 and then rounded to f32. The DAG combiner should not combine this -; fp_round and the subsequent fptrunc from float to half. define half @test_f80trunc_nodagcombine() #0 { -; CHECK-LABEL: test_f80trunc_nodagcombine: -; CHECK-I686-NOT: calll __truncxfhf2 +; CHECK-LIBCALL-LABEL: test_f80trunc_nodagcombine: +; CHECK-LIBCALL: # BB#0: +; CHECK-LIBCALL-NEXT: pushq %rax +; CHECK-LIBCALL-NEXT: callq test_floatret +; CHECK-LIBCALL-NEXT: callq __gnu_f2h_ieee +; CHECK-LIBCALL-NEXT: movzwl %ax, %edi +; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee +; CHECK-LIBCALL-NEXT: popq %rax +; CHECK-LIBCALL-NEXT: retq +; +; CHECK-F16C-LABEL: test_f80trunc_nodagcombine: +; CHECK-F16C: # BB#0: +; CHECK-F16C-NEXT: pushq %rax +; CHECK-F16C-NEXT: callq test_floatret +; CHECK-F16C-NEXT: vcvtps2ph $4, %xmm0, %xmm0 +; CHECK-F16C-NEXT: vcvtph2ps %xmm0, %xmm0 +; CHECK-F16C-NEXT: popq %rax +; CHECK-F16C-NEXT: retq +; +; CHECK-I686-LABEL: test_f80trunc_nodagcombine: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: subl $12, %esp +; CHECK-I686-NEXT: calll test_floatret +; CHECK-I686-NEXT: fstps (%esp) +; CHECK-I686-NEXT: calll __gnu_f2h_ieee +; CHECK-I686-NEXT: movzwl %ax, %eax +; CHECK-I686-NEXT: movl %eax, (%esp) +; CHECK-I686-NEXT: calll __gnu_h2f_ieee +; CHECK-I686-NEXT: addl $12, %esp +; CHECK-I686-NEXT: retl %1 = call float @test_floatret() %2 = fptrunc float %1 to half ret half %2 } -; CHECK-LABEL: test_sitofp_fadd_i32: -; CHECK-LIBCALL-NEXT: pushq %rbx -; CHECK-LIBCALL-NEXT: subq $16, %rsp -; CHECK-LIBCALL-NEXT: movl %edi, %ebx -; CHECK-LIBCALL-NEXT: movzwl (%rsi), %edi -; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee -; CHECK-LIBCALL-NEXT: movss %xmm0, 12(%rsp) -; CHECK-LIBCALL-NEXT: cvtsi2ssl %ebx, %xmm0 -; CHECK-LIBCALL-NEXT: callq __gnu_f2h_ieee -; CHECK-LIBCALL-NEXT: movzwl %ax, %edi -; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee -; CHECK-LIBCALL-NEXT: addss 12(%rsp), %xmm0 -; CHECK-LIBCALL-NEXT: addq $16, %rsp -; CHECK-LIBCALL-NEXT: popq %rbx -; CHECK-LIBCALL-NEXT: retq -; CHECK-F16C-NEXT: movswl (%rsi), %eax -; CHECK-F16C-NEXT: vmovd %eax, %xmm0 -; CHECK-F16C-NEXT: vcvtph2ps %xmm0, %xmm0 -; CHECK-F16C-NEXT: vcvtsi2ssl %edi, %xmm1, %xmm1 -; CHECK-F16C-NEXT: vcvtps2ph $4, %xmm1, %xmm1 -; CHECK-F16C-NEXT: vcvtph2ps %xmm1, %xmm1 -; CHECK-F16C-NEXT: vaddss %xmm1, %xmm0, %xmm0 -; CHECK-F16C-NEXT: retq define float @test_sitofp_fadd_i32(i32 %a, half* %b) #0 { +; CHECK-LIBCALL-LABEL: test_sitofp_fadd_i32: +; CHECK-LIBCALL: # BB#0: +; CHECK-LIBCALL-NEXT: pushq %rbx +; CHECK-LIBCALL-NEXT: subq $16, %rsp +; CHECK-LIBCALL-NEXT: movl %edi, %ebx +; CHECK-LIBCALL-NEXT: movzwl (%rsi), %edi +; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee +; CHECK-LIBCALL-NEXT: movss %xmm0, {{[0-9]+}}(%rsp) # 4-byte Spill +; CHECK-LIBCALL-NEXT: cvtsi2ssl %ebx, %xmm0 +; CHECK-LIBCALL-NEXT: callq __gnu_f2h_ieee +; CHECK-LIBCALL-NEXT: movzwl %ax, %edi +; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee +; CHECK-LIBCALL-NEXT: addss {{[0-9]+}}(%rsp), %xmm0 # 4-byte Folded Reload +; CHECK-LIBCALL-NEXT: addq $16, %rsp +; CHECK-LIBCALL-NEXT: popq %rbx +; CHECK-LIBCALL-NEXT: retq +; +; CHECK-F16C-LABEL: test_sitofp_fadd_i32: +; CHECK-F16C: # BB#0: +; CHECK-F16C-NEXT: movswl (%rsi), %eax +; CHECK-F16C-NEXT: vmovd %eax, %xmm0 +; CHECK-F16C-NEXT: vcvtsi2ssl %edi, %xmm1, %xmm1 +; CHECK-F16C-NEXT: vcvtph2ps %xmm0, %xmm0 +; CHECK-F16C-NEXT: vcvtps2ph $4, %xmm1, %xmm1 +; CHECK-F16C-NEXT: vcvtph2ps %xmm1, %xmm1 +; CHECK-F16C-NEXT: vaddss %xmm1, %xmm0, %xmm0 +; CHECK-F16C-NEXT: retq +; +; CHECK-I686-LABEL: test_sitofp_fadd_i32: +; CHECK-I686: # BB#0: +; CHECK-I686-NEXT: subl $28, %esp +; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %eax +; CHECK-I686-NEXT: movzwl (%eax), %eax +; CHECK-I686-NEXT: movl %eax, (%esp) +; CHECK-I686-NEXT: calll __gnu_h2f_ieee +; CHECK-I686-NEXT: fstps {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero +; CHECK-I686-NEXT: movss %xmm0, {{[0-9]+}}(%esp) # 4-byte Spill +; CHECK-I686-NEXT: xorps %xmm0, %xmm0 +; CHECK-I686-NEXT: cvtsi2ssl {{[0-9]+}}(%esp), %xmm0 +; CHECK-I686-NEXT: movss %xmm0, (%esp) +; CHECK-I686-NEXT: calll __gnu_f2h_ieee +; CHECK-I686-NEXT: movzwl %ax, %eax +; CHECK-I686-NEXT: movl %eax, (%esp) +; CHECK-I686-NEXT: calll __gnu_h2f_ieee +; CHECK-I686-NEXT: fstps {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: movss {{[0-9]+}}(%esp), %xmm0 # 4-byte Reload +; CHECK-I686-NEXT: # xmm0 = mem[0],zero,zero,zero +; CHECK-I686-NEXT: addss {{[0-9]+}}(%esp), %xmm0 +; CHECK-I686-NEXT: movss %xmm0, {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: flds {{[0-9]+}}(%esp) +; CHECK-I686-NEXT: addl $28, %esp +; CHECK-I686-NEXT: retl %tmp0 = load half, half* %b %tmp1 = sitofp i32 %a to half %tmp2 = fadd half %tmp0, %tmp1 diff --git a/llvm/test/CodeGen/X86/illegal-bitfield-loadstore.ll b/llvm/test/CodeGen/X86/illegal-bitfield-loadstore.ll index ceb46571190..54214764c9a 100644 --- a/llvm/test/CodeGen/X86/illegal-bitfield-loadstore.ll +++ b/llvm/test/CodeGen/X86/illegal-bitfield-loadstore.ll @@ -112,23 +112,23 @@ define void @i56_and_or(i56* %a) { define void @i56_insert_bit(i56* %a, i1 zeroext %bit) { ; CHECK-LABEL: i56_insert_bit: ; CHECK: # BB#0: -; CHECK-NEXT: movzbl %sil, %eax -; CHECK-NEXT: movzwl 4(%rdi), %ecx -; CHECK-NEXT: movzbl 6(%rdi), %edx -; CHECK-NEXT: movl (%rdi), %esi -; CHECK-NEXT: movb %dl, 6(%rdi) -; CHECK-NEXT: # kill: %EDX<def> %EDX<kill> %RDX<kill> %RDX<def> -; CHECK-NEXT: shll $16, %edx -; CHECK-NEXT: orl %ecx, %edx -; CHECK-NEXT: shlq $32, %rdx -; CHECK-NEXT: orq %rdx, %rsi -; CHECK-NEXT: shlq $13, %rax -; CHECK-NEXT: movabsq $72057594037919743, %rcx # imm = 0xFFFFFFFFFFDFFF -; CHECK-NEXT: andq %rsi, %rcx -; CHECK-NEXT: orq %rax, %rcx -; CHECK-NEXT: movl %ecx, (%rdi) -; CHECK-NEXT: shrq $32, %rcx -; CHECK-NEXT: movw %cx, 4(%rdi) +; CHECK-NEXT: movzwl 4(%rdi), %eax +; CHECK-NEXT: movzbl 6(%rdi), %ecx +; CHECK-NEXT: movl (%rdi), %edx +; CHECK-NEXT: movb %cl, 6(%rdi) +; CHECK-NEXT: movzbl %sil, %esi +; CHECK-NEXT: # kill: %ECX<def> %ECX<kill> %RCX<kill> %RCX<def> +; CHECK-NEXT: shll $16, %ecx +; CHECK-NEXT: orl %eax, %ecx +; CHECK-NEXT: shlq $32, %rcx +; CHECK-NEXT: orq %rcx, %rdx +; CHECK-NEXT: shlq $13, %rsi +; CHECK-NEXT: movabsq $72057594037919743, %rax # imm = 0xFFFFFFFFFFDFFF +; CHECK-NEXT: andq %rdx, %rax +; CHECK-NEXT: orq %rsi, %rax +; CHECK-NEXT: movl %eax, (%rdi) +; CHECK-NEXT: shrq $32, %rax +; CHECK-NEXT: movw %ax, 4(%rdi) ; CHECK-NEXT: retq %extbit = zext i1 %bit to i56 %b = load i56, i56* %a, align 1 diff --git a/llvm/test/CodeGen/X86/mul-constant-i32.ll b/llvm/test/CodeGen/X86/mul-constant-i32.ll index d545b477e10..a8b594904b6 100644 --- a/llvm/test/CodeGen/X86/mul-constant-i32.ll +++ b/llvm/test/CodeGen/X86/mul-constant-i32.ll @@ -17,7 +17,7 @@ define i32 @test_mul_by_1(i32 %x) { ; X64-HSW-LABEL: test_mul_by_1: ; X64-HSW: # BB#0: ; X64-HSW-NEXT: movl %edi, %eax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_1: ; X64-JAG: # BB#0: @@ -32,7 +32,7 @@ define i32 @test_mul_by_1(i32 %x) { ; HSW-NOOPT-LABEL: test_mul_by_1: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: movl %edi, %eax # sched: [1:0.25] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_1: ; JAG-NOOPT: # BB#0: @@ -63,7 +63,7 @@ define i32 @test_mul_by_2(i32 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: leal (%rdi,%rdi), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_2: ; X64-JAG: # BB#0: @@ -81,7 +81,7 @@ define i32 @test_mul_by_2(i32 %x) { ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; HSW-NOOPT-NEXT: leal (%rdi,%rdi), %eax # sched: [1:0.50] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_2: ; JAG-NOOPT: # BB#0: @@ -114,7 +114,7 @@ define i32 @test_mul_by_3(i32 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: leal (%rdi,%rdi,2), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_3: ; X64-JAG: # BB#0: @@ -131,7 +131,7 @@ define i32 @test_mul_by_3(i32 %x) { ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; HSW-NOOPT-NEXT: leal (%rdi,%rdi,2), %eax # sched: [1:0.50] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_3: ; JAG-NOOPT: # BB#0: @@ -165,7 +165,7 @@ define i32 @test_mul_by_4(i32 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: leal (,%rdi,4), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_4: ; X64-JAG: # BB#0: @@ -183,7 +183,7 @@ define i32 @test_mul_by_4(i32 %x) { ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; HSW-NOOPT-NEXT: leal (,%rdi,4), %eax # sched: [1:0.50] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_4: ; JAG-NOOPT: # BB#0: @@ -216,7 +216,7 @@ define i32 @test_mul_by_5(i32 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: leal (%rdi,%rdi,4), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_5: ; X64-JAG: # BB#0: @@ -233,7 +233,7 @@ define i32 @test_mul_by_5(i32 %x) { ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; HSW-NOOPT-NEXT: leal (%rdi,%rdi,4), %eax # sched: [1:0.50] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_5: ; JAG-NOOPT: # BB#0: @@ -269,7 +269,7 @@ define i32 @test_mul_by_6(i32 %x) { ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: addl %edi, %edi # sched: [1:0.25] ; X64-HSW-NEXT: leal (%rdi,%rdi,2), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_6: ; X64-JAG: # BB#0: @@ -285,8 +285,8 @@ define i32 @test_mul_by_6(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_6: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $6, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $6, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_6: ; JAG-NOOPT: # BB#0: @@ -321,7 +321,7 @@ define i32 @test_mul_by_7(i32 %x) { ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: leal (,%rdi,8), %eax # sched: [1:0.50] ; X64-HSW-NEXT: subl %edi, %eax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_7: ; X64-JAG: # BB#0: @@ -337,8 +337,8 @@ define i32 @test_mul_by_7(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_7: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $7, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $7, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_7: ; JAG-NOOPT: # BB#0: @@ -371,7 +371,7 @@ define i32 @test_mul_by_8(i32 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: leal (,%rdi,8), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_8: ; X64-JAG: # BB#0: @@ -389,7 +389,7 @@ define i32 @test_mul_by_8(i32 %x) { ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; HSW-NOOPT-NEXT: leal (,%rdi,8), %eax # sched: [1:0.50] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_8: ; JAG-NOOPT: # BB#0: @@ -422,7 +422,7 @@ define i32 @test_mul_by_9(i32 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: leal (%rdi,%rdi,8), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_9: ; X64-JAG: # BB#0: @@ -439,7 +439,7 @@ define i32 @test_mul_by_9(i32 %x) { ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; HSW-NOOPT-NEXT: leal (%rdi,%rdi,8), %eax # sched: [1:0.50] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_9: ; JAG-NOOPT: # BB#0: @@ -475,7 +475,7 @@ define i32 @test_mul_by_10(i32 %x) { ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: addl %edi, %edi # sched: [1:0.25] ; X64-HSW-NEXT: leal (%rdi,%rdi,4), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_10: ; X64-JAG: # BB#0: @@ -491,8 +491,8 @@ define i32 @test_mul_by_10(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_10: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $10, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $10, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_10: ; JAG-NOOPT: # BB#0: @@ -527,7 +527,7 @@ define i32 @test_mul_by_11(i32 %x) { ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: leal (%rdi,%rdi,4), %eax # sched: [1:0.50] ; X64-HSW-NEXT: leal (%rdi,%rax,2), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_11: ; X64-JAG: # BB#0: @@ -543,8 +543,8 @@ define i32 @test_mul_by_11(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_11: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $11, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $11, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_11: ; JAG-NOOPT: # BB#0: @@ -575,9 +575,9 @@ define i32 @test_mul_by_12(i32 %x) { ; X64-HSW-LABEL: test_mul_by_12: ; X64-HSW: # BB#0: ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> -; X64-HSW-NEXT: shll $2, %edi # sched: [1:0.50] +; X64-HSW-NEXT: shll $2, %edi # sched: [1:1.00] ; X64-HSW-NEXT: leal (%rdi,%rdi,2), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_12: ; X64-JAG: # BB#0: @@ -593,8 +593,8 @@ define i32 @test_mul_by_12(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_12: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $12, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $12, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_12: ; JAG-NOOPT: # BB#0: @@ -629,7 +629,7 @@ define i32 @test_mul_by_13(i32 %x) { ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: leal (%rdi,%rdi,2), %eax # sched: [1:0.50] ; X64-HSW-NEXT: leal (%rdi,%rax,4), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_13: ; X64-JAG: # BB#0: @@ -645,8 +645,8 @@ define i32 @test_mul_by_13(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_13: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $13, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $13, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_13: ; JAG-NOOPT: # BB#0: @@ -681,7 +681,7 @@ define i32 @test_mul_by_14(i32 %x) { ; X64-HSW-NEXT: leal (%rdi,%rdi,2), %eax # sched: [1:0.50] ; X64-HSW-NEXT: leal (%rdi,%rax,4), %eax # sched: [1:0.50] ; X64-HSW-NEXT: addl %edi, %eax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_14: ; X64-JAG: # BB#0: @@ -698,8 +698,8 @@ define i32 @test_mul_by_14(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_14: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $14, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $14, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_14: ; JAG-NOOPT: # BB#0: @@ -732,7 +732,7 @@ define i32 @test_mul_by_15(i32 %x) { ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: leal (%rdi,%rdi,4), %eax # sched: [1:0.50] ; X64-HSW-NEXT: leal (%rax,%rax,2), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_15: ; X64-JAG: # BB#0: @@ -748,8 +748,8 @@ define i32 @test_mul_by_15(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_15: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $15, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $15, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_15: ; JAG-NOOPT: # BB#0: @@ -780,9 +780,9 @@ define i32 @test_mul_by_16(i32 %x) { ; ; X64-HSW-LABEL: test_mul_by_16: ; X64-HSW: # BB#0: -; X64-HSW-NEXT: shll $4, %edi # sched: [1:0.50] +; X64-HSW-NEXT: shll $4, %edi # sched: [1:1.00] ; X64-HSW-NEXT: movl %edi, %eax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_16: ; X64-JAG: # BB#0: @@ -798,9 +798,9 @@ define i32 @test_mul_by_16(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_16: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: shll $4, %edi # sched: [1:0.50] +; HSW-NOOPT-NEXT: shll $4, %edi # sched: [1:1.00] ; HSW-NOOPT-NEXT: movl %edi, %eax # sched: [1:0.25] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_16: ; JAG-NOOPT: # BB#0: @@ -836,9 +836,9 @@ define i32 @test_mul_by_17(i32 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: movl %edi, %eax # sched: [1:0.25] -; X64-HSW-NEXT: shll $4, %eax # sched: [1:0.50] +; X64-HSW-NEXT: shll $4, %eax # sched: [1:1.00] ; X64-HSW-NEXT: leal (%rax,%rdi), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_17: ; X64-JAG: # BB#0: @@ -855,8 +855,8 @@ define i32 @test_mul_by_17(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_17: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $17, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $17, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_17: ; JAG-NOOPT: # BB#0: @@ -892,7 +892,7 @@ define i32 @test_mul_by_18(i32 %x) { ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: addl %edi, %edi # sched: [1:0.25] ; X64-HSW-NEXT: leal (%rdi,%rdi,8), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_18: ; X64-JAG: # BB#0: @@ -908,8 +908,8 @@ define i32 @test_mul_by_18(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_18: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $18, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $18, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_18: ; JAG-NOOPT: # BB#0: @@ -944,9 +944,9 @@ define i32 @test_mul_by_19(i32 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: leal (%rdi,%rdi,4), %eax # sched: [1:0.50] -; X64-HSW-NEXT: shll $2, %eax # sched: [1:0.50] +; X64-HSW-NEXT: shll $2, %eax # sched: [1:1.00] ; X64-HSW-NEXT: subl %edi, %eax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_19: ; X64-JAG: # BB#0: @@ -963,8 +963,8 @@ define i32 @test_mul_by_19(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_19: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $19, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $19, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_19: ; JAG-NOOPT: # BB#0: @@ -995,9 +995,9 @@ define i32 @test_mul_by_20(i32 %x) { ; X64-HSW-LABEL: test_mul_by_20: ; X64-HSW: # BB#0: ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> -; X64-HSW-NEXT: shll $2, %edi # sched: [1:0.50] +; X64-HSW-NEXT: shll $2, %edi # sched: [1:1.00] ; X64-HSW-NEXT: leal (%rdi,%rdi,4), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_20: ; X64-JAG: # BB#0: @@ -1013,8 +1013,8 @@ define i32 @test_mul_by_20(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_20: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $20, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $20, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_20: ; JAG-NOOPT: # BB#0: @@ -1049,7 +1049,7 @@ define i32 @test_mul_by_21(i32 %x) { ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: leal (%rdi,%rdi,4), %eax # sched: [1:0.50] ; X64-HSW-NEXT: leal (%rdi,%rax,4), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_21: ; X64-JAG: # BB#0: @@ -1065,8 +1065,8 @@ define i32 @test_mul_by_21(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_21: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $21, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $21, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_21: ; JAG-NOOPT: # BB#0: @@ -1101,7 +1101,7 @@ define i32 @test_mul_by_22(i32 %x) { ; X64-HSW-NEXT: leal (%rdi,%rdi,4), %eax # sched: [1:0.50] ; X64-HSW-NEXT: leal (%rdi,%rax,4), %eax # sched: [1:0.50] ; X64-HSW-NEXT: addl %edi, %eax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_22: ; X64-JAG: # BB#0: @@ -1118,8 +1118,8 @@ define i32 @test_mul_by_22(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_22: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $22, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $22, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_22: ; JAG-NOOPT: # BB#0: @@ -1152,9 +1152,9 @@ define i32 @test_mul_by_23(i32 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: leal (%rdi,%rdi,2), %eax # sched: [1:0.50] -; X64-HSW-NEXT: shll $3, %eax # sched: [1:0.50] +; X64-HSW-NEXT: shll $3, %eax # sched: [1:1.00] ; X64-HSW-NEXT: subl %edi, %eax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_23: ; X64-JAG: # BB#0: @@ -1171,8 +1171,8 @@ define i32 @test_mul_by_23(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_23: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $23, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $23, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_23: ; JAG-NOOPT: # BB#0: @@ -1203,9 +1203,9 @@ define i32 @test_mul_by_24(i32 %x) { ; X64-HSW-LABEL: test_mul_by_24: ; X64-HSW: # BB#0: ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> -; X64-HSW-NEXT: shll $3, %edi # sched: [1:0.50] +; X64-HSW-NEXT: shll $3, %edi # sched: [1:1.00] ; X64-HSW-NEXT: leal (%rdi,%rdi,2), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_24: ; X64-JAG: # BB#0: @@ -1221,8 +1221,8 @@ define i32 @test_mul_by_24(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_24: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $24, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $24, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_24: ; JAG-NOOPT: # BB#0: @@ -1257,7 +1257,7 @@ define i32 @test_mul_by_25(i32 %x) { ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: leal (%rdi,%rdi,4), %eax # sched: [1:0.50] ; X64-HSW-NEXT: leal (%rax,%rax,4), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_25: ; X64-JAG: # BB#0: @@ -1273,8 +1273,8 @@ define i32 @test_mul_by_25(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_25: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $25, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $25, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_25: ; JAG-NOOPT: # BB#0: @@ -1311,7 +1311,7 @@ define i32 @test_mul_by_26(i32 %x) { ; X64-HSW-NEXT: leal (%rdi,%rdi,8), %eax # sched: [1:0.50] ; X64-HSW-NEXT: leal (%rax,%rax,2), %eax # sched: [1:0.50] ; X64-HSW-NEXT: subl %edi, %eax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_26: ; X64-JAG: # BB#0: @@ -1328,8 +1328,8 @@ define i32 @test_mul_by_26(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_26: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $26, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $26, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_26: ; JAG-NOOPT: # BB#0: @@ -1362,7 +1362,7 @@ define i32 @test_mul_by_27(i32 %x) { ; X64-HSW-NEXT: # kill: %EDI<def> %EDI<kill> %RDI<def> ; X64-HSW-NEXT: leal (%rdi,%rdi,8), %eax # sched: [1:0.50] ; X64-HSW-NEXT: leal (%rax,%rax,2), %eax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_27: ; X64-JAG: # BB#0: @@ -1378,8 +1378,8 @@ define i32 @test_mul_by_27(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_27: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $27, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $27, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_27: ; JAG-NOOPT: # BB#0: @@ -1416,7 +1416,7 @@ define i32 @test_mul_by_28(i32 %x) { ; X64-HSW-NEXT: leal (%rdi,%rdi,8), %eax # sched: [1:0.50] ; X64-HSW-NEXT: leal (%rax,%rax,2), %eax # sched: [1:0.50] ; X64-HSW-NEXT: addl %edi, %eax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_28: ; X64-JAG: # BB#0: @@ -1433,8 +1433,8 @@ define i32 @test_mul_by_28(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_28: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $28, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $28, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_28: ; JAG-NOOPT: # BB#0: @@ -1471,7 +1471,7 @@ define i32 @test_mul_by_29(i32 %x) { ; X64-HSW-NEXT: leal (%rax,%rax,2), %eax # sched: [1:0.50] ; X64-HSW-NEXT: addl %edi, %eax # sched: [1:0.25] ; X64-HSW-NEXT: addl %edi, %eax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_29: ; X64-JAG: # BB#0: @@ -1489,8 +1489,8 @@ define i32 @test_mul_by_29(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_29: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $29, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $29, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_29: ; JAG-NOOPT: # BB#0: @@ -1523,10 +1523,10 @@ define i32 @test_mul_by_30(i32 %x) { ; X64-HSW-LABEL: test_mul_by_30: ; X64-HSW: # BB#0: ; X64-HSW-NEXT: movl %edi, %eax # sched: [1:0.25] -; X64-HSW-NEXT: shll $5, %eax # sched: [1:0.50] +; X64-HSW-NEXT: shll $5, %eax # sched: [1:1.00] ; X64-HSW-NEXT: subl %edi, %eax # sched: [1:0.25] ; X64-HSW-NEXT: subl %edi, %eax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_30: ; X64-JAG: # BB#0: @@ -1543,8 +1543,8 @@ define i32 @test_mul_by_30(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_30: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $30, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $30, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_30: ; JAG-NOOPT: # BB#0: @@ -1576,9 +1576,9 @@ define i32 @test_mul_by_31(i32 %x) { ; X64-HSW-LABEL: test_mul_by_31: ; X64-HSW: # BB#0: ; X64-HSW-NEXT: movl %edi, %eax # sched: [1:0.25] -; X64-HSW-NEXT: shll $5, %eax # sched: [1:0.50] +; X64-HSW-NEXT: shll $5, %eax # sched: [1:1.00] ; X64-HSW-NEXT: subl %edi, %eax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_31: ; X64-JAG: # BB#0: @@ -1594,8 +1594,8 @@ define i32 @test_mul_by_31(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_31: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: imull $31, %edi, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull $31, %edi, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_31: ; JAG-NOOPT: # BB#0: @@ -1626,9 +1626,9 @@ define i32 @test_mul_by_32(i32 %x) { ; ; X64-HSW-LABEL: test_mul_by_32: ; X64-HSW: # BB#0: -; X64-HSW-NEXT: shll $5, %edi # sched: [1:0.50] +; X64-HSW-NEXT: shll $5, %edi # sched: [1:1.00] ; X64-HSW-NEXT: movl %edi, %eax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_32: ; X64-JAG: # BB#0: @@ -1644,9 +1644,9 @@ define i32 @test_mul_by_32(i32 %x) { ; ; HSW-NOOPT-LABEL: test_mul_by_32: ; HSW-NOOPT: # BB#0: -; HSW-NOOPT-NEXT: shll $5, %edi # sched: [1:0.50] +; HSW-NOOPT-NEXT: shll $5, %edi # sched: [1:1.00] ; HSW-NOOPT-NEXT: movl %edi, %eax # sched: [1:0.25] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_32: ; JAG-NOOPT: # BB#0: @@ -1686,8 +1686,8 @@ define i32 @test_mul_spec(i32 %x) nounwind { ; X64-HSW-NEXT: addl $42, %ecx # sched: [1:0.25] ; X64-HSW-NEXT: leal (%rdi,%rdi,4), %eax # sched: [1:0.50] ; X64-HSW-NEXT: addl $2, %eax # sched: [1:0.25] -; X64-HSW-NEXT: imull %ecx, %eax # sched: [4:1.00] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: imull %ecx, %eax # sched: [3:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_spec: ; X64-JAG: # BB#0: @@ -1712,8 +1712,8 @@ define i32 @test_mul_spec(i32 %x) nounwind { ; HSW-NOOPT-NEXT: addl $42, %ecx # sched: [1:0.25] ; HSW-NOOPT-NEXT: leal (%rdi,%rdi,4), %eax # sched: [1:0.50] ; HSW-NOOPT-NEXT: addl $2, %eax # sched: [1:0.25] -; HSW-NOOPT-NEXT: imull %ecx, %eax # sched: [4:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: imull %ecx, %eax # sched: [3:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_spec: ; JAG-NOOPT: # BB#0: diff --git a/llvm/test/CodeGen/X86/mul-constant-i64.ll b/llvm/test/CodeGen/X86/mul-constant-i64.ll index ea841c761c7..bf732f2f536 100644 --- a/llvm/test/CodeGen/X86/mul-constant-i64.ll +++ b/llvm/test/CodeGen/X86/mul-constant-i64.ll @@ -18,7 +18,7 @@ define i64 @test_mul_by_1(i64 %x) nounwind { ; X64-HSW-LABEL: test_mul_by_1: ; X64-HSW: # BB#0: ; X64-HSW-NEXT: movq %rdi, %rax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_1: ; X64-JAG: # BB#0: @@ -34,7 +34,7 @@ define i64 @test_mul_by_1(i64 %x) nounwind { ; HSW-NOOPT-LABEL: test_mul_by_1: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: movq %rdi, %rax # sched: [1:0.25] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_1: ; JAG-NOOPT: # BB#0: @@ -66,7 +66,7 @@ define i64 @test_mul_by_2(i64 %x) { ; X64-HSW-LABEL: test_mul_by_2: ; X64-HSW: # BB#0: ; X64-HSW-NEXT: leaq (%rdi,%rdi), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_2: ; X64-JAG: # BB#0: @@ -84,7 +84,7 @@ define i64 @test_mul_by_2(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_2: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: leaq (%rdi,%rdi), %rax # sched: [1:0.50] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_2: ; JAG-NOOPT: # BB#0: @@ -116,7 +116,7 @@ define i64 @test_mul_by_3(i64 %x) { ; X64-HSW-LABEL: test_mul_by_3: ; X64-HSW: # BB#0: ; X64-HSW-NEXT: leaq (%rdi,%rdi,2), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_3: ; X64-JAG: # BB#0: @@ -134,7 +134,7 @@ define i64 @test_mul_by_3(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_3: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: leaq (%rdi,%rdi,2), %rax # sched: [1:0.50] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_3: ; JAG-NOOPT: # BB#0: @@ -166,7 +166,7 @@ define i64 @test_mul_by_4(i64 %x) { ; X64-HSW-LABEL: test_mul_by_4: ; X64-HSW: # BB#0: ; X64-HSW-NEXT: leaq (,%rdi,4), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_4: ; X64-JAG: # BB#0: @@ -184,7 +184,7 @@ define i64 @test_mul_by_4(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_4: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: leaq (,%rdi,4), %rax # sched: [1:0.50] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_4: ; JAG-NOOPT: # BB#0: @@ -216,7 +216,7 @@ define i64 @test_mul_by_5(i64 %x) { ; X64-HSW-LABEL: test_mul_by_5: ; X64-HSW: # BB#0: ; X64-HSW-NEXT: leaq (%rdi,%rdi,4), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_5: ; X64-JAG: # BB#0: @@ -234,7 +234,7 @@ define i64 @test_mul_by_5(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_5: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: leaq (%rdi,%rdi,4), %rax # sched: [1:0.50] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_5: ; JAG-NOOPT: # BB#0: @@ -268,7 +268,7 @@ define i64 @test_mul_by_6(i64 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: addq %rdi, %rdi # sched: [1:0.25] ; X64-HSW-NEXT: leaq (%rdi,%rdi,2), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_6: ; X64-JAG: # BB#0: @@ -287,7 +287,7 @@ define i64 @test_mul_by_6(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_6: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $6, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_6: ; JAG-NOOPT: # BB#0: @@ -323,7 +323,7 @@ define i64 @test_mul_by_7(i64 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: leaq (,%rdi,8), %rax # sched: [1:0.50] ; X64-HSW-NEXT: subq %rdi, %rax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_7: ; X64-JAG: # BB#0: @@ -342,7 +342,7 @@ define i64 @test_mul_by_7(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_7: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $7, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_7: ; JAG-NOOPT: # BB#0: @@ -375,7 +375,7 @@ define i64 @test_mul_by_8(i64 %x) { ; X64-HSW-LABEL: test_mul_by_8: ; X64-HSW: # BB#0: ; X64-HSW-NEXT: leaq (,%rdi,8), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_8: ; X64-JAG: # BB#0: @@ -393,7 +393,7 @@ define i64 @test_mul_by_8(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_8: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: leaq (,%rdi,8), %rax # sched: [1:0.50] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_8: ; JAG-NOOPT: # BB#0: @@ -425,7 +425,7 @@ define i64 @test_mul_by_9(i64 %x) { ; X64-HSW-LABEL: test_mul_by_9: ; X64-HSW: # BB#0: ; X64-HSW-NEXT: leaq (%rdi,%rdi,8), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_9: ; X64-JAG: # BB#0: @@ -443,7 +443,7 @@ define i64 @test_mul_by_9(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_9: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: leaq (%rdi,%rdi,8), %rax # sched: [1:0.50] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_9: ; JAG-NOOPT: # BB#0: @@ -477,7 +477,7 @@ define i64 @test_mul_by_10(i64 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: addq %rdi, %rdi # sched: [1:0.25] ; X64-HSW-NEXT: leaq (%rdi,%rdi,4), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_10: ; X64-JAG: # BB#0: @@ -496,7 +496,7 @@ define i64 @test_mul_by_10(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_10: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $10, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_10: ; JAG-NOOPT: # BB#0: @@ -532,7 +532,7 @@ define i64 @test_mul_by_11(i64 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: leaq (%rdi,%rdi,4), %rax # sched: [1:0.50] ; X64-HSW-NEXT: leaq (%rdi,%rax,2), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_11: ; X64-JAG: # BB#0: @@ -551,7 +551,7 @@ define i64 @test_mul_by_11(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_11: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $11, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_11: ; JAG-NOOPT: # BB#0: @@ -585,7 +585,7 @@ define i64 @test_mul_by_12(i64 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: shlq $2, %rdi # sched: [1:0.50] ; X64-HSW-NEXT: leaq (%rdi,%rdi,2), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_12: ; X64-JAG: # BB#0: @@ -604,7 +604,7 @@ define i64 @test_mul_by_12(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_12: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $12, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_12: ; JAG-NOOPT: # BB#0: @@ -640,7 +640,7 @@ define i64 @test_mul_by_13(i64 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: leaq (%rdi,%rdi,2), %rax # sched: [1:0.50] ; X64-HSW-NEXT: leaq (%rdi,%rax,4), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_13: ; X64-JAG: # BB#0: @@ -659,7 +659,7 @@ define i64 @test_mul_by_13(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_13: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $13, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_13: ; JAG-NOOPT: # BB#0: @@ -696,7 +696,7 @@ define i64 @test_mul_by_14(i64 %x) { ; X64-HSW-NEXT: leaq (%rdi,%rdi,2), %rax # sched: [1:0.50] ; X64-HSW-NEXT: leaq (%rdi,%rax,4), %rax # sched: [1:0.50] ; X64-HSW-NEXT: addq %rdi, %rax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_14: ; X64-JAG: # BB#0: @@ -716,7 +716,7 @@ define i64 @test_mul_by_14(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_14: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $14, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_14: ; JAG-NOOPT: # BB#0: @@ -751,7 +751,7 @@ define i64 @test_mul_by_15(i64 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: leaq (%rdi,%rdi,4), %rax # sched: [1:0.50] ; X64-HSW-NEXT: leaq (%rax,%rax,2), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_15: ; X64-JAG: # BB#0: @@ -770,7 +770,7 @@ define i64 @test_mul_by_15(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_15: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $15, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_15: ; JAG-NOOPT: # BB#0: @@ -804,7 +804,7 @@ define i64 @test_mul_by_16(i64 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: shlq $4, %rdi # sched: [1:0.50] ; X64-HSW-NEXT: movq %rdi, %rax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_16: ; X64-JAG: # BB#0: @@ -824,7 +824,7 @@ define i64 @test_mul_by_16(i64 %x) { ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: shlq $4, %rdi # sched: [1:0.50] ; HSW-NOOPT-NEXT: movq %rdi, %rax # sched: [1:0.25] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_16: ; JAG-NOOPT: # BB#0: @@ -864,7 +864,7 @@ define i64 @test_mul_by_17(i64 %x) { ; X64-HSW-NEXT: movq %rdi, %rax # sched: [1:0.25] ; X64-HSW-NEXT: shlq $4, %rax # sched: [1:0.50] ; X64-HSW-NEXT: leaq (%rax,%rdi), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_17: ; X64-JAG: # BB#0: @@ -884,7 +884,7 @@ define i64 @test_mul_by_17(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_17: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $17, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_17: ; JAG-NOOPT: # BB#0: @@ -920,7 +920,7 @@ define i64 @test_mul_by_18(i64 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: addq %rdi, %rdi # sched: [1:0.25] ; X64-HSW-NEXT: leaq (%rdi,%rdi,8), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_18: ; X64-JAG: # BB#0: @@ -939,7 +939,7 @@ define i64 @test_mul_by_18(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_18: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $18, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_18: ; JAG-NOOPT: # BB#0: @@ -977,7 +977,7 @@ define i64 @test_mul_by_19(i64 %x) { ; X64-HSW-NEXT: leaq (%rdi,%rdi,4), %rax # sched: [1:0.50] ; X64-HSW-NEXT: shlq $2, %rax # sched: [1:0.50] ; X64-HSW-NEXT: subq %rdi, %rax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_19: ; X64-JAG: # BB#0: @@ -997,7 +997,7 @@ define i64 @test_mul_by_19(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_19: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $19, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_19: ; JAG-NOOPT: # BB#0: @@ -1031,7 +1031,7 @@ define i64 @test_mul_by_20(i64 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: shlq $2, %rdi # sched: [1:0.50] ; X64-HSW-NEXT: leaq (%rdi,%rdi,4), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_20: ; X64-JAG: # BB#0: @@ -1050,7 +1050,7 @@ define i64 @test_mul_by_20(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_20: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $20, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_20: ; JAG-NOOPT: # BB#0: @@ -1086,7 +1086,7 @@ define i64 @test_mul_by_21(i64 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: leaq (%rdi,%rdi,4), %rax # sched: [1:0.50] ; X64-HSW-NEXT: leaq (%rdi,%rax,4), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_21: ; X64-JAG: # BB#0: @@ -1105,7 +1105,7 @@ define i64 @test_mul_by_21(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_21: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $21, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_21: ; JAG-NOOPT: # BB#0: @@ -1142,7 +1142,7 @@ define i64 @test_mul_by_22(i64 %x) { ; X64-HSW-NEXT: leaq (%rdi,%rdi,4), %rax # sched: [1:0.50] ; X64-HSW-NEXT: leaq (%rdi,%rax,4), %rax # sched: [1:0.50] ; X64-HSW-NEXT: addq %rdi, %rax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_22: ; X64-JAG: # BB#0: @@ -1162,7 +1162,7 @@ define i64 @test_mul_by_22(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_22: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $22, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_22: ; JAG-NOOPT: # BB#0: @@ -1199,7 +1199,7 @@ define i64 @test_mul_by_23(i64 %x) { ; X64-HSW-NEXT: leaq (%rdi,%rdi,2), %rax # sched: [1:0.50] ; X64-HSW-NEXT: shlq $3, %rax # sched: [1:0.50] ; X64-HSW-NEXT: subq %rdi, %rax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_23: ; X64-JAG: # BB#0: @@ -1219,7 +1219,7 @@ define i64 @test_mul_by_23(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_23: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $23, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_23: ; JAG-NOOPT: # BB#0: @@ -1253,7 +1253,7 @@ define i64 @test_mul_by_24(i64 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: shlq $3, %rdi # sched: [1:0.50] ; X64-HSW-NEXT: leaq (%rdi,%rdi,2), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_24: ; X64-JAG: # BB#0: @@ -1272,7 +1272,7 @@ define i64 @test_mul_by_24(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_24: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $24, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_24: ; JAG-NOOPT: # BB#0: @@ -1308,7 +1308,7 @@ define i64 @test_mul_by_25(i64 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: leaq (%rdi,%rdi,4), %rax # sched: [1:0.50] ; X64-HSW-NEXT: leaq (%rax,%rax,4), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_25: ; X64-JAG: # BB#0: @@ -1327,7 +1327,7 @@ define i64 @test_mul_by_25(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_25: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $25, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_25: ; JAG-NOOPT: # BB#0: @@ -1365,7 +1365,7 @@ define i64 @test_mul_by_26(i64 %x) { ; X64-HSW-NEXT: leaq (%rdi,%rdi,8), %rax # sched: [1:0.50] ; X64-HSW-NEXT: leaq (%rax,%rax,2), %rax # sched: [1:0.50] ; X64-HSW-NEXT: subq %rdi, %rax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_26: ; X64-JAG: # BB#0: @@ -1385,7 +1385,7 @@ define i64 @test_mul_by_26(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_26: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $26, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_26: ; JAG-NOOPT: # BB#0: @@ -1420,7 +1420,7 @@ define i64 @test_mul_by_27(i64 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: leaq (%rdi,%rdi,8), %rax # sched: [1:0.50] ; X64-HSW-NEXT: leaq (%rax,%rax,2), %rax # sched: [1:0.50] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_27: ; X64-JAG: # BB#0: @@ -1439,7 +1439,7 @@ define i64 @test_mul_by_27(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_27: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $27, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_27: ; JAG-NOOPT: # BB#0: @@ -1477,7 +1477,7 @@ define i64 @test_mul_by_28(i64 %x) { ; X64-HSW-NEXT: leaq (%rdi,%rdi,8), %rax # sched: [1:0.50] ; X64-HSW-NEXT: leaq (%rax,%rax,2), %rax # sched: [1:0.50] ; X64-HSW-NEXT: addq %rdi, %rax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_28: ; X64-JAG: # BB#0: @@ -1497,7 +1497,7 @@ define i64 @test_mul_by_28(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_28: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $28, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_28: ; JAG-NOOPT: # BB#0: @@ -1536,7 +1536,7 @@ define i64 @test_mul_by_29(i64 %x) { ; X64-HSW-NEXT: leaq (%rax,%rax,2), %rax # sched: [1:0.50] ; X64-HSW-NEXT: addq %rdi, %rax # sched: [1:0.25] ; X64-HSW-NEXT: addq %rdi, %rax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_29: ; X64-JAG: # BB#0: @@ -1557,7 +1557,7 @@ define i64 @test_mul_by_29(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_29: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $29, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_29: ; JAG-NOOPT: # BB#0: @@ -1596,7 +1596,7 @@ define i64 @test_mul_by_30(i64 %x) { ; X64-HSW-NEXT: shlq $5, %rax # sched: [1:0.50] ; X64-HSW-NEXT: subq %rdi, %rax # sched: [1:0.25] ; X64-HSW-NEXT: subq %rdi, %rax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_30: ; X64-JAG: # BB#0: @@ -1617,7 +1617,7 @@ define i64 @test_mul_by_30(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_30: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $30, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_30: ; JAG-NOOPT: # BB#0: @@ -1654,7 +1654,7 @@ define i64 @test_mul_by_31(i64 %x) { ; X64-HSW-NEXT: movq %rdi, %rax # sched: [1:0.25] ; X64-HSW-NEXT: shlq $5, %rax # sched: [1:0.50] ; X64-HSW-NEXT: subq %rdi, %rax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_31: ; X64-JAG: # BB#0: @@ -1674,7 +1674,7 @@ define i64 @test_mul_by_31(i64 %x) { ; HSW-NOOPT-LABEL: test_mul_by_31: ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: imulq $31, %rdi, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_31: ; JAG-NOOPT: # BB#0: @@ -1709,7 +1709,7 @@ define i64 @test_mul_by_32(i64 %x) { ; X64-HSW: # BB#0: ; X64-HSW-NEXT: shlq $5, %rdi # sched: [1:0.50] ; X64-HSW-NEXT: movq %rdi, %rax # sched: [1:0.25] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_by_32: ; X64-JAG: # BB#0: @@ -1729,7 +1729,7 @@ define i64 @test_mul_by_32(i64 %x) { ; HSW-NOOPT: # BB#0: ; HSW-NOOPT-NEXT: shlq $5, %rdi # sched: [1:0.50] ; HSW-NOOPT-NEXT: movq %rdi, %rax # sched: [1:0.25] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_by_32: ; JAG-NOOPT: # BB#0: @@ -1793,7 +1793,7 @@ define i64 @test_mul_spec(i64 %x) nounwind { ; X64-HSW-NEXT: leaq (%rdi,%rdi,4), %rax # sched: [1:0.50] ; X64-HSW-NEXT: addq $2, %rax # sched: [1:0.25] ; X64-HSW-NEXT: imulq %rcx, %rax # sched: [3:1.00] -; X64-HSW-NEXT: retq # sched: [1:1.00] +; X64-HSW-NEXT: retq # sched: [2:1.00] ; ; X64-JAG-LABEL: test_mul_spec: ; X64-JAG: # BB#0: @@ -1841,7 +1841,7 @@ define i64 @test_mul_spec(i64 %x) nounwind { ; HSW-NOOPT-NEXT: leaq (%rdi,%rdi,4), %rax # sched: [1:0.50] ; HSW-NOOPT-NEXT: addq $2, %rax # sched: [1:0.25] ; HSW-NOOPT-NEXT: imulq %rcx, %rax # sched: [3:1.00] -; HSW-NOOPT-NEXT: retq # sched: [1:1.00] +; HSW-NOOPT-NEXT: retq # sched: [2:1.00] ; ; JAG-NOOPT-LABEL: test_mul_spec: ; JAG-NOOPT: # BB#0: diff --git a/llvm/test/CodeGen/X86/pr32329.ll b/llvm/test/CodeGen/X86/pr32329.ll index f2b79b67877..7cb38863e89 100644 --- a/llvm/test/CodeGen/X86/pr32329.ll +++ b/llvm/test/CodeGen/X86/pr32329.ll @@ -59,8 +59,8 @@ define void @foo() local_unnamed_addr { ; X86-NEXT: cmovnel %ecx, %esi ; X86-NEXT: cmpl %edx, %edi ; X86-NEXT: movl %ebp, var_50+4 -; X86-NEXT: movl %esi, var_50 ; X86-NEXT: setge var_205 +; X86-NEXT: movl %esi, var_50 ; X86-NEXT: imull %eax, %ebx ; X86-NEXT: movb %bl, var_218 ; X86-NEXT: popl %esi diff --git a/llvm/test/CodeGen/X86/recip-fastmath.ll b/llvm/test/CodeGen/X86/recip-fastmath.ll index 16e261bf3c5..e9e1ee5e742 100644 --- a/llvm/test/CodeGen/X86/recip-fastmath.ll +++ b/llvm/test/CodeGen/X86/recip-fastmath.ll @@ -45,15 +45,15 @@ define float @f32_no_estimate(float %x) #0 { ; ; SANDY-LABEL: f32_no_estimate: ; SANDY: # BB#0: -; SANDY-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [4:0.50] -; SANDY-NEXT: vdivss %xmm0, %xmm1, %xmm0 # sched: [12:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [6:0.50] +; SANDY-NEXT: vdivss %xmm0, %xmm1, %xmm0 # sched: [14:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: f32_no_estimate: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [4:0.50] -; HASWELL-NEXT: vdivss %xmm0, %xmm1, %xmm0 # sched: [12:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [?:5.000000e-01] +; HASWELL-NEXT: vdivss %xmm0, %xmm1, %xmm0 # sched: [13:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: f32_no_estimate: ; HASWELL-NO-FMA: # BB#0: @@ -63,9 +63,9 @@ define float @f32_no_estimate(float %x) #0 { ; ; AVX512-LABEL: f32_no_estimate: ; AVX512: # BB#0: -; AVX512-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [4:0.50] -; AVX512-NEXT: vdivss %xmm0, %xmm1, %xmm0 # sched: [12:1.00] -; AVX512-NEXT: retq # sched: [1:1.00] +; AVX512-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [?:5.000000e-01] +; AVX512-NEXT: vdivss %xmm0, %xmm1, %xmm0 # sched: [13:1.00] +; AVX512-NEXT: retq # sched: [2:1.00] %div = fdiv fast float 1.0, %x ret float %div } @@ -113,18 +113,18 @@ define float @f32_one_step(float %x) #1 { ; SANDY: # BB#0: ; SANDY-NEXT: vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00] ; SANDY-NEXT: vmulss %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50] +; SANDY-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [6:0.50] ; SANDY-NEXT: vsubss %xmm0, %xmm2, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00] ; SANDY-NEXT: vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: f32_one_step: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00] -; HASWELL-NEXT: vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 -; HASWELL-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm0 -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: f32_one_step: ; HASWELL-NO-FMA: # BB#0: @@ -139,9 +139,9 @@ define float @f32_one_step(float %x) #1 { ; AVX512-LABEL: f32_one_step: ; AVX512: # BB#0: ; AVX512-NEXT: vrcp14ss %xmm0, %xmm0, %xmm1 -; AVX512-NEXT: vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 -; AVX512-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm0 -; AVX512-NEXT: retq # sched: [1:1.00] +; AVX512-NEXT: vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 # sched: [5:0.50] +; AVX512-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm0 # sched: [5:0.50] +; AVX512-NEXT: retq # sched: [2:1.00] %div = fdiv fast float 1.0, %x ret float %div } @@ -207,7 +207,7 @@ define float @f32_two_step(float %x) #2 { ; SANDY: # BB#0: ; SANDY-NEXT: vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00] ; SANDY-NEXT: vmulss %xmm1, %xmm0, %xmm2 # sched: [5:1.00] -; SANDY-NEXT: vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero sched: [4:0.50] +; SANDY-NEXT: vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero sched: [6:0.50] ; SANDY-NEXT: vsubss %xmm2, %xmm3, %xmm2 # sched: [3:1.00] ; SANDY-NEXT: vmulss %xmm2, %xmm1, %xmm2 # sched: [5:1.00] ; SANDY-NEXT: vaddss %xmm2, %xmm1, %xmm1 # sched: [3:1.00] @@ -215,18 +215,18 @@ define float @f32_two_step(float %x) #2 { ; SANDY-NEXT: vsubss %xmm0, %xmm3, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00] ; SANDY-NEXT: vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: f32_two_step: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00] -; HASWELL-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50] +; HASWELL-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [?:5.000000e-01] ; HASWELL-NEXT: vmovaps %xmm1, %xmm3 # sched: [1:1.00] -; HASWELL-NEXT: vfnmadd213ss %xmm2, %xmm0, %xmm3 -; HASWELL-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm3 -; HASWELL-NEXT: vfnmadd213ss %xmm2, %xmm3, %xmm0 -; HASWELL-NEXT: vfmadd132ss %xmm3, %xmm3, %xmm0 -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vfnmadd213ss %xmm2, %xmm0, %xmm3 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm3 # sched: [5:0.50] +; HASWELL-NEXT: vfnmadd213ss %xmm2, %xmm3, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ss %xmm3, %xmm3, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: f32_two_step: ; HASWELL-NO-FMA: # BB#0: @@ -245,13 +245,13 @@ define float @f32_two_step(float %x) #2 { ; AVX512-LABEL: f32_two_step: ; AVX512: # BB#0: ; AVX512-NEXT: vrcp14ss %xmm0, %xmm0, %xmm1 -; AVX512-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50] +; AVX512-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [?:5.000000e-01] ; AVX512-NEXT: vmovaps %xmm1, %xmm3 # sched: [1:1.00] -; AVX512-NEXT: vfnmadd213ss %xmm2, %xmm0, %xmm3 -; AVX512-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm3 -; AVX512-NEXT: vfnmadd213ss %xmm2, %xmm3, %xmm0 -; AVX512-NEXT: vfmadd132ss %xmm3, %xmm3, %xmm0 -; AVX512-NEXT: retq # sched: [1:1.00] +; AVX512-NEXT: vfnmadd213ss %xmm2, %xmm0, %xmm3 # sched: [5:0.50] +; AVX512-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm3 # sched: [5:0.50] +; AVX512-NEXT: vfnmadd213ss %xmm2, %xmm3, %xmm0 # sched: [5:0.50] +; AVX512-NEXT: vfmadd132ss %xmm3, %xmm3, %xmm0 # sched: [5:0.50] +; AVX512-NEXT: retq # sched: [2:1.00] %div = fdiv fast float 1.0, %x ret float %div } @@ -284,15 +284,15 @@ define <4 x float> @v4f32_no_estimate(<4 x float> %x) #0 { ; ; SANDY-LABEL: v4f32_no_estimate: ; SANDY: # BB#0: -; SANDY-NEXT: vmovaps {{.*#+}} xmm1 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50] -; SANDY-NEXT: vdivps %xmm0, %xmm1, %xmm0 # sched: [12:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovaps {{.*#+}} xmm1 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [6:0.50] +; SANDY-NEXT: vdivps %xmm0, %xmm1, %xmm0 # sched: [14:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: v4f32_no_estimate: ; HASWELL: # BB#0: -; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %xmm1 # sched: [4:0.50] -; HASWELL-NEXT: vdivps %xmm0, %xmm1, %xmm0 # sched: [12:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %xmm1 # sched: [?:5.000000e-01] +; HASWELL-NEXT: vdivps %xmm0, %xmm1, %xmm0 # sched: [13:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: v4f32_no_estimate: ; HASWELL-NO-FMA: # BB#0: @@ -302,9 +302,9 @@ define <4 x float> @v4f32_no_estimate(<4 x float> %x) #0 { ; ; AVX512-LABEL: v4f32_no_estimate: ; AVX512: # BB#0: -; AVX512-NEXT: vbroadcastss {{.*}}(%rip), %xmm1 # sched: [4:0.50] -; AVX512-NEXT: vdivps %xmm0, %xmm1, %xmm0 # sched: [12:1.00] -; AVX512-NEXT: retq # sched: [1:1.00] +; AVX512-NEXT: vbroadcastss {{.*}}(%rip), %xmm1 # sched: [?:5.000000e-01] +; AVX512-NEXT: vdivps %xmm0, %xmm1, %xmm0 # sched: [13:1.00] +; AVX512-NEXT: retq # sched: [2:1.00] %div = fdiv fast <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>, %x ret <4 x float> %div } @@ -350,21 +350,21 @@ define <4 x float> @v4f32_one_step(<4 x float> %x) #1 { ; ; SANDY-LABEL: v4f32_one_step: ; SANDY: # BB#0: -; SANDY-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] +; SANDY-NEXT: vrcpps %xmm0, %xmm1 # sched: [7:3.00] ; SANDY-NEXT: vmulps %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [6:0.50] ; SANDY-NEXT: vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00] ; SANDY-NEXT: vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: v4f32_one_step: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] -; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50] -; HASWELL-NEXT: vfnmadd213ps %xmm2, %xmm1, %xmm0 -; HASWELL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01] +; HASWELL-NEXT: vfnmadd213ps %xmm2, %xmm1, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: v4f32_one_step: ; HASWELL-NO-FMA: # BB#0: @@ -379,17 +379,17 @@ define <4 x float> @v4f32_one_step(<4 x float> %x) #1 { ; KNL-LABEL: v4f32_one_step: ; KNL: # BB#0: ; KNL-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] -; KNL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50] -; KNL-NEXT: vfnmadd213ps %xmm2, %xmm1, %xmm0 -; KNL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 -; KNL-NEXT: retq # sched: [1:1.00] +; KNL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01] +; KNL-NEXT: vfnmadd213ps %xmm2, %xmm1, %xmm0 # sched: [5:0.50] +; KNL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50] +; KNL-NEXT: retq # sched: [2:1.00] ; ; SKX-LABEL: v4f32_one_step: ; SKX: # BB#0: ; SKX-NEXT: vrcp14ps %xmm0, %xmm1 ; SKX-NEXT: vfnmadd213ps {{.*}}(%rip){1to4}, %xmm1, %xmm0 -; SKX-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 -; SKX-NEXT: retq # sched: [1:1.00] +; SKX-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50] +; SKX-NEXT: retq # sched: [2:1.00] %div = fdiv fast <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>, %x ret <4 x float> %div } @@ -453,9 +453,9 @@ define <4 x float> @v4f32_two_step(<4 x float> %x) #2 { ; ; SANDY-LABEL: v4f32_two_step: ; SANDY: # BB#0: -; SANDY-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] +; SANDY-NEXT: vrcpps %xmm0, %xmm1 # sched: [7:3.00] ; SANDY-NEXT: vmulps %xmm1, %xmm0, %xmm2 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} xmm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} xmm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [6:0.50] ; SANDY-NEXT: vsubps %xmm2, %xmm3, %xmm2 # sched: [3:1.00] ; SANDY-NEXT: vmulps %xmm2, %xmm1, %xmm2 # sched: [5:1.00] ; SANDY-NEXT: vaddps %xmm2, %xmm1, %xmm1 # sched: [3:1.00] @@ -463,18 +463,18 @@ define <4 x float> @v4f32_two_step(<4 x float> %x) #2 { ; SANDY-NEXT: vsubps %xmm0, %xmm3, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00] ; SANDY-NEXT: vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: v4f32_two_step: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] -; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50] +; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01] ; HASWELL-NEXT: vmovaps %xmm1, %xmm3 # sched: [1:1.00] -; HASWELL-NEXT: vfnmadd213ps %xmm2, %xmm0, %xmm3 -; HASWELL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm3 -; HASWELL-NEXT: vfnmadd213ps %xmm2, %xmm3, %xmm0 -; HASWELL-NEXT: vfmadd132ps %xmm3, %xmm3, %xmm0 -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vfnmadd213ps %xmm2, %xmm0, %xmm3 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm3 # sched: [5:0.50] +; HASWELL-NEXT: vfnmadd213ps %xmm2, %xmm3, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ps %xmm3, %xmm3, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: v4f32_two_step: ; HASWELL-NO-FMA: # BB#0: @@ -493,24 +493,24 @@ define <4 x float> @v4f32_two_step(<4 x float> %x) #2 { ; KNL-LABEL: v4f32_two_step: ; KNL: # BB#0: ; KNL-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] -; KNL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50] +; KNL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01] ; KNL-NEXT: vmovaps %xmm1, %xmm3 # sched: [1:1.00] -; KNL-NEXT: vfnmadd213ps %xmm2, %xmm0, %xmm3 -; KNL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm3 -; KNL-NEXT: vfnmadd213ps %xmm2, %xmm3, %xmm0 -; KNL-NEXT: vfmadd132ps %xmm3, %xmm3, %xmm0 -; KNL-NEXT: retq # sched: [1:1.00] +; KNL-NEXT: vfnmadd213ps %xmm2, %xmm0, %xmm3 # sched: [5:0.50] +; KNL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm3 # sched: [5:0.50] +; KNL-NEXT: vfnmadd213ps %xmm2, %xmm3, %xmm0 # sched: [5:0.50] +; KNL-NEXT: vfmadd132ps %xmm3, %xmm3, %xmm0 # sched: [5:0.50] +; KNL-NEXT: retq # sched: [2:1.00] ; ; SKX-LABEL: v4f32_two_step: ; SKX: # BB#0: ; SKX-NEXT: vrcp14ps %xmm0, %xmm1 -; SKX-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50] +; SKX-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01] ; SKX-NEXT: vmovaps %xmm1, %xmm3 # sched: [1:1.00] -; SKX-NEXT: vfnmadd213ps %xmm2, %xmm0, %xmm3 -; SKX-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm3 -; SKX-NEXT: vfnmadd213ps %xmm2, %xmm3, %xmm0 -; SKX-NEXT: vfmadd132ps %xmm3, %xmm3, %xmm0 -; SKX-NEXT: retq # sched: [1:1.00] +; SKX-NEXT: vfnmadd213ps %xmm2, %xmm0, %xmm3 # sched: [5:0.50] +; SKX-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm3 # sched: [5:0.50] +; SKX-NEXT: vfnmadd213ps %xmm2, %xmm3, %xmm0 # sched: [5:0.50] +; SKX-NEXT: vfmadd132ps %xmm3, %xmm3, %xmm0 # sched: [5:0.50] +; SKX-NEXT: retq # sched: [2:1.00] %div = fdiv fast <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>, %x ret <4 x float> %div } @@ -546,15 +546,15 @@ define <8 x float> @v8f32_no_estimate(<8 x float> %x) #0 { ; ; SANDY-LABEL: v8f32_no_estimate: ; SANDY: # BB#0: -; SANDY-NEXT: vmovaps {{.*#+}} ymm1 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50] -; SANDY-NEXT: vdivps %ymm0, %ymm1, %ymm0 # sched: [12:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovaps {{.*#+}} ymm1 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50] +; SANDY-NEXT: vdivps %ymm0, %ymm1, %ymm0 # sched: [29:3.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: v8f32_no_estimate: ; HASWELL: # BB#0: ; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %ymm1 # sched: [5:1.00] -; HASWELL-NEXT: vdivps %ymm0, %ymm1, %ymm0 # sched: [19:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vdivps %ymm0, %ymm1, %ymm0 # sched: [21:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: v8f32_no_estimate: ; HASWELL-NO-FMA: # BB#0: @@ -565,8 +565,8 @@ define <8 x float> @v8f32_no_estimate(<8 x float> %x) #0 { ; AVX512-LABEL: v8f32_no_estimate: ; AVX512: # BB#0: ; AVX512-NEXT: vbroadcastss {{.*}}(%rip), %ymm1 # sched: [5:1.00] -; AVX512-NEXT: vdivps %ymm0, %ymm1, %ymm0 # sched: [19:2.00] -; AVX512-NEXT: retq # sched: [1:1.00] +; AVX512-NEXT: vdivps %ymm0, %ymm1, %ymm0 # sched: [21:2.00] +; AVX512-NEXT: retq # sched: [2:1.00] %div = fdiv fast <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %x ret <8 x float> %div } @@ -621,19 +621,19 @@ define <8 x float> @v8f32_one_step(<8 x float> %x) #1 { ; SANDY: # BB#0: ; SANDY-NEXT: vrcpps %ymm0, %ymm1 # sched: [5:1.00] ; SANDY-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50] ; SANDY-NEXT: vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00] ; SANDY-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00] ; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: v8f32_one_step: ; HASWELL: # BB#0: -; HASWELL-NEXT: vrcpps %ymm0, %ymm1 # sched: [7:2.00] +; HASWELL-NEXT: vrcpps %ymm0, %ymm1 # sched: [11:2.00] ; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00] -; HASWELL-NEXT: vfnmadd213ps %ymm2, %ymm1, %ymm0 -; HASWELL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vfnmadd213ps %ymm2, %ymm1, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: v8f32_one_step: ; HASWELL-NO-FMA: # BB#0: @@ -647,18 +647,18 @@ define <8 x float> @v8f32_one_step(<8 x float> %x) #1 { ; ; KNL-LABEL: v8f32_one_step: ; KNL: # BB#0: -; KNL-NEXT: vrcpps %ymm0, %ymm1 # sched: [7:2.00] +; KNL-NEXT: vrcpps %ymm0, %ymm1 # sched: [11:2.00] ; KNL-NEXT: vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00] -; KNL-NEXT: vfnmadd213ps %ymm2, %ymm1, %ymm0 -; KNL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 -; KNL-NEXT: retq # sched: [1:1.00] +; KNL-NEXT: vfnmadd213ps %ymm2, %ymm1, %ymm0 # sched: [5:0.50] +; KNL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50] +; KNL-NEXT: retq # sched: [2:1.00] ; ; SKX-LABEL: v8f32_one_step: ; SKX: # BB#0: ; SKX-NEXT: vrcp14ps %ymm0, %ymm1 ; SKX-NEXT: vfnmadd213ps {{.*}}(%rip){1to8}, %ymm1, %ymm0 -; SKX-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 -; SKX-NEXT: retq # sched: [1:1.00] +; SKX-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50] +; SKX-NEXT: retq # sched: [2:1.00] %div = fdiv fast <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %x ret <8 x float> %div } @@ -737,7 +737,7 @@ define <8 x float> @v8f32_two_step(<8 x float> %x) #2 { ; SANDY: # BB#0: ; SANDY-NEXT: vrcpps %ymm0, %ymm1 # sched: [5:1.00] ; SANDY-NEXT: vmulps %ymm1, %ymm0, %ymm2 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50] ; SANDY-NEXT: vsubps %ymm2, %ymm3, %ymm2 # sched: [3:1.00] ; SANDY-NEXT: vmulps %ymm2, %ymm1, %ymm2 # sched: [5:1.00] ; SANDY-NEXT: vaddps %ymm2, %ymm1, %ymm1 # sched: [3:1.00] @@ -745,18 +745,18 @@ define <8 x float> @v8f32_two_step(<8 x float> %x) #2 { ; SANDY-NEXT: vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00] ; SANDY-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00] ; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: v8f32_two_step: ; HASWELL: # BB#0: -; HASWELL-NEXT: vrcpps %ymm0, %ymm1 # sched: [7:2.00] +; HASWELL-NEXT: vrcpps %ymm0, %ymm1 # sched: [11:2.00] ; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00] ; HASWELL-NEXT: vmovaps %ymm1, %ymm3 # sched: [1:1.00] -; HASWELL-NEXT: vfnmadd213ps %ymm2, %ymm0, %ymm3 -; HASWELL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm3 -; HASWELL-NEXT: vfnmadd213ps %ymm2, %ymm3, %ymm0 -; HASWELL-NEXT: vfmadd132ps %ymm3, %ymm3, %ymm0 -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vfnmadd213ps %ymm2, %ymm0, %ymm3 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm3 # sched: [5:0.50] +; HASWELL-NEXT: vfnmadd213ps %ymm2, %ymm3, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ps %ymm3, %ymm3, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: v8f32_two_step: ; HASWELL-NO-FMA: # BB#0: @@ -774,25 +774,25 @@ define <8 x float> @v8f32_two_step(<8 x float> %x) #2 { ; ; KNL-LABEL: v8f32_two_step: ; KNL: # BB#0: -; KNL-NEXT: vrcpps %ymm0, %ymm1 # sched: [7:2.00] +; KNL-NEXT: vrcpps %ymm0, %ymm1 # sched: [11:2.00] ; KNL-NEXT: vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00] ; KNL-NEXT: vmovaps %ymm1, %ymm3 # sched: [1:1.00] -; KNL-NEXT: vfnmadd213ps %ymm2, %ymm0, %ymm3 -; KNL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm3 -; KNL-NEXT: vfnmadd213ps %ymm2, %ymm3, %ymm0 -; KNL-NEXT: vfmadd132ps %ymm3, %ymm3, %ymm0 -; KNL-NEXT: retq # sched: [1:1.00] +; KNL-NEXT: vfnmadd213ps %ymm2, %ymm0, %ymm3 # sched: [5:0.50] +; KNL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm3 # sched: [5:0.50] +; KNL-NEXT: vfnmadd213ps %ymm2, %ymm3, %ymm0 # sched: [5:0.50] +; KNL-NEXT: vfmadd132ps %ymm3, %ymm3, %ymm0 # sched: [5:0.50] +; KNL-NEXT: retq # sched: [2:1.00] ; ; SKX-LABEL: v8f32_two_step: ; SKX: # BB#0: ; SKX-NEXT: vrcp14ps %ymm0, %ymm1 ; SKX-NEXT: vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00] ; SKX-NEXT: vmovaps %ymm1, %ymm3 # sched: [1:1.00] -; SKX-NEXT: vfnmadd213ps %ymm2, %ymm0, %ymm3 -; SKX-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm3 -; SKX-NEXT: vfnmadd213ps %ymm2, %ymm3, %ymm0 -; SKX-NEXT: vfmadd132ps %ymm3, %ymm3, %ymm0 -; SKX-NEXT: retq # sched: [1:1.00] +; SKX-NEXT: vfnmadd213ps %ymm2, %ymm0, %ymm3 # sched: [5:0.50] +; SKX-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm3 # sched: [5:0.50] +; SKX-NEXT: vfnmadd213ps %ymm2, %ymm3, %ymm0 # sched: [5:0.50] +; SKX-NEXT: vfmadd132ps %ymm3, %ymm3, %ymm0 # sched: [5:0.50] +; SKX-NEXT: retq # sched: [2:1.00] %div = fdiv fast <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %x ret <8 x float> %div } diff --git a/llvm/test/CodeGen/X86/recip-fastmath2.ll b/llvm/test/CodeGen/X86/recip-fastmath2.ll index 440a6f0bef1..a3ac4596c07 100644 --- a/llvm/test/CodeGen/X86/recip-fastmath2.ll +++ b/llvm/test/CodeGen/X86/recip-fastmath2.ll @@ -39,26 +39,26 @@ define float @f32_no_step_2(float %x) #3 { ; SANDY-LABEL: f32_no_step_2: ; SANDY: # BB#0: ; SANDY-NEXT: vrcpss %xmm0, %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: f32_no_step_2: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrcpss %xmm0, %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: f32_no_step_2: ; HASWELL-NO-FMA: # BB#0: ; HASWELL-NO-FMA-NEXT: vrcpss %xmm0, %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NO-FMA-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; HASWELL-NO-FMA-NEXT: retq # sched: [1:1.00] +; HASWELL-NO-FMA-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NO-FMA-NEXT: retq # sched: [2:1.00] ; ; AVX512-LABEL: f32_no_step_2: ; AVX512: # BB#0: ; AVX512-NEXT: vrcp14ss %xmm0, %xmm0, %xmm0 -; AVX512-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; AVX512-NEXT: retq # sched: [1:1.00] +; AVX512-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; AVX512-NEXT: retq # sched: [2:1.00] %div = fdiv fast float 1234.0, %x ret float %div } @@ -110,39 +110,39 @@ define float @f32_one_step_2(float %x) #1 { ; SANDY: # BB#0: ; SANDY-NEXT: vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00] ; SANDY-NEXT: vmulss %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50] +; SANDY-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [6:0.50] ; SANDY-NEXT: vsubss %xmm0, %xmm2, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00] ; SANDY-NEXT: vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: f32_one_step_2: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00] -; HASWELL-NEXT: vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 -; HASWELL-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm0 -; HASWELL-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: f32_one_step_2: ; HASWELL-NO-FMA: # BB#0: ; HASWELL-NO-FMA-NEXT: vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00] ; HASWELL-NO-FMA-NEXT: vmulss %xmm1, %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NO-FMA-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50] +; HASWELL-NO-FMA-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [?:5.000000e-01] ; HASWELL-NO-FMA-NEXT: vsubss %xmm0, %xmm2, %xmm0 # sched: [3:1.00] ; HASWELL-NO-FMA-NEXT: vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NO-FMA-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; HASWELL-NO-FMA-NEXT: retq # sched: [1:1.00] +; HASWELL-NO-FMA-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NO-FMA-NEXT: retq # sched: [2:1.00] ; ; AVX512-LABEL: f32_one_step_2: ; AVX512: # BB#0: ; AVX512-NEXT: vrcp14ss %xmm0, %xmm0, %xmm1 -; AVX512-NEXT: vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 -; AVX512-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm0 -; AVX512-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; AVX512-NEXT: retq # sched: [1:1.00] +; AVX512-NEXT: vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 # sched: [5:0.50] +; AVX512-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm0 # sched: [5:0.50] +; AVX512-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; AVX512-NEXT: retq # sched: [2:1.00] %div = fdiv fast float 3456.0, %x ret float %div } @@ -198,43 +198,43 @@ define float @f32_one_step_2_divs(float %x) #1 { ; SANDY: # BB#0: ; SANDY-NEXT: vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00] ; SANDY-NEXT: vmulss %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50] +; SANDY-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [6:0.50] ; SANDY-NEXT: vsubss %xmm0, %xmm2, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00] ; SANDY-NEXT: vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:1.00] +; SANDY-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [11:1.00] ; SANDY-NEXT: vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: f32_one_step_2_divs: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00] -; HASWELL-NEXT: vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 -; HASWELL-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm0 -; HASWELL-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50] +; HASWELL-NEXT: vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [5:0.50] ; HASWELL-NEXT: vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: f32_one_step_2_divs: ; HASWELL-NO-FMA: # BB#0: ; HASWELL-NO-FMA-NEXT: vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00] ; HASWELL-NO-FMA-NEXT: vmulss %xmm1, %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NO-FMA-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50] +; HASWELL-NO-FMA-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [?:5.000000e-01] ; HASWELL-NO-FMA-NEXT: vsubss %xmm0, %xmm2, %xmm0 # sched: [3:1.00] ; HASWELL-NO-FMA-NEXT: vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NO-FMA-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50] +; HASWELL-NO-FMA-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50] -; HASWELL-NO-FMA-NEXT: retq # sched: [1:1.00] +; HASWELL-NO-FMA-NEXT: retq # sched: [2:1.00] ; ; AVX512-LABEL: f32_one_step_2_divs: ; AVX512: # BB#0: ; AVX512-NEXT: vrcp14ss %xmm0, %xmm0, %xmm1 -; AVX512-NEXT: vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 -; AVX512-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm0 -; AVX512-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50] +; AVX512-NEXT: vfnmadd213ss {{.*}}(%rip), %xmm1, %xmm0 # sched: [5:0.50] +; AVX512-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm0 # sched: [5:0.50] +; AVX512-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm1 # sched: [5:0.50] ; AVX512-NEXT: vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50] -; AVX512-NEXT: retq # sched: [1:1.00] +; AVX512-NEXT: retq # sched: [2:1.00] %div = fdiv fast float 3456.0, %x %div2 = fdiv fast float %div, %x ret float %div2 @@ -305,7 +305,7 @@ define float @f32_two_step_2(float %x) #2 { ; SANDY: # BB#0: ; SANDY-NEXT: vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00] ; SANDY-NEXT: vmulss %xmm1, %xmm0, %xmm2 # sched: [5:1.00] -; SANDY-NEXT: vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero sched: [4:0.50] +; SANDY-NEXT: vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero sched: [6:0.50] ; SANDY-NEXT: vsubss %xmm2, %xmm3, %xmm2 # sched: [3:1.00] ; SANDY-NEXT: vmulss %xmm2, %xmm1, %xmm2 # sched: [5:1.00] ; SANDY-NEXT: vaddss %xmm2, %xmm1, %xmm1 # sched: [3:1.00] @@ -313,26 +313,26 @@ define float @f32_two_step_2(float %x) #2 { ; SANDY-NEXT: vsubss %xmm0, %xmm3, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vmulss %xmm0, %xmm1, %xmm0 # sched: [5:1.00] ; SANDY-NEXT: vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: f32_two_step_2: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00] -; HASWELL-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50] +; HASWELL-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [?:5.000000e-01] ; HASWELL-NEXT: vmovaps %xmm1, %xmm3 # sched: [1:1.00] -; HASWELL-NEXT: vfnmadd213ss %xmm2, %xmm0, %xmm3 -; HASWELL-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm3 -; HASWELL-NEXT: vfnmadd213ss %xmm2, %xmm3, %xmm0 -; HASWELL-NEXT: vfmadd132ss %xmm3, %xmm3, %xmm0 -; HASWELL-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vfnmadd213ss %xmm2, %xmm0, %xmm3 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm3 # sched: [5:0.50] +; HASWELL-NEXT: vfnmadd213ss %xmm2, %xmm3, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ss %xmm3, %xmm3, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: f32_two_step_2: ; HASWELL-NO-FMA: # BB#0: ; HASWELL-NO-FMA-NEXT: vrcpss %xmm0, %xmm0, %xmm1 # sched: [5:1.00] ; HASWELL-NO-FMA-NEXT: vmulss %xmm1, %xmm0, %xmm2 # sched: [5:0.50] -; HASWELL-NO-FMA-NEXT: vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero sched: [4:0.50] +; HASWELL-NO-FMA-NEXT: vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero sched: [?:5.000000e-01] ; HASWELL-NO-FMA-NEXT: vsubss %xmm2, %xmm3, %xmm2 # sched: [3:1.00] ; HASWELL-NO-FMA-NEXT: vmulss %xmm2, %xmm1, %xmm2 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vaddss %xmm2, %xmm1, %xmm1 # sched: [3:1.00] @@ -340,20 +340,20 @@ define float @f32_two_step_2(float %x) #2 { ; HASWELL-NO-FMA-NEXT: vsubss %xmm0, %xmm3, %xmm0 # sched: [3:1.00] ; HASWELL-NO-FMA-NEXT: vmulss %xmm0, %xmm1, %xmm0 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vaddss %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NO-FMA-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; HASWELL-NO-FMA-NEXT: retq # sched: [1:1.00] +; HASWELL-NO-FMA-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NO-FMA-NEXT: retq # sched: [2:1.00] ; ; AVX512-LABEL: f32_two_step_2: ; AVX512: # BB#0: ; AVX512-NEXT: vrcp14ss %xmm0, %xmm0, %xmm1 -; AVX512-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50] +; AVX512-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [?:5.000000e-01] ; AVX512-NEXT: vmovaps %xmm1, %xmm3 # sched: [1:1.00] -; AVX512-NEXT: vfnmadd213ss %xmm2, %xmm0, %xmm3 -; AVX512-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm3 -; AVX512-NEXT: vfnmadd213ss %xmm2, %xmm3, %xmm0 -; AVX512-NEXT: vfmadd132ss %xmm3, %xmm3, %xmm0 -; AVX512-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; AVX512-NEXT: retq # sched: [1:1.00] +; AVX512-NEXT: vfnmadd213ss %xmm2, %xmm0, %xmm3 # sched: [5:0.50] +; AVX512-NEXT: vfmadd132ss %xmm1, %xmm1, %xmm3 # sched: [5:0.50] +; AVX512-NEXT: vfnmadd213ss %xmm2, %xmm3, %xmm0 # sched: [5:0.50] +; AVX512-NEXT: vfmadd132ss %xmm3, %xmm3, %xmm0 # sched: [5:0.50] +; AVX512-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; AVX512-NEXT: retq # sched: [2:1.00] %div = fdiv fast float 6789.0, %x ret float %div } @@ -403,51 +403,51 @@ define <4 x float> @v4f32_one_step2(<4 x float> %x) #1 { ; ; SANDY-LABEL: v4f32_one_step2: ; SANDY: # BB#0: -; SANDY-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] +; SANDY-NEXT: vrcpps %xmm0, %xmm1 # sched: [7:3.00] ; SANDY-NEXT: vmulps %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [6:0.50] ; SANDY-NEXT: vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00] ; SANDY-NEXT: vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: v4f32_one_step2: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] -; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50] -; HASWELL-NEXT: vfnmadd213ps %xmm2, %xmm1, %xmm0 -; HASWELL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 -; HASWELL-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01] +; HASWELL-NEXT: vfnmadd213ps %xmm2, %xmm1, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: v4f32_one_step2: ; HASWELL-NO-FMA: # BB#0: ; HASWELL-NO-FMA-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] ; HASWELL-NO-FMA-NEXT: vmulps %xmm1, %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NO-FMA-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50] +; HASWELL-NO-FMA-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01] ; HASWELL-NO-FMA-NEXT: vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00] ; HASWELL-NO-FMA-NEXT: vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; HASWELL-NO-FMA-NEXT: retq # sched: [1:1.00] +; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NO-FMA-NEXT: retq # sched: [2:1.00] ; ; KNL-LABEL: v4f32_one_step2: ; KNL: # BB#0: ; KNL-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] -; KNL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50] -; KNL-NEXT: vfnmadd213ps %xmm2, %xmm1, %xmm0 -; KNL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 -; KNL-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; KNL-NEXT: retq # sched: [1:1.00] +; KNL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01] +; KNL-NEXT: vfnmadd213ps %xmm2, %xmm1, %xmm0 # sched: [5:0.50] +; KNL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50] +; KNL-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; KNL-NEXT: retq # sched: [2:1.00] ; ; SKX-LABEL: v4f32_one_step2: ; SKX: # BB#0: ; SKX-NEXT: vrcp14ps %xmm0, %xmm1 ; SKX-NEXT: vfnmadd213ps {{.*}}(%rip){1to4}, %xmm1, %xmm0 -; SKX-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 -; SKX-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; SKX-NEXT: retq # sched: [1:1.00] +; SKX-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50] +; SKX-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; SKX-NEXT: retq # sched: [2:1.00] %div = fdiv fast <4 x float> <float 1.0, float 2.0, float 3.0, float 4.0>, %x ret <4 x float> %div } @@ -501,56 +501,56 @@ define <4 x float> @v4f32_one_step_2_divs(<4 x float> %x) #1 { ; ; SANDY-LABEL: v4f32_one_step_2_divs: ; SANDY: # BB#0: -; SANDY-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] +; SANDY-NEXT: vrcpps %xmm0, %xmm1 # sched: [7:3.00] ; SANDY-NEXT: vmulps %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [6:0.50] ; SANDY-NEXT: vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00] ; SANDY-NEXT: vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:1.00] +; SANDY-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [11:1.00] ; SANDY-NEXT: vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: v4f32_one_step_2_divs: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] -; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50] -; HASWELL-NEXT: vfnmadd213ps %xmm2, %xmm1, %xmm0 -; HASWELL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 -; HASWELL-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50] +; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01] +; HASWELL-NEXT: vfnmadd213ps %xmm2, %xmm1, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [5:0.50] ; HASWELL-NEXT: vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: v4f32_one_step_2_divs: ; HASWELL-NO-FMA: # BB#0: ; HASWELL-NO-FMA-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] ; HASWELL-NO-FMA-NEXT: vmulps %xmm1, %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NO-FMA-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50] +; HASWELL-NO-FMA-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01] ; HASWELL-NO-FMA-NEXT: vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00] ; HASWELL-NO-FMA-NEXT: vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50] +; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50] -; HASWELL-NO-FMA-NEXT: retq # sched: [1:1.00] +; HASWELL-NO-FMA-NEXT: retq # sched: [2:1.00] ; ; KNL-LABEL: v4f32_one_step_2_divs: ; KNL: # BB#0: ; KNL-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] -; KNL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50] -; KNL-NEXT: vfnmadd213ps %xmm2, %xmm1, %xmm0 -; KNL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 -; KNL-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50] +; KNL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01] +; KNL-NEXT: vfnmadd213ps %xmm2, %xmm1, %xmm0 # sched: [5:0.50] +; KNL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50] +; KNL-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [5:0.50] ; KNL-NEXT: vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50] -; KNL-NEXT: retq # sched: [1:1.00] +; KNL-NEXT: retq # sched: [2:1.00] ; ; SKX-LABEL: v4f32_one_step_2_divs: ; SKX: # BB#0: ; SKX-NEXT: vrcp14ps %xmm0, %xmm1 ; SKX-NEXT: vfnmadd213ps {{.*}}(%rip){1to4}, %xmm1, %xmm0 -; SKX-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 -; SKX-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [9:0.50] +; SKX-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm0 # sched: [5:0.50] +; SKX-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm1 # sched: [5:0.50] ; SKX-NEXT: vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50] -; SKX-NEXT: retq # sched: [1:1.00] +; SKX-NEXT: retq # sched: [2:1.00] %div = fdiv fast <4 x float> <float 1.0, float 2.0, float 3.0, float 4.0>, %x %div2 = fdiv fast <4 x float> %div, %x ret <4 x float> %div2 @@ -619,9 +619,9 @@ define <4 x float> @v4f32_two_step2(<4 x float> %x) #2 { ; ; SANDY-LABEL: v4f32_two_step2: ; SANDY: # BB#0: -; SANDY-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] +; SANDY-NEXT: vrcpps %xmm0, %xmm1 # sched: [7:3.00] ; SANDY-NEXT: vmulps %xmm1, %xmm0, %xmm2 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} xmm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} xmm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [6:0.50] ; SANDY-NEXT: vsubps %xmm2, %xmm3, %xmm2 # sched: [3:1.00] ; SANDY-NEXT: vmulps %xmm2, %xmm1, %xmm2 # sched: [5:1.00] ; SANDY-NEXT: vaddps %xmm2, %xmm1, %xmm1 # sched: [3:1.00] @@ -629,26 +629,26 @@ define <4 x float> @v4f32_two_step2(<4 x float> %x) #2 { ; SANDY-NEXT: vsubps %xmm0, %xmm3, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00] ; SANDY-NEXT: vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [11:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: v4f32_two_step2: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] -; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50] +; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01] ; HASWELL-NEXT: vmovaps %xmm1, %xmm3 # sched: [1:1.00] -; HASWELL-NEXT: vfnmadd213ps %xmm2, %xmm0, %xmm3 -; HASWELL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm3 -; HASWELL-NEXT: vfnmadd213ps %xmm2, %xmm3, %xmm0 -; HASWELL-NEXT: vfmadd132ps %xmm3, %xmm3, %xmm0 -; HASWELL-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vfnmadd213ps %xmm2, %xmm0, %xmm3 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm3 # sched: [5:0.50] +; HASWELL-NEXT: vfnmadd213ps %xmm2, %xmm3, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ps %xmm3, %xmm3, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: v4f32_two_step2: ; HASWELL-NO-FMA: # BB#0: ; HASWELL-NO-FMA-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] ; HASWELL-NO-FMA-NEXT: vmulps %xmm1, %xmm0, %xmm2 # sched: [5:0.50] -; HASWELL-NO-FMA-NEXT: vbroadcastss {{.*}}(%rip), %xmm3 # sched: [4:0.50] +; HASWELL-NO-FMA-NEXT: vbroadcastss {{.*}}(%rip), %xmm3 # sched: [?:5.000000e-01] ; HASWELL-NO-FMA-NEXT: vsubps %xmm2, %xmm3, %xmm2 # sched: [3:1.00] ; HASWELL-NO-FMA-NEXT: vmulps %xmm2, %xmm1, %xmm2 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vaddps %xmm2, %xmm1, %xmm1 # sched: [3:1.00] @@ -656,32 +656,32 @@ define <4 x float> @v4f32_two_step2(<4 x float> %x) #2 { ; HASWELL-NO-FMA-NEXT: vsubps %xmm0, %xmm3, %xmm0 # sched: [3:1.00] ; HASWELL-NO-FMA-NEXT: vmulps %xmm0, %xmm1, %xmm0 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; HASWELL-NO-FMA-NEXT: retq # sched: [1:1.00] +; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NO-FMA-NEXT: retq # sched: [2:1.00] ; ; KNL-LABEL: v4f32_two_step2: ; KNL: # BB#0: ; KNL-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] -; KNL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50] +; KNL-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01] ; KNL-NEXT: vmovaps %xmm1, %xmm3 # sched: [1:1.00] -; KNL-NEXT: vfnmadd213ps %xmm2, %xmm0, %xmm3 -; KNL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm3 -; KNL-NEXT: vfnmadd213ps %xmm2, %xmm3, %xmm0 -; KNL-NEXT: vfmadd132ps %xmm3, %xmm3, %xmm0 -; KNL-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; KNL-NEXT: retq # sched: [1:1.00] +; KNL-NEXT: vfnmadd213ps %xmm2, %xmm0, %xmm3 # sched: [5:0.50] +; KNL-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm3 # sched: [5:0.50] +; KNL-NEXT: vfnmadd213ps %xmm2, %xmm3, %xmm0 # sched: [5:0.50] +; KNL-NEXT: vfmadd132ps %xmm3, %xmm3, %xmm0 # sched: [5:0.50] +; KNL-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; KNL-NEXT: retq # sched: [2:1.00] ; ; SKX-LABEL: v4f32_two_step2: ; SKX: # BB#0: ; SKX-NEXT: vrcp14ps %xmm0, %xmm1 -; SKX-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [4:0.50] +; SKX-NEXT: vbroadcastss {{.*}}(%rip), %xmm2 # sched: [?:5.000000e-01] ; SKX-NEXT: vmovaps %xmm1, %xmm3 # sched: [1:1.00] -; SKX-NEXT: vfnmadd213ps %xmm2, %xmm0, %xmm3 -; SKX-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm3 -; SKX-NEXT: vfnmadd213ps %xmm2, %xmm3, %xmm0 -; SKX-NEXT: vfmadd132ps %xmm3, %xmm3, %xmm0 -; SKX-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [9:0.50] -; SKX-NEXT: retq # sched: [1:1.00] +; SKX-NEXT: vfnmadd213ps %xmm2, %xmm0, %xmm3 # sched: [5:0.50] +; SKX-NEXT: vfmadd132ps %xmm1, %xmm1, %xmm3 # sched: [5:0.50] +; SKX-NEXT: vfnmadd213ps %xmm2, %xmm3, %xmm0 # sched: [5:0.50] +; SKX-NEXT: vfmadd132ps %xmm3, %xmm3, %xmm0 # sched: [5:0.50] +; SKX-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 # sched: [5:0.50] +; SKX-NEXT: retq # sched: [2:1.00] %div = fdiv fast <4 x float> <float 1.0, float 2.0, float 3.0, float 4.0>, %x ret <4 x float> %div } @@ -741,49 +741,49 @@ define <8 x float> @v8f32_one_step2(<8 x float> %x) #1 { ; SANDY: # BB#0: ; SANDY-NEXT: vrcpps %ymm0, %ymm1 # sched: [5:1.00] ; SANDY-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50] ; SANDY-NEXT: vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00] ; SANDY-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00] ; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:2.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: v8f32_one_step2: ; HASWELL: # BB#0: -; HASWELL-NEXT: vrcpps %ymm0, %ymm1 # sched: [7:2.00] +; HASWELL-NEXT: vrcpps %ymm0, %ymm1 # sched: [11:2.00] ; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00] -; HASWELL-NEXT: vfnmadd213ps %ymm2, %ymm1, %ymm0 -; HASWELL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 -; HASWELL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vfnmadd213ps %ymm2, %ymm1, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: v8f32_one_step2: ; HASWELL-NO-FMA: # BB#0: -; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm1 # sched: [7:2.00] -; HASWELL-NO-FMA-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00] +; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm1 # sched: [11:2.00] +; HASWELL-NO-FMA-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00] ; HASWELL-NO-FMA-NEXT: vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00] -; HASWELL-NO-FMA-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00] +; HASWELL-NO-FMA-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00] -; HASWELL-NO-FMA-NEXT: retq # sched: [1:1.00] +; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50] +; HASWELL-NO-FMA-NEXT: retq # sched: [2:1.00] ; ; KNL-LABEL: v8f32_one_step2: ; KNL: # BB#0: -; KNL-NEXT: vrcpps %ymm0, %ymm1 # sched: [7:2.00] +; KNL-NEXT: vrcpps %ymm0, %ymm1 # sched: [11:2.00] ; KNL-NEXT: vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00] -; KNL-NEXT: vfnmadd213ps %ymm2, %ymm1, %ymm0 -; KNL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 -; KNL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00] -; KNL-NEXT: retq # sched: [1:1.00] +; KNL-NEXT: vfnmadd213ps %ymm2, %ymm1, %ymm0 # sched: [5:0.50] +; KNL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50] +; KNL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50] +; KNL-NEXT: retq # sched: [2:1.00] ; ; SKX-LABEL: v8f32_one_step2: ; SKX: # BB#0: ; SKX-NEXT: vrcp14ps %ymm0, %ymm1 ; SKX-NEXT: vfnmadd213ps {{.*}}(%rip){1to8}, %ymm1, %ymm0 -; SKX-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 -; SKX-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00] -; SKX-NEXT: retq # sched: [1:1.00] +; SKX-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50] +; SKX-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50] +; SKX-NEXT: retq # sched: [2:1.00] %div = fdiv fast <8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, %x ret <8 x float> %div } @@ -848,54 +848,54 @@ define <8 x float> @v8f32_one_step_2_divs(<8 x float> %x) #1 { ; SANDY: # BB#0: ; SANDY-NEXT: vrcpps %ymm0, %ymm1 # sched: [5:1.00] ; SANDY-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50] ; SANDY-NEXT: vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00] ; SANDY-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00] ; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [9:1.00] +; SANDY-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [12:2.00] ; SANDY-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: v8f32_one_step_2_divs: ; HASWELL: # BB#0: -; HASWELL-NEXT: vrcpps %ymm0, %ymm1 # sched: [7:2.00] +; HASWELL-NEXT: vrcpps %ymm0, %ymm1 # sched: [11:2.00] ; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00] -; HASWELL-NEXT: vfnmadd213ps %ymm2, %ymm1, %ymm0 -; HASWELL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 -; HASWELL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [9:1.00] -; HASWELL-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vfnmadd213ps %ymm2, %ymm1, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [5:0.50] +; HASWELL-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: v8f32_one_step_2_divs: ; HASWELL-NO-FMA: # BB#0: -; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm1 # sched: [7:2.00] -; HASWELL-NO-FMA-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00] +; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm1 # sched: [11:2.00] +; HASWELL-NO-FMA-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00] ; HASWELL-NO-FMA-NEXT: vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00] -; HASWELL-NO-FMA-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00] +; HASWELL-NO-FMA-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [9:1.00] -; HASWELL-NO-FMA-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00] -; HASWELL-NO-FMA-NEXT: retq # sched: [1:1.00] +; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [5:0.50] +; HASWELL-NO-FMA-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50] +; HASWELL-NO-FMA-NEXT: retq # sched: [2:1.00] ; ; KNL-LABEL: v8f32_one_step_2_divs: ; KNL: # BB#0: -; KNL-NEXT: vrcpps %ymm0, %ymm1 # sched: [7:2.00] +; KNL-NEXT: vrcpps %ymm0, %ymm1 # sched: [11:2.00] ; KNL-NEXT: vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00] -; KNL-NEXT: vfnmadd213ps %ymm2, %ymm1, %ymm0 -; KNL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 -; KNL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [9:1.00] -; KNL-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00] -; KNL-NEXT: retq # sched: [1:1.00] +; KNL-NEXT: vfnmadd213ps %ymm2, %ymm1, %ymm0 # sched: [5:0.50] +; KNL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50] +; KNL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [5:0.50] +; KNL-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50] +; KNL-NEXT: retq # sched: [2:1.00] ; ; SKX-LABEL: v8f32_one_step_2_divs: ; SKX: # BB#0: ; SKX-NEXT: vrcp14ps %ymm0, %ymm1 ; SKX-NEXT: vfnmadd213ps {{.*}}(%rip){1to8}, %ymm1, %ymm0 -; SKX-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 -; SKX-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [9:1.00] -; SKX-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00] -; SKX-NEXT: retq # sched: [1:1.00] +; SKX-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm0 # sched: [5:0.50] +; SKX-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm1 # sched: [5:0.50] +; SKX-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50] +; SKX-NEXT: retq # sched: [2:1.00] %div = fdiv fast <8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, %x %div2 = fdiv fast <8 x float> %div, %x ret <8 x float> %div2 @@ -980,7 +980,7 @@ define <8 x float> @v8f32_two_step2(<8 x float> %x) #2 { ; SANDY: # BB#0: ; SANDY-NEXT: vrcpps %ymm0, %ymm1 # sched: [5:1.00] ; SANDY-NEXT: vmulps %ymm1, %ymm0, %ymm2 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [4:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50] ; SANDY-NEXT: vsubps %ymm2, %ymm3, %ymm2 # sched: [3:1.00] ; SANDY-NEXT: vmulps %ymm2, %ymm1, %ymm2 # sched: [5:1.00] ; SANDY-NEXT: vaddps %ymm2, %ymm1, %ymm1 # sched: [3:1.00] @@ -988,59 +988,59 @@ define <8 x float> @v8f32_two_step2(<8 x float> %x) #2 { ; SANDY-NEXT: vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00] ; SANDY-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00] ; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; SANDY-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:2.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: v8f32_two_step2: ; HASWELL: # BB#0: -; HASWELL-NEXT: vrcpps %ymm0, %ymm1 # sched: [7:2.00] +; HASWELL-NEXT: vrcpps %ymm0, %ymm1 # sched: [11:2.00] ; HASWELL-NEXT: vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00] ; HASWELL-NEXT: vmovaps %ymm1, %ymm3 # sched: [1:1.00] -; HASWELL-NEXT: vfnmadd213ps %ymm2, %ymm0, %ymm3 -; HASWELL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm3 -; HASWELL-NEXT: vfnmadd213ps %ymm2, %ymm3, %ymm0 -; HASWELL-NEXT: vfmadd132ps %ymm3, %ymm3, %ymm0 -; HASWELL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vfnmadd213ps %ymm2, %ymm0, %ymm3 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm3 # sched: [5:0.50] +; HASWELL-NEXT: vfnmadd213ps %ymm2, %ymm3, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: vfmadd132ps %ymm3, %ymm3, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: v8f32_two_step2: ; HASWELL-NO-FMA: # BB#0: -; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm1 # sched: [7:2.00] -; HASWELL-NO-FMA-NEXT: vmulps %ymm1, %ymm0, %ymm2 # sched: [5:1.00] +; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm1 # sched: [11:2.00] +; HASWELL-NO-FMA-NEXT: vmulps %ymm1, %ymm0, %ymm2 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vbroadcastss {{.*}}(%rip), %ymm3 # sched: [5:1.00] ; HASWELL-NO-FMA-NEXT: vsubps %ymm2, %ymm3, %ymm2 # sched: [3:1.00] -; HASWELL-NO-FMA-NEXT: vmulps %ymm2, %ymm1, %ymm2 # sched: [5:1.00] +; HASWELL-NO-FMA-NEXT: vmulps %ymm2, %ymm1, %ymm2 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vaddps %ymm2, %ymm1, %ymm1 # sched: [3:1.00] -; HASWELL-NO-FMA-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00] +; HASWELL-NO-FMA-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00] -; HASWELL-NO-FMA-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00] +; HASWELL-NO-FMA-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:0.50] ; HASWELL-NO-FMA-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00] -; HASWELL-NO-FMA-NEXT: retq # sched: [1:1.00] +; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50] +; HASWELL-NO-FMA-NEXT: retq # sched: [2:1.00] ; ; KNL-LABEL: v8f32_two_step2: ; KNL: # BB#0: -; KNL-NEXT: vrcpps %ymm0, %ymm1 # sched: [7:2.00] +; KNL-NEXT: vrcpps %ymm0, %ymm1 # sched: [11:2.00] ; KNL-NEXT: vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00] ; KNL-NEXT: vmovaps %ymm1, %ymm3 # sched: [1:1.00] -; KNL-NEXT: vfnmadd213ps %ymm2, %ymm0, %ymm3 -; KNL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm3 -; KNL-NEXT: vfnmadd213ps %ymm2, %ymm3, %ymm0 -; KNL-NEXT: vfmadd132ps %ymm3, %ymm3, %ymm0 -; KNL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00] -; KNL-NEXT: retq # sched: [1:1.00] +; KNL-NEXT: vfnmadd213ps %ymm2, %ymm0, %ymm3 # sched: [5:0.50] +; KNL-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm3 # sched: [5:0.50] +; KNL-NEXT: vfnmadd213ps %ymm2, %ymm3, %ymm0 # sched: [5:0.50] +; KNL-NEXT: vfmadd132ps %ymm3, %ymm3, %ymm0 # sched: [5:0.50] +; KNL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50] +; KNL-NEXT: retq # sched: [2:1.00] ; ; SKX-LABEL: v8f32_two_step2: ; SKX: # BB#0: ; SKX-NEXT: vrcp14ps %ymm0, %ymm1 ; SKX-NEXT: vbroadcastss {{.*}}(%rip), %ymm2 # sched: [5:1.00] ; SKX-NEXT: vmovaps %ymm1, %ymm3 # sched: [1:1.00] -; SKX-NEXT: vfnmadd213ps %ymm2, %ymm0, %ymm3 -; SKX-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm3 -; SKX-NEXT: vfnmadd213ps %ymm2, %ymm3, %ymm0 -; SKX-NEXT: vfmadd132ps %ymm3, %ymm3, %ymm0 -; SKX-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00] -; SKX-NEXT: retq # sched: [1:1.00] +; SKX-NEXT: vfnmadd213ps %ymm2, %ymm0, %ymm3 # sched: [5:0.50] +; SKX-NEXT: vfmadd132ps %ymm1, %ymm1, %ymm3 # sched: [5:0.50] +; SKX-NEXT: vfnmadd213ps %ymm2, %ymm3, %ymm0 # sched: [5:0.50] +; SKX-NEXT: vfmadd132ps %ymm3, %ymm3, %ymm0 # sched: [5:0.50] +; SKX-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50] +; SKX-NEXT: retq # sched: [2:1.00] %div = fdiv fast <8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, %x ret <8 x float> %div } @@ -1070,27 +1070,27 @@ define <8 x float> @v8f32_no_step(<8 x float> %x) #3 { ; SANDY-LABEL: v8f32_no_step: ; SANDY: # BB#0: ; SANDY-NEXT: vrcpps %ymm0, %ymm0 # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: v8f32_no_step: ; HASWELL: # BB#0: -; HASWELL-NEXT: vrcpps %ymm0, %ymm0 # sched: [7:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vrcpps %ymm0, %ymm0 # sched: [11:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: v8f32_no_step: ; HASWELL-NO-FMA: # BB#0: -; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm0 # sched: [7:2.00] -; HASWELL-NO-FMA-NEXT: retq # sched: [1:1.00] +; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm0 # sched: [11:2.00] +; HASWELL-NO-FMA-NEXT: retq # sched: [2:1.00] ; ; KNL-LABEL: v8f32_no_step: ; KNL: # BB#0: -; KNL-NEXT: vrcpps %ymm0, %ymm0 # sched: [7:2.00] -; KNL-NEXT: retq # sched: [1:1.00] +; KNL-NEXT: vrcpps %ymm0, %ymm0 # sched: [11:2.00] +; KNL-NEXT: retq # sched: [2:1.00] ; ; SKX-LABEL: v8f32_no_step: ; SKX: # BB#0: ; SKX-NEXT: vrcp14ps %ymm0, %ymm0 -; SKX-NEXT: retq # sched: [1:1.00] +; SKX-NEXT: retq # sched: [2:1.00] %div = fdiv fast <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %x ret <8 x float> %div } @@ -1125,32 +1125,32 @@ define <8 x float> @v8f32_no_step2(<8 x float> %x) #3 { ; SANDY-LABEL: v8f32_no_step2: ; SANDY: # BB#0: ; SANDY-NEXT: vrcpps %ymm0, %ymm0 # sched: [5:1.00] -; SANDY-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:2.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: v8f32_no_step2: ; HASWELL: # BB#0: -; HASWELL-NEXT: vrcpps %ymm0, %ymm0 # sched: [7:2.00] -; HASWELL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vrcpps %ymm0, %ymm0 # sched: [11:2.00] +; HASWELL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; HASWELL-NO-FMA-LABEL: v8f32_no_step2: ; HASWELL-NO-FMA: # BB#0: -; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm0 # sched: [7:2.00] -; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00] -; HASWELL-NO-FMA-NEXT: retq # sched: [1:1.00] +; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm0 # sched: [11:2.00] +; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50] +; HASWELL-NO-FMA-NEXT: retq # sched: [2:1.00] ; ; KNL-LABEL: v8f32_no_step2: ; KNL: # BB#0: -; KNL-NEXT: vrcpps %ymm0, %ymm0 # sched: [7:2.00] -; KNL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00] -; KNL-NEXT: retq # sched: [1:1.00] +; KNL-NEXT: vrcpps %ymm0, %ymm0 # sched: [11:2.00] +; KNL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50] +; KNL-NEXT: retq # sched: [2:1.00] ; ; SKX-LABEL: v8f32_no_step2: ; SKX: # BB#0: ; SKX-NEXT: vrcp14ps %ymm0, %ymm0 -; SKX-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [9:1.00] -; SKX-NEXT: retq # sched: [1:1.00] +; SKX-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [5:0.50] +; SKX-NEXT: retq # sched: [2:1.00] %div = fdiv fast <8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, %x ret <8 x float> %div } diff --git a/llvm/test/CodeGen/X86/sse-schedule.ll b/llvm/test/CodeGen/X86/sse-schedule.ll index 52e6b61aedf..e94935972c1 100644 --- a/llvm/test/CodeGen/X86/sse-schedule.ll +++ b/llvm/test/CodeGen/X86/sse-schedule.ll @@ -31,14 +31,14 @@ define <4 x float> @test_addps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a ; SANDY-LABEL: test_addps: ; SANDY: # BB#0: ; SANDY-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vaddps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vaddps (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_addps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vaddps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vaddps (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_addps: ; BTVER2: # BB#0: @@ -73,14 +73,14 @@ define float @test_addss(float %a0, float %a1, float *%a2) { ; SANDY-LABEL: test_addss: ; SANDY: # BB#0: ; SANDY-NEXT: vaddss %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vaddss (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vaddss (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_addss: ; HASWELL: # BB#0: ; HASWELL-NEXT: vaddss %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vaddss (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vaddss (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_addss: ; BTVER2: # BB#0: @@ -122,15 +122,15 @@ define <4 x float> @test_andps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a ; ; SANDY-LABEL: test_andps: ; SANDY: # BB#0: -; SANDY-NEXT: vandps %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: vandps (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vandps %xmm1, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: vandps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_andps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vandps %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vandps (%rdi), %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vandps (%rdi), %xmm0, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_andps: ; BTVER2: # BB#0: @@ -176,15 +176,15 @@ define <4 x float> @test_andnotps(<4 x float> %a0, <4 x float> %a1, <4 x float> ; ; SANDY-LABEL: test_andnotps: ; SANDY: # BB#0: -; SANDY-NEXT: vandnps %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: vandnps (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vandnps %xmm1, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: vandnps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_andnotps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vandnps %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vandnps (%rdi), %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vandnps (%rdi), %xmm0, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_andnotps: ; BTVER2: # BB#0: @@ -228,16 +228,16 @@ define <4 x float> @test_cmpps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a ; SANDY-LABEL: test_cmpps: ; SANDY: # BB#0: ; SANDY-NEXT: vcmpeqps %xmm1, %xmm0, %xmm1 # sched: [3:1.00] -; SANDY-NEXT: vcmpeqps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: vorps %xmm0, %xmm1, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vcmpeqps (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: vorps %xmm0, %xmm1, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cmpps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcmpeqps %xmm1, %xmm0, %xmm1 # sched: [3:1.00] -; HASWELL-NEXT: vcmpeqps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] +; HASWELL-NEXT: vcmpeqps (%rdi), %xmm0, %xmm0 # sched: [3:1.00] ; HASWELL-NEXT: vorps %xmm0, %xmm1, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cmpps: ; BTVER2: # BB#0: @@ -277,13 +277,13 @@ define float @test_cmpss(float %a0, float %a1, float *%a2) { ; SANDY: # BB#0: ; SANDY-NEXT: vcmpeqss %xmm1, %xmm0, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vcmpeqss (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cmpss: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcmpeqss %xmm1, %xmm0, %xmm0 # sched: [3:1.00] ; HASWELL-NEXT: vcmpeqss (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cmpss: ; BTVER2: # BB#0: @@ -347,30 +347,30 @@ define i32 @test_comiss(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) { ; SANDY-LABEL: test_comiss: ; SANDY: # BB#0: ; SANDY-NEXT: vcomiss %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: setnp %al # sched: [1:0.33] -; SANDY-NEXT: sete %cl # sched: [1:0.33] +; SANDY-NEXT: setnp %al # sched: [1:1.00] +; SANDY-NEXT: sete %cl # sched: [1:1.00] ; SANDY-NEXT: andb %al, %cl # sched: [1:0.33] ; SANDY-NEXT: vcomiss (%rdi), %xmm0 # sched: [7:1.00] -; SANDY-NEXT: setnp %al # sched: [1:0.33] -; SANDY-NEXT: sete %dl # sched: [1:0.33] +; SANDY-NEXT: setnp %al # sched: [1:1.00] +; SANDY-NEXT: sete %dl # sched: [1:1.00] ; SANDY-NEXT: andb %al, %dl # sched: [1:0.33] ; SANDY-NEXT: orb %cl, %dl # sched: [1:0.33] ; SANDY-NEXT: movzbl %dl, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_comiss: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcomiss %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: setnp %al # sched: [1:0.50] -; HASWELL-NEXT: sete %cl # sched: [1:0.50] +; HASWELL-NEXT: setnp %al # sched: [1:1.00] +; HASWELL-NEXT: sete %cl # sched: [1:1.00] ; HASWELL-NEXT: andb %al, %cl # sched: [1:0.25] ; HASWELL-NEXT: vcomiss (%rdi), %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: setnp %al # sched: [1:0.50] -; HASWELL-NEXT: sete %dl # sched: [1:0.50] +; HASWELL-NEXT: setnp %al # sched: [1:1.00] +; HASWELL-NEXT: sete %dl # sched: [1:1.00] ; HASWELL-NEXT: andb %al, %dl # sched: [1:0.25] ; HASWELL-NEXT: orb %cl, %dl # sched: [1:0.25] ; HASWELL-NEXT: movzbl %dl, %eax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_comiss: ; BTVER2: # BB#0: @@ -417,17 +417,17 @@ define float @test_cvtsi2ss(i32 %a0, i32 *%a1) { ; ; SANDY-LABEL: test_cvtsi2ss: ; SANDY: # BB#0: -; SANDY-NEXT: vcvtsi2ssl %edi, %xmm0, %xmm0 # sched: [4:1.00] -; SANDY-NEXT: vcvtsi2ssl (%rsi), %xmm1, %xmm1 # sched: [8:1.00] +; SANDY-NEXT: vcvtsi2ssl %edi, %xmm0, %xmm0 # sched: [5:2.00] +; SANDY-NEXT: vcvtsi2ssl (%rsi), %xmm1, %xmm1 # sched: [10:1.00] ; SANDY-NEXT: vaddss %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtsi2ss: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvtsi2ssl %edi, %xmm0, %xmm0 # sched: [4:1.00] ; HASWELL-NEXT: vcvtsi2ssl (%rsi), %xmm1, %xmm1 # sched: [8:1.00] ; HASWELL-NEXT: vaddss %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtsi2ss: ; BTVER2: # BB#0: @@ -466,17 +466,17 @@ define float @test_cvtsi2ssq(i64 %a0, i64 *%a1) { ; ; SANDY-LABEL: test_cvtsi2ssq: ; SANDY: # BB#0: -; SANDY-NEXT: vcvtsi2ssq %rdi, %xmm0, %xmm0 # sched: [4:1.00] -; SANDY-NEXT: vcvtsi2ssq (%rsi), %xmm1, %xmm1 # sched: [8:1.00] +; SANDY-NEXT: vcvtsi2ssq %rdi, %xmm0, %xmm0 # sched: [5:2.00] +; SANDY-NEXT: vcvtsi2ssq (%rsi), %xmm1, %xmm1 # sched: [10:1.00] ; SANDY-NEXT: vaddss %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtsi2ssq: ; HASWELL: # BB#0: -; HASWELL-NEXT: vcvtsi2ssq %rdi, %xmm0, %xmm0 # sched: [4:1.00] +; HASWELL-NEXT: vcvtsi2ssq %rdi, %xmm0, %xmm0 # sched: [5:2.00] ; HASWELL-NEXT: vcvtsi2ssq (%rsi), %xmm1, %xmm1 # sched: [8:1.00] ; HASWELL-NEXT: vaddss %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtsi2ssq: ; BTVER2: # BB#0: @@ -515,17 +515,17 @@ define i32 @test_cvtss2si(float %a0, float *%a1) { ; ; SANDY-LABEL: test_cvtss2si: ; SANDY: # BB#0: -; SANDY-NEXT: vcvtss2si %xmm0, %ecx # sched: [3:1.00] -; SANDY-NEXT: vcvtss2si (%rdi), %eax # sched: [7:1.00] +; SANDY-NEXT: vcvtss2si %xmm0, %ecx # sched: [5:1.00] +; SANDY-NEXT: vcvtss2si (%rdi), %eax # sched: [10:1.00] ; SANDY-NEXT: addl %ecx, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtss2si: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvtss2si %xmm0, %ecx # sched: [4:1.00] -; HASWELL-NEXT: vcvtss2si (%rdi), %eax # sched: [8:1.00] +; HASWELL-NEXT: vcvtss2si (%rdi), %eax # sched: [4:1.00] ; HASWELL-NEXT: addl %ecx, %eax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtss2si: ; BTVER2: # BB#0: @@ -567,17 +567,17 @@ define i64 @test_cvtss2siq(float %a0, float *%a1) { ; ; SANDY-LABEL: test_cvtss2siq: ; SANDY: # BB#0: -; SANDY-NEXT: vcvtss2si %xmm0, %rcx # sched: [3:1.00] -; SANDY-NEXT: vcvtss2si (%rdi), %rax # sched: [7:1.00] +; SANDY-NEXT: vcvtss2si %xmm0, %rcx # sched: [5:1.00] +; SANDY-NEXT: vcvtss2si (%rdi), %rax # sched: [10:1.00] ; SANDY-NEXT: addq %rcx, %rax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtss2siq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvtss2si %xmm0, %rcx # sched: [4:1.00] -; HASWELL-NEXT: vcvtss2si (%rdi), %rax # sched: [8:1.00] +; HASWELL-NEXT: vcvtss2si (%rdi), %rax # sched: [4:1.00] ; HASWELL-NEXT: addq %rcx, %rax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtss2siq: ; BTVER2: # BB#0: @@ -619,17 +619,17 @@ define i32 @test_cvttss2si(float %a0, float *%a1) { ; ; SANDY-LABEL: test_cvttss2si: ; SANDY: # BB#0: -; SANDY-NEXT: vcvttss2si %xmm0, %ecx # sched: [3:1.00] -; SANDY-NEXT: vcvttss2si (%rdi), %eax # sched: [7:1.00] +; SANDY-NEXT: vcvttss2si %xmm0, %ecx # sched: [5:1.00] +; SANDY-NEXT: vcvttss2si (%rdi), %eax # sched: [10:1.00] ; SANDY-NEXT: addl %ecx, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvttss2si: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvttss2si %xmm0, %ecx # sched: [4:1.00] -; HASWELL-NEXT: vcvttss2si (%rdi), %eax # sched: [8:1.00] +; HASWELL-NEXT: vcvttss2si (%rdi), %eax # sched: [4:1.00] ; HASWELL-NEXT: addl %ecx, %eax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvttss2si: ; BTVER2: # BB#0: @@ -668,17 +668,17 @@ define i64 @test_cvttss2siq(float %a0, float *%a1) { ; ; SANDY-LABEL: test_cvttss2siq: ; SANDY: # BB#0: -; SANDY-NEXT: vcvttss2si %xmm0, %rcx # sched: [3:1.00] -; SANDY-NEXT: vcvttss2si (%rdi), %rax # sched: [7:1.00] +; SANDY-NEXT: vcvttss2si %xmm0, %rcx # sched: [5:1.00] +; SANDY-NEXT: vcvttss2si (%rdi), %rax # sched: [10:1.00] ; SANDY-NEXT: addq %rcx, %rax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvttss2siq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvttss2si %xmm0, %rcx # sched: [4:1.00] -; HASWELL-NEXT: vcvttss2si (%rdi), %rax # sched: [8:1.00] +; HASWELL-NEXT: vcvttss2si (%rdi), %rax # sched: [4:1.00] ; HASWELL-NEXT: addq %rcx, %rax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvttss2siq: ; BTVER2: # BB#0: @@ -714,15 +714,15 @@ define <4 x float> @test_divps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a ; ; SANDY-LABEL: test_divps: ; SANDY: # BB#0: -; SANDY-NEXT: vdivps %xmm1, %xmm0, %xmm0 # sched: [12:1.00] -; SANDY-NEXT: vdivps (%rdi), %xmm0, %xmm0 # sched: [16:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vdivps %xmm1, %xmm0, %xmm0 # sched: [14:1.00] +; SANDY-NEXT: vdivps (%rdi), %xmm0, %xmm0 # sched: [20:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_divps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vdivps %xmm1, %xmm0, %xmm0 # sched: [12:1.00] -; HASWELL-NEXT: vdivps (%rdi), %xmm0, %xmm0 # sched: [16:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vdivps %xmm1, %xmm0, %xmm0 # sched: [13:1.00] +; HASWELL-NEXT: vdivps (%rdi), %xmm0, %xmm0 # sched: [13:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_divps: ; BTVER2: # BB#0: @@ -756,15 +756,15 @@ define float @test_divss(float %a0, float %a1, float *%a2) { ; ; SANDY-LABEL: test_divss: ; SANDY: # BB#0: -; SANDY-NEXT: vdivss %xmm1, %xmm0, %xmm0 # sched: [12:1.00] -; SANDY-NEXT: vdivss (%rdi), %xmm0, %xmm0 # sched: [16:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vdivss %xmm1, %xmm0, %xmm0 # sched: [14:1.00] +; SANDY-NEXT: vdivss (%rdi), %xmm0, %xmm0 # sched: [20:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_divss: ; HASWELL: # BB#0: -; HASWELL-NEXT: vdivss %xmm1, %xmm0, %xmm0 # sched: [12:1.00] -; HASWELL-NEXT: vdivss (%rdi), %xmm0, %xmm0 # sched: [16:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vdivss %xmm1, %xmm0, %xmm0 # sched: [13:1.00] +; HASWELL-NEXT: vdivss (%rdi), %xmm0, %xmm0 # sched: [13:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_divss: ; BTVER2: # BB#0: @@ -799,14 +799,14 @@ define void @test_ldmxcsr(i32 %a0) { ; SANDY-LABEL: test_ldmxcsr: ; SANDY: # BB#0: ; SANDY-NEXT: movl %edi, -{{[0-9]+}}(%rsp) # sched: [1:1.00] -; SANDY-NEXT: vldmxcsr -{{[0-9]+}}(%rsp) # sched: [4:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vldmxcsr -{{[0-9]+}}(%rsp) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_ldmxcsr: ; HASWELL: # BB#0: ; HASWELL-NEXT: movl %edi, -{{[0-9]+}}(%rsp) # sched: [1:1.00] -; HASWELL-NEXT: vldmxcsr -{{[0-9]+}}(%rsp) # sched: [6:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vldmxcsr -{{[0-9]+}}(%rsp) # sched: [2:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_ldmxcsr: ; BTVER2: # BB#0: @@ -843,14 +843,14 @@ define <4 x float> @test_maxps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a ; SANDY-LABEL: test_maxps: ; SANDY: # BB#0: ; SANDY-NEXT: vmaxps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmaxps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmaxps (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_maxps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmaxps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vmaxps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmaxps (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_maxps: ; BTVER2: # BB#0: @@ -886,14 +886,14 @@ define <4 x float> @test_maxss(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a ; SANDY-LABEL: test_maxss: ; SANDY: # BB#0: ; SANDY-NEXT: vmaxss %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmaxss (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmaxss (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_maxss: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmaxss %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vmaxss (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmaxss (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_maxss: ; BTVER2: # BB#0: @@ -929,14 +929,14 @@ define <4 x float> @test_minps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a ; SANDY-LABEL: test_minps: ; SANDY: # BB#0: ; SANDY-NEXT: vminps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vminps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vminps (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_minps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vminps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vminps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vminps (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_minps: ; BTVER2: # BB#0: @@ -972,14 +972,14 @@ define <4 x float> @test_minss(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a ; SANDY-LABEL: test_minss: ; SANDY: # BB#0: ; SANDY-NEXT: vminss %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vminss (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vminss (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_minss: ; HASWELL: # BB#0: ; HASWELL-NEXT: vminss %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vminss (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vminss (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_minss: ; BTVER2: # BB#0: @@ -1017,17 +1017,17 @@ define void @test_movaps(<4 x float> *%a0, <4 x float> *%a1) { ; ; SANDY-LABEL: test_movaps: ; SANDY: # BB#0: -; SANDY-NEXT: vmovaps (%rdi), %xmm0 # sched: [4:0.50] +; SANDY-NEXT: vmovaps (%rdi), %xmm0 # sched: [6:0.50] ; SANDY-NEXT: vaddps %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmovaps %xmm0, (%rsi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovaps %xmm0, (%rsi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movaps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovaps (%rdi), %xmm0 # sched: [4:0.50] +; HASWELL-NEXT: vmovaps (%rdi), %xmm0 # sched: [?:5.000000e-01] ; HASWELL-NEXT: vaddps %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vmovaps %xmm0, (%rsi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovaps %xmm0, (%rsi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movaps: ; BTVER2: # BB#0: @@ -1068,12 +1068,12 @@ define <4 x float> @test_movhlps(<4 x float> %a0, <4 x float> %a1) { ; SANDY-LABEL: test_movhlps: ; SANDY: # BB#0: ; SANDY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1] sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movhlps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1] sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movhlps: ; BTVER2: # BB#0: @@ -1111,17 +1111,17 @@ define void @test_movhps(<4 x float> %a0, <4 x float> %a1, x86_mmx *%a2) { ; ; SANDY-LABEL: test_movhps: ; SANDY: # BB#0: -; SANDY-NEXT: vmovhpd {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [5:1.00] +; SANDY-NEXT: vmovhpd {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [7:1.00] ; SANDY-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vpextrq $1, %xmm0, (%rdi) # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movhps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovhpd {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [5:1.00] +; HASWELL-NEXT: vmovhpd {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [1:1.00] ; HASWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vpextrq $1, %xmm0, (%rdi) # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpextrq $1, %xmm0, (%rdi) # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movhps: ; BTVER2: # BB#0: @@ -1164,13 +1164,13 @@ define <4 x float> @test_movlhps(<4 x float> %a0, <4 x float> %a1) { ; SANDY: # BB#0: ; SANDY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0] sched: [1:1.00] ; SANDY-NEXT: vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movlhps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0] sched: [1:1.00] ; HASWELL-NEXT: vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movlhps: ; BTVER2: # BB#0: @@ -1206,17 +1206,17 @@ define void @test_movlps(<4 x float> %a0, <4 x float> %a1, x86_mmx *%a2) { ; ; SANDY-LABEL: test_movlps: ; SANDY: # BB#0: -; SANDY-NEXT: vmovlpd {{.*#+}} xmm1 = mem[0],xmm1[1] sched: [5:1.00] +; SANDY-NEXT: vmovlpd {{.*#+}} xmm1 = mem[0],xmm1[1] sched: [7:1.00] ; SANDY-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmovlps %xmm0, (%rdi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovlps %xmm0, (%rdi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movlps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovlpd {{.*#+}} xmm1 = mem[0],xmm1[1] sched: [5:1.00] +; HASWELL-NEXT: vmovlpd {{.*#+}} xmm1 = mem[0],xmm1[1] sched: [1:1.00] ; HASWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vmovlps %xmm0, (%rdi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovlps %xmm0, (%rdi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movlps: ; BTVER2: # BB#0: @@ -1254,13 +1254,13 @@ define i32 @test_movmskps(<4 x float> %a0) { ; ; SANDY-LABEL: test_movmskps: ; SANDY: # BB#0: -; SANDY-NEXT: vmovmskps %xmm0, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovmskps %xmm0, %eax # sched: [2:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movmskps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmovmskps %xmm0, %eax # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movmskps: ; BTVER2: # BB#0: @@ -1295,13 +1295,13 @@ define void @test_movntps(<4 x float> %a0, <4 x float> *%a1) { ; ; SANDY-LABEL: test_movntps: ; SANDY: # BB#0: -; SANDY-NEXT: vmovntps %xmm0, (%rdi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovntps %xmm0, (%rdi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movntps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovntps %xmm0, (%rdi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovntps %xmm0, (%rdi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movntps: ; BTVER2: # BB#0: @@ -1335,17 +1335,17 @@ define void @test_movss_mem(float* %a0, float* %a1) { ; ; SANDY-LABEL: test_movss_mem: ; SANDY: # BB#0: -; SANDY-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero sched: [4:0.50] +; SANDY-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero sched: [6:0.50] ; SANDY-NEXT: vaddss %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmovss %xmm0, (%rsi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovss %xmm0, (%rsi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movss_mem: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero sched: [4:0.50] +; HASWELL-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero sched: [?:5.000000e-01] ; HASWELL-NEXT: vaddss %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vmovss %xmm0, (%rsi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovss %xmm0, (%rsi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movss_mem: ; BTVER2: # BB#0: @@ -1383,13 +1383,13 @@ define <4 x float> @test_movss_reg(<4 x float> %a0, <4 x float> %a1) { ; ; SANDY-LABEL: test_movss_reg: ; SANDY: # BB#0: -; SANDY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3] sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3] sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movss_reg: ; HASWELL: # BB#0: ; HASWELL-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3] sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movss_reg: ; BTVER2: # BB#0: @@ -1423,17 +1423,17 @@ define void @test_movups(<4 x float> *%a0, <4 x float> *%a1) { ; ; SANDY-LABEL: test_movups: ; SANDY: # BB#0: -; SANDY-NEXT: vmovups (%rdi), %xmm0 # sched: [4:0.50] +; SANDY-NEXT: vmovups (%rdi), %xmm0 # sched: [6:0.50] ; SANDY-NEXT: vaddps %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmovups %xmm0, (%rsi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovups %xmm0, (%rsi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movups: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovups (%rdi), %xmm0 # sched: [4:0.50] +; HASWELL-NEXT: vmovups (%rdi), %xmm0 # sched: [?:5.000000e-01] ; HASWELL-NEXT: vaddps %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vmovups %xmm0, (%rsi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovups %xmm0, (%rsi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movups: ; BTVER2: # BB#0: @@ -1469,14 +1469,14 @@ define <4 x float> @test_mulps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a ; SANDY-LABEL: test_mulps: ; SANDY: # BB#0: ; SANDY-NEXT: vmulps %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: vmulps (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmulps (%rdi), %xmm0, %xmm0 # sched: [11:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_mulps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmulps %xmm1, %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: vmulps (%rdi), %xmm0, %xmm0 # sched: [9:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmulps (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_mulps: ; BTVER2: # BB#0: @@ -1511,14 +1511,14 @@ define float @test_mulss(float %a0, float %a1, float *%a2) { ; SANDY-LABEL: test_mulss: ; SANDY: # BB#0: ; SANDY-NEXT: vmulss %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: vmulss (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmulss (%rdi), %xmm0, %xmm0 # sched: [11:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_mulss: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmulss %xmm1, %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: vmulss (%rdi), %xmm0, %xmm0 # sched: [9:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmulss (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_mulss: ; BTVER2: # BB#0: @@ -1560,15 +1560,15 @@ define <4 x float> @test_orps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2 ; ; SANDY-LABEL: test_orps: ; SANDY: # BB#0: -; SANDY-NEXT: vorps %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: vorps (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vorps %xmm1, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: vorps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_orps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vorps %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vorps (%rdi), %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vorps (%rdi), %xmm0, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_orps: ; BTVER2: # BB#0: @@ -1609,13 +1609,13 @@ define void @test_prefetchnta(i8* %a0) { ; ; SANDY-LABEL: test_prefetchnta: ; SANDY: # BB#0: -; SANDY-NEXT: prefetchnta (%rdi) # sched: [4:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: prefetchnta (%rdi) # sched: [5:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_prefetchnta: ; HASWELL: # BB#0: -; HASWELL-NEXT: prefetchnta (%rdi) # sched: [4:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: prefetchnta (%rdi) # sched: [?:5.000000e-01] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_prefetchnta: ; BTVER2: # BB#0: @@ -1652,17 +1652,17 @@ define <4 x float> @test_rcpps(<4 x float> %a0, <4 x float> *%a1) { ; ; SANDY-LABEL: test_rcpps: ; SANDY: # BB#0: -; SANDY-NEXT: vrcpps %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: vrcpps (%rdi), %xmm1 # sched: [9:1.00] +; SANDY-NEXT: vrcpps %xmm0, %xmm0 # sched: [7:3.00] +; SANDY-NEXT: vrcpps (%rdi), %xmm1 # sched: [11:1.00] ; SANDY-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_rcpps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrcpps %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: vrcpps (%rdi), %xmm1 # sched: [9:1.00] +; HASWELL-NEXT: vrcpps (%rdi), %xmm1 # sched: [5:1.00] ; HASWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_rcpps: ; BTVER2: # BB#0: @@ -1708,18 +1708,18 @@ define <4 x float> @test_rcpss(float %a0, float *%a1) { ; SANDY-LABEL: test_rcpss: ; SANDY: # BB#0: ; SANDY-NEXT: vrcpss %xmm0, %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [4:0.50] +; SANDY-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [6:0.50] ; SANDY-NEXT: vrcpss %xmm1, %xmm1, %xmm1 # sched: [9:1.00] ; SANDY-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_rcpss: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrcpss %xmm0, %xmm0, %xmm0 # sched: [9:1.00] -; HASWELL-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [4:0.50] +; HASWELL-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [?:5.000000e-01] ; HASWELL-NEXT: vrcpss %xmm1, %xmm1, %xmm1 # sched: [9:1.00] ; HASWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_rcpss: ; BTVER2: # BB#0: @@ -1765,16 +1765,16 @@ define <4 x float> @test_rsqrtps(<4 x float> %a0, <4 x float> *%a1) { ; SANDY-LABEL: test_rsqrtps: ; SANDY: # BB#0: ; SANDY-NEXT: vrsqrtps %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: vrsqrtps (%rdi), %xmm1 # sched: [9:1.00] +; SANDY-NEXT: vrsqrtps (%rdi), %xmm1 # sched: [11:1.00] ; SANDY-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_rsqrtps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrsqrtps %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: vrsqrtps (%rdi), %xmm1 # sched: [9:1.00] +; HASWELL-NEXT: vrsqrtps (%rdi), %xmm1 # sched: [5:1.00] ; HASWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_rsqrtps: ; BTVER2: # BB#0: @@ -1819,19 +1819,19 @@ define <4 x float> @test_rsqrtss(float %a0, float *%a1) { ; ; SANDY-LABEL: test_rsqrtss: ; SANDY: # BB#0: -; SANDY-NEXT: vrsqrtss %xmm0, %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [4:0.50] -; SANDY-NEXT: vrsqrtss %xmm1, %xmm1, %xmm1 # sched: [9:1.00] +; SANDY-NEXT: vrsqrtss %xmm0, %xmm0, %xmm0 # sched: [5:1.00] +; SANDY-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [6:0.50] +; SANDY-NEXT: vrsqrtss %xmm1, %xmm1, %xmm1 # sched: [5:1.00] ; SANDY-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_rsqrtss: ; HASWELL: # BB#0: ; HASWELL-NEXT: vrsqrtss %xmm0, %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [4:0.50] +; HASWELL-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [?:5.000000e-01] ; HASWELL-NEXT: vrsqrtss %xmm1, %xmm1, %xmm1 # sched: [5:1.00] ; HASWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_rsqrtss: ; BTVER2: # BB#0: @@ -1875,12 +1875,12 @@ define void @test_sfence() { ; SANDY-LABEL: test_sfence: ; SANDY: # BB#0: ; SANDY-NEXT: sfence # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_sfence: ; HASWELL: # BB#0: -; HASWELL-NEXT: sfence # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: sfence # sched: [1:0.33] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_sfence: ; BTVER2: # BB#0: @@ -1917,14 +1917,14 @@ define <4 x float> @test_shufps(<4 x float> %a0, <4 x float> %a1, <4 x float> *% ; SANDY-LABEL: test_shufps: ; SANDY: # BB#0: ; SANDY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0] sched: [1:1.00] -; SANDY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3],mem[0,0] sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3],mem[0,0] sched: [7:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_shufps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0] sched: [1:1.00] -; HASWELL-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3],mem[0,0] sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3],mem[0,0] sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_shufps: ; BTVER2: # BB#0: @@ -1962,17 +1962,17 @@ define <4 x float> @test_sqrtps(<4 x float> %a0, <4 x float> *%a1) { ; ; SANDY-LABEL: test_sqrtps: ; SANDY: # BB#0: -; SANDY-NEXT: vsqrtps %xmm0, %xmm0 # sched: [15:1.00] -; SANDY-NEXT: vsqrtps (%rdi), %xmm1 # sched: [19:1.00] +; SANDY-NEXT: vsqrtps %xmm0, %xmm0 # sched: [14:1.00] +; SANDY-NEXT: vsqrtps (%rdi), %xmm1 # sched: [20:1.00] ; SANDY-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_sqrtps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vsqrtps %xmm0, %xmm0 # sched: [15:1.00] -; HASWELL-NEXT: vsqrtps (%rdi), %xmm1 # sched: [19:1.00] +; HASWELL-NEXT: vsqrtps %xmm0, %xmm0 # sched: [14:1.00] +; HASWELL-NEXT: vsqrtps (%rdi), %xmm1 # sched: [14:1.00] ; HASWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_sqrtps: ; BTVER2: # BB#0: @@ -2017,19 +2017,19 @@ define <4 x float> @test_sqrtss(<4 x float> %a0, <4 x float> *%a1) { ; ; SANDY-LABEL: test_sqrtss: ; SANDY: # BB#0: -; SANDY-NEXT: vsqrtss %xmm0, %xmm0, %xmm0 # sched: [19:1.00] -; SANDY-NEXT: vmovaps (%rdi), %xmm1 # sched: [4:0.50] -; SANDY-NEXT: vsqrtss %xmm1, %xmm1, %xmm1 # sched: [19:1.00] +; SANDY-NEXT: vsqrtss %xmm0, %xmm0, %xmm0 # sched: [114:1.00] +; SANDY-NEXT: vmovaps (%rdi), %xmm1 # sched: [6:0.50] +; SANDY-NEXT: vsqrtss %xmm1, %xmm1, %xmm1 # sched: [114:1.00] ; SANDY-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_sqrtss: ; HASWELL: # BB#0: -; HASWELL-NEXT: vsqrtss %xmm0, %xmm0, %xmm0 # sched: [19:1.00] -; HASWELL-NEXT: vmovaps (%rdi), %xmm1 # sched: [4:0.50] -; HASWELL-NEXT: vsqrtss %xmm1, %xmm1, %xmm1 # sched: [19:1.00] +; HASWELL-NEXT: vsqrtss %xmm0, %xmm0, %xmm0 # sched: [14:1.00] +; HASWELL-NEXT: vmovaps (%rdi), %xmm1 # sched: [?:5.000000e-01] +; HASWELL-NEXT: vsqrtss %xmm1, %xmm1, %xmm1 # sched: [14:1.00] ; HASWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_sqrtss: ; BTVER2: # BB#0: @@ -2067,15 +2067,15 @@ define i32 @test_stmxcsr() { ; ; SANDY-LABEL: test_stmxcsr: ; SANDY: # BB#0: -; SANDY-NEXT: vstmxcsr -{{[0-9]+}}(%rsp) # sched: [1:1.00] -; SANDY-NEXT: movl -{{[0-9]+}}(%rsp), %eax # sched: [4:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vstmxcsr -{{[0-9]+}}(%rsp) # sched: [5:1.00] +; SANDY-NEXT: movl -{{[0-9]+}}(%rsp), %eax # sched: [5:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_stmxcsr: ; HASWELL: # BB#0: -; HASWELL-NEXT: vstmxcsr -{{[0-9]+}}(%rsp) # sched: [7:1.00] -; HASWELL-NEXT: movl -{{[0-9]+}}(%rsp), %eax # sched: [4:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vstmxcsr -{{[0-9]+}}(%rsp) # sched: [1:1.00] +; HASWELL-NEXT: movl -{{[0-9]+}}(%rsp), %eax # sched: [?:5.000000e-01] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_stmxcsr: ; BTVER2: # BB#0: @@ -2112,14 +2112,14 @@ define <4 x float> @test_subps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a ; SANDY-LABEL: test_subps: ; SANDY: # BB#0: ; SANDY-NEXT: vsubps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vsubps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vsubps (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_subps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vsubps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vsubps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vsubps (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_subps: ; BTVER2: # BB#0: @@ -2154,14 +2154,14 @@ define float @test_subss(float %a0, float %a1, float *%a2) { ; SANDY-LABEL: test_subss: ; SANDY: # BB#0: ; SANDY-NEXT: vsubss %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vsubss (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vsubss (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_subss: ; HASWELL: # BB#0: ; HASWELL-NEXT: vsubss %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vsubss (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vsubss (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_subss: ; BTVER2: # BB#0: @@ -2220,30 +2220,30 @@ define i32 @test_ucomiss(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) { ; SANDY-LABEL: test_ucomiss: ; SANDY: # BB#0: ; SANDY-NEXT: vucomiss %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: setnp %al # sched: [1:0.33] -; SANDY-NEXT: sete %cl # sched: [1:0.33] +; SANDY-NEXT: setnp %al # sched: [1:1.00] +; SANDY-NEXT: sete %cl # sched: [1:1.00] ; SANDY-NEXT: andb %al, %cl # sched: [1:0.33] ; SANDY-NEXT: vucomiss (%rdi), %xmm0 # sched: [7:1.00] -; SANDY-NEXT: setnp %al # sched: [1:0.33] -; SANDY-NEXT: sete %dl # sched: [1:0.33] +; SANDY-NEXT: setnp %al # sched: [1:1.00] +; SANDY-NEXT: sete %dl # sched: [1:1.00] ; SANDY-NEXT: andb %al, %dl # sched: [1:0.33] ; SANDY-NEXT: orb %cl, %dl # sched: [1:0.33] ; SANDY-NEXT: movzbl %dl, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_ucomiss: ; HASWELL: # BB#0: ; HASWELL-NEXT: vucomiss %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: setnp %al # sched: [1:0.50] -; HASWELL-NEXT: sete %cl # sched: [1:0.50] +; HASWELL-NEXT: setnp %al # sched: [1:1.00] +; HASWELL-NEXT: sete %cl # sched: [1:1.00] ; HASWELL-NEXT: andb %al, %cl # sched: [1:0.25] ; HASWELL-NEXT: vucomiss (%rdi), %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: setnp %al # sched: [1:0.50] -; HASWELL-NEXT: sete %dl # sched: [1:0.50] +; HASWELL-NEXT: setnp %al # sched: [1:1.00] +; HASWELL-NEXT: sete %dl # sched: [1:1.00] ; HASWELL-NEXT: andb %al, %dl # sched: [1:0.25] ; HASWELL-NEXT: orb %cl, %dl # sched: [1:0.25] ; HASWELL-NEXT: movzbl %dl, %eax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_ucomiss: ; BTVER2: # BB#0: @@ -2292,14 +2292,14 @@ define <4 x float> @test_unpckhps(<4 x float> %a0, <4 x float> %a1, <4 x float> ; SANDY-LABEL: test_unpckhps: ; SANDY: # BB#0: ; SANDY-NEXT: vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3] sched: [1:1.00] -; SANDY-NEXT: vunpckhps {{.*#+}} xmm0 = xmm0[2],mem[2],xmm0[3],mem[3] sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vunpckhps {{.*#+}} xmm0 = xmm0[2],mem[2],xmm0[3],mem[3] sched: [7:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_unpckhps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3] sched: [1:1.00] -; HASWELL-NEXT: vunpckhps {{.*#+}} xmm0 = xmm0[2],mem[2],xmm0[3],mem[3] sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vunpckhps {{.*#+}} xmm0 = xmm0[2],mem[2],xmm0[3],mem[3] sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_unpckhps: ; BTVER2: # BB#0: @@ -2338,14 +2338,14 @@ define <4 x float> @test_unpcklps(<4 x float> %a0, <4 x float> %a1, <4 x float> ; SANDY-LABEL: test_unpcklps: ; SANDY: # BB#0: ; SANDY-NEXT: vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] sched: [1:1.00] -; SANDY-NEXT: vunpcklps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vunpcklps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] sched: [7:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_unpcklps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] sched: [1:1.00] -; HASWELL-NEXT: vunpcklps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vunpcklps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_unpcklps: ; BTVER2: # BB#0: @@ -2387,15 +2387,15 @@ define <4 x float> @test_xorps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a ; ; SANDY-LABEL: test_xorps: ; SANDY: # BB#0: -; SANDY-NEXT: vxorps %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: vxorps (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vxorps %xmm1, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: vxorps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_xorps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vxorps %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vxorps (%rdi), %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vxorps (%rdi), %xmm0, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_xorps: ; BTVER2: # BB#0: diff --git a/llvm/test/CodeGen/X86/sse2-schedule.ll b/llvm/test/CodeGen/X86/sse2-schedule.ll index 14c155c8c6c..29782d8c01a 100644 --- a/llvm/test/CodeGen/X86/sse2-schedule.ll +++ b/llvm/test/CodeGen/X86/sse2-schedule.ll @@ -31,14 +31,14 @@ define <2 x double> @test_addpd(<2 x double> %a0, <2 x double> %a1, <2 x double> ; SANDY-LABEL: test_addpd: ; SANDY: # BB#0: ; SANDY-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vaddpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vaddpd (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_addpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vaddpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vaddpd (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_addpd: ; BTVER2: # BB#0: @@ -73,14 +73,14 @@ define double @test_addsd(double %a0, double %a1, double *%a2) { ; SANDY-LABEL: test_addsd: ; SANDY: # BB#0: ; SANDY-NEXT: vaddsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vaddsd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vaddsd (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_addsd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vaddsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vaddsd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vaddsd (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_addsd: ; BTVER2: # BB#0: @@ -117,17 +117,17 @@ define <2 x double> @test_andpd(<2 x double> %a0, <2 x double> %a1, <2 x double> ; ; SANDY-LABEL: test_andpd: ; SANDY: # BB#0: -; SANDY-NEXT: vandpd %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: vandpd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; SANDY-NEXT: vandpd %xmm1, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: vandpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] ; SANDY-NEXT: vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_andpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vandpd %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vandpd (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: vandpd (%rdi), %xmm0, %xmm0 # sched: [1:1.00] ; HASWELL-NEXT: vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_andpd: ; BTVER2: # BB#0: @@ -170,17 +170,17 @@ define <2 x double> @test_andnotpd(<2 x double> %a0, <2 x double> %a1, <2 x doub ; ; SANDY-LABEL: test_andnotpd: ; SANDY: # BB#0: -; SANDY-NEXT: vandnpd %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: vandnpd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; SANDY-NEXT: vandnpd %xmm1, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: vandnpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] ; SANDY-NEXT: vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_andnotpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vandnpd %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vandnpd (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: vandnpd (%rdi), %xmm0, %xmm0 # sched: [1:1.00] ; HASWELL-NEXT: vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_andnotpd: ; BTVER2: # BB#0: @@ -226,16 +226,16 @@ define <2 x double> @test_cmppd(<2 x double> %a0, <2 x double> %a1, <2 x double> ; SANDY-LABEL: test_cmppd: ; SANDY: # BB#0: ; SANDY-NEXT: vcmpeqpd %xmm1, %xmm0, %xmm1 # sched: [3:1.00] -; SANDY-NEXT: vcmpeqpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: vorpd %xmm0, %xmm1, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vcmpeqpd (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: vorpd %xmm0, %xmm1, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cmppd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcmpeqpd %xmm1, %xmm0, %xmm1 # sched: [3:1.00] -; HASWELL-NEXT: vcmpeqpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] +; HASWELL-NEXT: vcmpeqpd (%rdi), %xmm0, %xmm0 # sched: [3:1.00] ; HASWELL-NEXT: vorpd %xmm0, %xmm1, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cmppd: ; BTVER2: # BB#0: @@ -275,13 +275,13 @@ define double @test_cmpsd(double %a0, double %a1, double *%a2) { ; SANDY: # BB#0: ; SANDY-NEXT: vcmpeqsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vcmpeqsd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cmpsd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcmpeqsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] ; HASWELL-NEXT: vcmpeqsd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cmpsd: ; BTVER2: # BB#0: @@ -345,30 +345,30 @@ define i32 @test_comisd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) { ; SANDY-LABEL: test_comisd: ; SANDY: # BB#0: ; SANDY-NEXT: vcomisd %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: setnp %al # sched: [1:0.33] -; SANDY-NEXT: sete %cl # sched: [1:0.33] +; SANDY-NEXT: setnp %al # sched: [1:1.00] +; SANDY-NEXT: sete %cl # sched: [1:1.00] ; SANDY-NEXT: andb %al, %cl # sched: [1:0.33] ; SANDY-NEXT: vcomisd (%rdi), %xmm0 # sched: [7:1.00] -; SANDY-NEXT: setnp %al # sched: [1:0.33] -; SANDY-NEXT: sete %dl # sched: [1:0.33] +; SANDY-NEXT: setnp %al # sched: [1:1.00] +; SANDY-NEXT: sete %dl # sched: [1:1.00] ; SANDY-NEXT: andb %al, %dl # sched: [1:0.33] ; SANDY-NEXT: orb %cl, %dl # sched: [1:0.33] ; SANDY-NEXT: movzbl %dl, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_comisd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcomisd %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: setnp %al # sched: [1:0.50] -; HASWELL-NEXT: sete %cl # sched: [1:0.50] +; HASWELL-NEXT: setnp %al # sched: [1:1.00] +; HASWELL-NEXT: sete %cl # sched: [1:1.00] ; HASWELL-NEXT: andb %al, %cl # sched: [1:0.25] ; HASWELL-NEXT: vcomisd (%rdi), %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: setnp %al # sched: [1:0.50] -; HASWELL-NEXT: sete %dl # sched: [1:0.50] +; HASWELL-NEXT: setnp %al # sched: [1:1.00] +; HASWELL-NEXT: sete %dl # sched: [1:1.00] ; HASWELL-NEXT: andb %al, %dl # sched: [1:0.25] ; HASWELL-NEXT: orb %cl, %dl # sched: [1:0.25] ; HASWELL-NEXT: movzbl %dl, %eax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_comisd: ; BTVER2: # BB#0: @@ -416,16 +416,16 @@ define <2 x double> @test_cvtdq2pd(<4 x i32> %a0, <4 x i32> *%a1) { ; SANDY-LABEL: test_cvtdq2pd: ; SANDY: # BB#0: ; SANDY-NEXT: vcvtdq2pd %xmm0, %xmm0 # sched: [4:1.00] -; SANDY-NEXT: vcvtdq2pd (%rdi), %xmm1 # sched: [8:1.00] +; SANDY-NEXT: vcvtdq2pd (%rdi), %xmm1 # sched: [10:1.00] ; SANDY-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtdq2pd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvtdq2pd %xmm0, %xmm0 # sched: [4:1.00] -; HASWELL-NEXT: vcvtdq2pd (%rdi), %xmm1 # sched: [8:1.00] +; HASWELL-NEXT: vcvtdq2pd (%rdi), %xmm1 # sched: [4:1.00] ; HASWELL-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtdq2pd: ; BTVER2: # BB#0: @@ -467,17 +467,17 @@ define <4 x float> @test_cvtdq2ps(<4 x i32> %a0, <4 x i32> *%a1) { ; ; SANDY-LABEL: test_cvtdq2ps: ; SANDY: # BB#0: -; SANDY-NEXT: vcvtdq2ps %xmm0, %xmm0 # sched: [4:1.00] -; SANDY-NEXT: vcvtdq2ps (%rdi), %xmm1 # sched: [8:1.00] +; SANDY-NEXT: vcvtdq2ps %xmm0, %xmm0 # sched: [3:1.00] +; SANDY-NEXT: vcvtdq2ps (%rdi), %xmm1 # sched: [9:1.00] ; SANDY-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtdq2ps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vcvtdq2ps %xmm0, %xmm0 # sched: [4:1.00] -; HASWELL-NEXT: vcvtdq2ps (%rdi), %xmm1 # sched: [8:1.00] +; HASWELL-NEXT: vcvtdq2ps %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: vcvtdq2ps (%rdi), %xmm1 # sched: [3:1.00] ; HASWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtdq2ps: ; BTVER2: # BB#0: @@ -517,17 +517,17 @@ define <4 x i32> @test_cvtpd2dq(<2 x double> %a0, <2 x double> *%a1) { ; ; SANDY-LABEL: test_cvtpd2dq: ; SANDY: # BB#0: -; SANDY-NEXT: vcvtpd2dq %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vcvtpd2dqx (%rdi), %xmm1 # sched: [7:1.00] +; SANDY-NEXT: vcvtpd2dq %xmm0, %xmm0 # sched: [4:1.00] +; SANDY-NEXT: vcvtpd2dqx (%rdi), %xmm1 # sched: [10:1.00] ; SANDY-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtpd2dq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvtpd2dq %xmm0, %xmm0 # sched: [4:1.00] -; HASWELL-NEXT: vcvtpd2dqx (%rdi), %xmm1 # sched: [8:1.00] +; HASWELL-NEXT: vcvtpd2dqx (%rdi), %xmm1 # sched: [7:1.00] ; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtpd2dq: ; BTVER2: # BB#0: @@ -568,17 +568,17 @@ define <4 x float> @test_cvtpd2ps(<2 x double> %a0, <2 x double> *%a1) { ; ; SANDY-LABEL: test_cvtpd2ps: ; SANDY: # BB#0: -; SANDY-NEXT: vcvtpd2ps %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vcvtpd2psx (%rdi), %xmm1 # sched: [7:1.00] +; SANDY-NEXT: vcvtpd2ps %xmm0, %xmm0 # sched: [4:1.00] +; SANDY-NEXT: vcvtpd2psx (%rdi), %xmm1 # sched: [10:1.00] ; SANDY-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtpd2ps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvtpd2ps %xmm0, %xmm0 # sched: [4:1.00] -; HASWELL-NEXT: vcvtpd2psx (%rdi), %xmm1 # sched: [8:1.00] +; HASWELL-NEXT: vcvtpd2psx (%rdi), %xmm1 # sched: [7:1.00] ; HASWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtpd2ps: ; BTVER2: # BB#0: @@ -620,16 +620,16 @@ define <4 x i32> @test_cvtps2dq(<4 x float> %a0, <4 x float> *%a1) { ; SANDY-LABEL: test_cvtps2dq: ; SANDY: # BB#0: ; SANDY-NEXT: vcvtps2dq %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vcvtps2dq (%rdi), %xmm1 # sched: [7:1.00] +; SANDY-NEXT: vcvtps2dq (%rdi), %xmm1 # sched: [9:1.00] ; SANDY-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtps2dq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvtps2dq %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vcvtps2dq (%rdi), %xmm1 # sched: [7:1.00] +; HASWELL-NEXT: vcvtps2dq (%rdi), %xmm1 # sched: [3:1.00] ; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtps2dq: ; BTVER2: # BB#0: @@ -670,17 +670,17 @@ define <2 x double> @test_cvtps2pd(<4 x float> %a0, <4 x float> *%a1) { ; ; SANDY-LABEL: test_cvtps2pd: ; SANDY: # BB#0: -; SANDY-NEXT: vcvtps2pd %xmm0, %xmm0 # sched: [3:1.00] +; SANDY-NEXT: vcvtps2pd %xmm0, %xmm0 # sched: [2:1.00] ; SANDY-NEXT: vcvtps2pd (%rdi), %xmm1 # sched: [7:1.00] ; SANDY-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtps2pd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvtps2pd %xmm0, %xmm0 # sched: [2:1.00] -; HASWELL-NEXT: vcvtps2pd (%rdi), %xmm1 # sched: [5:1.00] +; HASWELL-NEXT: vcvtps2pd (%rdi), %xmm1 # sched: [1:1.00] ; HASWELL-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtps2pd: ; BTVER2: # BB#0: @@ -724,14 +724,14 @@ define i32 @test_cvtsd2si(double %a0, double *%a1) { ; SANDY-NEXT: vcvtsd2si %xmm0, %ecx # sched: [3:1.00] ; SANDY-NEXT: vcvtsd2si (%rdi), %eax # sched: [7:1.00] ; SANDY-NEXT: addl %ecx, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtsd2si: ; HASWELL: # BB#0: -; HASWELL-NEXT: vcvtsd2si %xmm0, %ecx # sched: [4:1.00] -; HASWELL-NEXT: vcvtsd2si (%rdi), %eax # sched: [8:1.00] +; HASWELL-NEXT: vcvtsd2si %xmm0, %ecx # sched: [3:1.00] +; HASWELL-NEXT: vcvtsd2si (%rdi), %eax # sched: [7:1.00] ; HASWELL-NEXT: addl %ecx, %eax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtsd2si: ; BTVER2: # BB#0: @@ -773,17 +773,17 @@ define i64 @test_cvtsd2siq(double %a0, double *%a1) { ; ; SANDY-LABEL: test_cvtsd2siq: ; SANDY: # BB#0: -; SANDY-NEXT: vcvtsd2si %xmm0, %rcx # sched: [3:1.00] -; SANDY-NEXT: vcvtsd2si (%rdi), %rax # sched: [7:1.00] +; SANDY-NEXT: vcvtsd2si %xmm0, %rcx # sched: [5:1.00] +; SANDY-NEXT: vcvtsd2si (%rdi), %rax # sched: [10:1.00] ; SANDY-NEXT: addq %rcx, %rax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtsd2siq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvtsd2si %xmm0, %rcx # sched: [4:1.00] -; HASWELL-NEXT: vcvtsd2si (%rdi), %rax # sched: [8:1.00] +; HASWELL-NEXT: vcvtsd2si (%rdi), %rax # sched: [4:1.00] ; HASWELL-NEXT: addq %rcx, %rax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtsd2siq: ; BTVER2: # BB#0: @@ -830,18 +830,18 @@ define float @test_cvtsd2ss(double %a0, double *%a1) { ; SANDY-LABEL: test_cvtsd2ss: ; SANDY: # BB#0: ; SANDY-NEXT: vcvtsd2ss %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero sched: [4:0.50] +; SANDY-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero sched: [6:0.50] ; SANDY-NEXT: vcvtsd2ss %xmm1, %xmm1, %xmm1 # sched: [3:1.00] ; SANDY-NEXT: vaddss %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtsd2ss: ; HASWELL: # BB#0: -; HASWELL-NEXT: vcvtsd2ss %xmm0, %xmm0, %xmm0 # sched: [4:1.00] -; HASWELL-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero sched: [4:0.50] -; HASWELL-NEXT: vcvtsd2ss %xmm1, %xmm1, %xmm1 # sched: [4:1.00] +; HASWELL-NEXT: vcvtsd2ss %xmm0, %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero sched: [?:5.000000e-01] +; HASWELL-NEXT: vcvtsd2ss %xmm1, %xmm1, %xmm1 # sched: [3:1.00] ; HASWELL-NEXT: vaddss %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtsd2ss: ; BTVER2: # BB#0: @@ -882,16 +882,16 @@ define double @test_cvtsi2sd(i32 %a0, i32 *%a1) { ; SANDY-LABEL: test_cvtsi2sd: ; SANDY: # BB#0: ; SANDY-NEXT: vcvtsi2sdl %edi, %xmm0, %xmm0 # sched: [4:1.00] -; SANDY-NEXT: vcvtsi2sdl (%rsi), %xmm1, %xmm1 # sched: [8:1.00] +; SANDY-NEXT: vcvtsi2sdl (%rsi), %xmm1, %xmm1 # sched: [9:1.00] ; SANDY-NEXT: vaddsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtsi2sd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvtsi2sdl %edi, %xmm0, %xmm0 # sched: [4:1.00] ; HASWELL-NEXT: vcvtsi2sdl (%rsi), %xmm1, %xmm1 # sched: [8:1.00] ; HASWELL-NEXT: vaddsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtsi2sd: ; BTVER2: # BB#0: @@ -931,16 +931,16 @@ define double @test_cvtsi2sdq(i64 %a0, i64 *%a1) { ; SANDY-LABEL: test_cvtsi2sdq: ; SANDY: # BB#0: ; SANDY-NEXT: vcvtsi2sdq %rdi, %xmm0, %xmm0 # sched: [4:1.00] -; SANDY-NEXT: vcvtsi2sdq (%rsi), %xmm1, %xmm1 # sched: [8:1.00] +; SANDY-NEXT: vcvtsi2sdq (%rsi), %xmm1, %xmm1 # sched: [9:1.00] ; SANDY-NEXT: vaddsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtsi2sdq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvtsi2sdq %rdi, %xmm0, %xmm0 # sched: [4:1.00] ; HASWELL-NEXT: vcvtsi2sdq (%rsi), %xmm1, %xmm1 # sched: [8:1.00] ; HASWELL-NEXT: vaddsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtsi2sdq: ; BTVER2: # BB#0: @@ -985,19 +985,19 @@ define double @test_cvtss2sd(float %a0, float *%a1) { ; ; SANDY-LABEL: test_cvtss2sd: ; SANDY: # BB#0: -; SANDY-NEXT: vcvtss2sd %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [4:0.50] -; SANDY-NEXT: vcvtss2sd %xmm1, %xmm1, %xmm1 # sched: [3:1.00] +; SANDY-NEXT: vcvtss2sd %xmm0, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [6:0.50] +; SANDY-NEXT: vcvtss2sd %xmm1, %xmm1, %xmm1 # sched: [1:1.00] ; SANDY-NEXT: vaddsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvtss2sd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvtss2sd %xmm0, %xmm0, %xmm0 # sched: [2:1.00] -; HASWELL-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [4:0.50] +; HASWELL-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [?:5.000000e-01] ; HASWELL-NEXT: vcvtss2sd %xmm1, %xmm1, %xmm1 # sched: [2:1.00] ; HASWELL-NEXT: vaddsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvtss2sd: ; BTVER2: # BB#0: @@ -1038,17 +1038,17 @@ define <4 x i32> @test_cvttpd2dq(<2 x double> %a0, <2 x double> *%a1) { ; ; SANDY-LABEL: test_cvttpd2dq: ; SANDY: # BB#0: -; SANDY-NEXT: vcvttpd2dq %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vcvttpd2dqx (%rdi), %xmm1 # sched: [7:1.00] +; SANDY-NEXT: vcvttpd2dq %xmm0, %xmm0 # sched: [4:1.00] +; SANDY-NEXT: vcvttpd2dqx (%rdi), %xmm1 # sched: [10:1.00] ; SANDY-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvttpd2dq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvttpd2dq %xmm0, %xmm0 # sched: [4:1.00] -; HASWELL-NEXT: vcvttpd2dqx (%rdi), %xmm1 # sched: [8:1.00] +; HASWELL-NEXT: vcvttpd2dqx (%rdi), %xmm1 # sched: [7:1.00] ; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvttpd2dq: ; BTVER2: # BB#0: @@ -1091,16 +1091,16 @@ define <4 x i32> @test_cvttps2dq(<4 x float> %a0, <4 x float> *%a1) { ; SANDY-LABEL: test_cvttps2dq: ; SANDY: # BB#0: ; SANDY-NEXT: vcvttps2dq %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vcvttps2dq (%rdi), %xmm1 # sched: [7:1.00] +; SANDY-NEXT: vcvttps2dq (%rdi), %xmm1 # sched: [9:1.00] ; SANDY-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvttps2dq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvttps2dq %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vcvttps2dq (%rdi), %xmm1 # sched: [7:1.00] +; HASWELL-NEXT: vcvttps2dq (%rdi), %xmm1 # sched: [3:1.00] ; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvttps2dq: ; BTVER2: # BB#0: @@ -1139,17 +1139,17 @@ define i32 @test_cvttsd2si(double %a0, double *%a1) { ; ; SANDY-LABEL: test_cvttsd2si: ; SANDY: # BB#0: -; SANDY-NEXT: vcvttsd2si %xmm0, %ecx # sched: [3:1.00] +; SANDY-NEXT: vcvttsd2si %xmm0, %ecx # sched: [5:1.00] ; SANDY-NEXT: vcvttsd2si (%rdi), %eax # sched: [7:1.00] ; SANDY-NEXT: addl %ecx, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvttsd2si: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvttsd2si %xmm0, %ecx # sched: [4:1.00] -; HASWELL-NEXT: vcvttsd2si (%rdi), %eax # sched: [8:1.00] +; HASWELL-NEXT: vcvttsd2si (%rdi), %eax # sched: [7:1.00] ; HASWELL-NEXT: addl %ecx, %eax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvttsd2si: ; BTVER2: # BB#0: @@ -1188,17 +1188,17 @@ define i64 @test_cvttsd2siq(double %a0, double *%a1) { ; ; SANDY-LABEL: test_cvttsd2siq: ; SANDY: # BB#0: -; SANDY-NEXT: vcvttsd2si %xmm0, %rcx # sched: [3:1.00] -; SANDY-NEXT: vcvttsd2si (%rdi), %rax # sched: [7:1.00] +; SANDY-NEXT: vcvttsd2si %xmm0, %rcx # sched: [5:1.00] +; SANDY-NEXT: vcvttsd2si (%rdi), %rax # sched: [10:1.00] ; SANDY-NEXT: addq %rcx, %rax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_cvttsd2siq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vcvttsd2si %xmm0, %rcx # sched: [4:1.00] -; HASWELL-NEXT: vcvttsd2si (%rdi), %rax # sched: [8:1.00] +; HASWELL-NEXT: vcvttsd2si (%rdi), %rax # sched: [4:1.00] ; HASWELL-NEXT: addq %rcx, %rax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_cvttsd2siq: ; BTVER2: # BB#0: @@ -1234,15 +1234,15 @@ define <2 x double> @test_divpd(<2 x double> %a0, <2 x double> %a1, <2 x double> ; ; SANDY-LABEL: test_divpd: ; SANDY: # BB#0: -; SANDY-NEXT: vdivpd %xmm1, %xmm0, %xmm0 # sched: [12:1.00] -; SANDY-NEXT: vdivpd (%rdi), %xmm0, %xmm0 # sched: [16:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vdivpd %xmm1, %xmm0, %xmm0 # sched: [22:1.00] +; SANDY-NEXT: vdivpd (%rdi), %xmm0, %xmm0 # sched: [28:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_divpd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vdivpd %xmm1, %xmm0, %xmm0 # sched: [12:1.00] -; HASWELL-NEXT: vdivpd (%rdi), %xmm0, %xmm0 # sched: [16:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vdivpd %xmm1, %xmm0, %xmm0 # sched: [19:4.00] +; HASWELL-NEXT: vdivpd (%rdi), %xmm0, %xmm0 # sched: [19:4.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_divpd: ; BTVER2: # BB#0: @@ -1276,15 +1276,15 @@ define double @test_divsd(double %a0, double %a1, double *%a2) { ; ; SANDY-LABEL: test_divsd: ; SANDY: # BB#0: -; SANDY-NEXT: vdivsd %xmm1, %xmm0, %xmm0 # sched: [12:1.00] -; SANDY-NEXT: vdivsd (%rdi), %xmm0, %xmm0 # sched: [16:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vdivsd %xmm1, %xmm0, %xmm0 # sched: [22:1.00] +; SANDY-NEXT: vdivsd (%rdi), %xmm0, %xmm0 # sched: [28:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_divsd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vdivsd %xmm1, %xmm0, %xmm0 # sched: [12:1.00] -; HASWELL-NEXT: vdivsd (%rdi), %xmm0, %xmm0 # sched: [16:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vdivsd %xmm1, %xmm0, %xmm0 # sched: [19:4.00] +; HASWELL-NEXT: vdivsd (%rdi), %xmm0, %xmm0 # sched: [19:4.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_divsd: ; BTVER2: # BB#0: @@ -1322,12 +1322,12 @@ define void @test_lfence() { ; SANDY-LABEL: test_lfence: ; SANDY: # BB#0: ; SANDY-NEXT: lfence # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_lfence: ; HASWELL: # BB#0: -; HASWELL-NEXT: lfence # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: lfence # sched: [2:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_lfence: ; BTVER2: # BB#0: @@ -1363,12 +1363,12 @@ define void @test_mfence() { ; SANDY-LABEL: test_mfence: ; SANDY: # BB#0: ; SANDY-NEXT: mfence # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_mfence: ; HASWELL: # BB#0: -; HASWELL-NEXT: mfence # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: mfence # sched: [2:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_mfence: ; BTVER2: # BB#0: @@ -1402,12 +1402,12 @@ define void @test_maskmovdqu(<16 x i8> %a0, <16 x i8> %a1, i8* %a2) { ; SANDY-LABEL: test_maskmovdqu: ; SANDY: # BB#0: ; SANDY-NEXT: vmaskmovdqu %xmm1, %xmm0 # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_maskmovdqu: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmaskmovdqu %xmm1, %xmm0 # sched: [14:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmaskmovdqu %xmm1, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_maskmovdqu: ; BTVER2: # BB#0: @@ -1440,14 +1440,14 @@ define <2 x double> @test_maxpd(<2 x double> %a0, <2 x double> %a1, <2 x double> ; SANDY-LABEL: test_maxpd: ; SANDY: # BB#0: ; SANDY-NEXT: vmaxpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmaxpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmaxpd (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_maxpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmaxpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vmaxpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmaxpd (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_maxpd: ; BTVER2: # BB#0: @@ -1483,14 +1483,14 @@ define <2 x double> @test_maxsd(<2 x double> %a0, <2 x double> %a1, <2 x double> ; SANDY-LABEL: test_maxsd: ; SANDY: # BB#0: ; SANDY-NEXT: vmaxsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmaxsd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmaxsd (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_maxsd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmaxsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vmaxsd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmaxsd (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_maxsd: ; BTVER2: # BB#0: @@ -1526,14 +1526,14 @@ define <2 x double> @test_minpd(<2 x double> %a0, <2 x double> %a1, <2 x double> ; SANDY-LABEL: test_minpd: ; SANDY: # BB#0: ; SANDY-NEXT: vminpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vminpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vminpd (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_minpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vminpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vminpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vminpd (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_minpd: ; BTVER2: # BB#0: @@ -1569,14 +1569,14 @@ define <2 x double> @test_minsd(<2 x double> %a0, <2 x double> %a1, <2 x double> ; SANDY-LABEL: test_minsd: ; SANDY: # BB#0: ; SANDY-NEXT: vminsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vminsd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vminsd (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_minsd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vminsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vminsd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vminsd (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_minsd: ; BTVER2: # BB#0: @@ -1614,17 +1614,17 @@ define void @test_movapd(<2 x double> *%a0, <2 x double> *%a1) { ; ; SANDY-LABEL: test_movapd: ; SANDY: # BB#0: -; SANDY-NEXT: vmovapd (%rdi), %xmm0 # sched: [4:0.50] +; SANDY-NEXT: vmovapd (%rdi), %xmm0 # sched: [6:0.50] ; SANDY-NEXT: vaddpd %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmovapd %xmm0, (%rsi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovapd %xmm0, (%rsi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movapd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovapd (%rdi), %xmm0 # sched: [4:0.50] +; HASWELL-NEXT: vmovapd (%rdi), %xmm0 # sched: [?:5.000000e-01] ; HASWELL-NEXT: vaddpd %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vmovapd %xmm0, (%rsi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovapd %xmm0, (%rsi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movapd: ; BTVER2: # BB#0: @@ -1662,17 +1662,17 @@ define void @test_movdqa(<2 x i64> *%a0, <2 x i64> *%a1) { ; ; SANDY-LABEL: test_movdqa: ; SANDY: # BB#0: -; SANDY-NEXT: vmovdqa (%rdi), %xmm0 # sched: [4:0.50] +; SANDY-NEXT: vmovdqa (%rdi), %xmm0 # sched: [6:0.50] ; SANDY-NEXT: vpaddq %xmm0, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vmovdqa %xmm0, (%rsi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovdqa %xmm0, (%rsi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movdqa: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovdqa (%rdi), %xmm0 # sched: [4:0.50] +; HASWELL-NEXT: vmovdqa (%rdi), %xmm0 # sched: [?:5.000000e-01] ; HASWELL-NEXT: vpaddq %xmm0, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vmovdqa %xmm0, (%rsi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovdqa %xmm0, (%rsi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movdqa: ; BTVER2: # BB#0: @@ -1710,17 +1710,17 @@ define void @test_movdqu(<2 x i64> *%a0, <2 x i64> *%a1) { ; ; SANDY-LABEL: test_movdqu: ; SANDY: # BB#0: -; SANDY-NEXT: vmovdqu (%rdi), %xmm0 # sched: [4:0.50] +; SANDY-NEXT: vmovdqu (%rdi), %xmm0 # sched: [6:0.50] ; SANDY-NEXT: vpaddq %xmm0, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vmovdqu %xmm0, (%rsi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovdqu %xmm0, (%rsi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movdqu: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovdqu (%rdi), %xmm0 # sched: [4:0.50] +; HASWELL-NEXT: vmovdqu (%rdi), %xmm0 # sched: [?:5.000000e-01] ; HASWELL-NEXT: vpaddq %xmm0, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vmovdqu %xmm0, (%rsi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovdqu %xmm0, (%rsi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movdqu: ; BTVER2: # BB#0: @@ -1768,22 +1768,22 @@ define i32 @test_movd(<4 x i32> %a0, i32 %a1, i32 *%a2) { ; SANDY-LABEL: test_movd: ; SANDY: # BB#0: ; SANDY-NEXT: vmovd %edi, %xmm1 # sched: [1:0.33] -; SANDY-NEXT: vmovd {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50] +; SANDY-NEXT: vmovd {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [6:0.50] ; SANDY-NEXT: vpaddd %xmm1, %xmm0, %xmm1 # sched: [1:0.50] ; SANDY-NEXT: vpaddd %xmm2, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vmovd %xmm0, %eax # sched: [1:0.33] -; SANDY-NEXT: vmovd %xmm1, (%rsi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovd %xmm0, %eax # sched: [2:1.00] +; SANDY-NEXT: vmovd %xmm1, (%rsi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovd %edi, %xmm1 # sched: [1:1.00] -; HASWELL-NEXT: vmovd {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [4:0.50] +; HASWELL-NEXT: vmovd %edi, %xmm1 # sched: [1:0.25] +; HASWELL-NEXT: vmovd {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [?:5.000000e-01] ; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm1 # sched: [1:0.50] ; HASWELL-NEXT: vpaddd %xmm2, %xmm0, %xmm0 # sched: [1:0.50] ; HASWELL-NEXT: vmovd %xmm0, %eax # sched: [1:1.00] -; HASWELL-NEXT: vmovd %xmm1, (%rsi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovd %xmm1, (%rsi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movd: ; BTVER2: # BB#0: @@ -1838,23 +1838,23 @@ define i64 @test_movd_64(<2 x i64> %a0, i64 %a1, i64 *%a2) { ; ; SANDY-LABEL: test_movd_64: ; SANDY: # BB#0: -; SANDY-NEXT: vmovq %rdi, %xmm1 # sched: [1:0.33] -; SANDY-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero sched: [4:0.50] +; SANDY-NEXT: vmovq %rdi, %xmm1 # sched: [1:1.00] +; SANDY-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero sched: [6:0.50] ; SANDY-NEXT: vpaddq %xmm1, %xmm0, %xmm1 # sched: [1:0.50] ; SANDY-NEXT: vpaddq %xmm2, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vmovq %xmm0, %rax # sched: [1:0.33] -; SANDY-NEXT: vmovq %xmm1, (%rsi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovq %xmm0, %rax # sched: [2:1.00] +; SANDY-NEXT: vmovq %xmm1, (%rsi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movd_64: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmovq %rdi, %xmm1 # sched: [1:1.00] -; HASWELL-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero sched: [4:0.50] +; HASWELL-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero sched: [?:5.000000e-01] ; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm1 # sched: [1:0.50] ; HASWELL-NEXT: vpaddq %xmm2, %xmm0, %xmm0 # sched: [1:0.50] ; HASWELL-NEXT: vmovq %xmm0, %rax # sched: [1:1.00] -; HASWELL-NEXT: vmovq %xmm1, (%rsi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovq %xmm1, (%rsi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movd_64: ; BTVER2: # BB#0: @@ -1900,17 +1900,17 @@ define void @test_movhpd(<2 x double> %a0, <2 x double> %a1, x86_mmx *%a2) { ; ; SANDY-LABEL: test_movhpd: ; SANDY: # BB#0: -; SANDY-NEXT: vmovhpd {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [5:1.00] +; SANDY-NEXT: vmovhpd {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [7:1.00] ; SANDY-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmovhpd %xmm0, (%rdi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovhpd %xmm0, (%rdi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movhpd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovhpd {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [5:1.00] +; HASWELL-NEXT: vmovhpd {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [1:1.00] ; HASWELL-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vmovhpd %xmm0, (%rdi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovhpd %xmm0, (%rdi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movhpd: ; BTVER2: # BB#0: @@ -1951,17 +1951,17 @@ define void @test_movlpd(<2 x double> %a0, <2 x double> %a1, x86_mmx *%a2) { ; ; SANDY-LABEL: test_movlpd: ; SANDY: # BB#0: -; SANDY-NEXT: vmovlpd {{.*#+}} xmm1 = mem[0],xmm1[1] sched: [5:1.00] +; SANDY-NEXT: vmovlpd {{.*#+}} xmm1 = mem[0],xmm1[1] sched: [7:1.00] ; SANDY-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmovlpd %xmm0, (%rdi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovlpd %xmm0, (%rdi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movlpd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovlpd {{.*#+}} xmm1 = mem[0],xmm1[1] sched: [5:1.00] +; HASWELL-NEXT: vmovlpd {{.*#+}} xmm1 = mem[0],xmm1[1] sched: [1:1.00] ; HASWELL-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vmovlpd %xmm0, (%rdi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovlpd %xmm0, (%rdi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movlpd: ; BTVER2: # BB#0: @@ -1998,13 +1998,13 @@ define i32 @test_movmskpd(<2 x double> %a0) { ; ; SANDY-LABEL: test_movmskpd: ; SANDY: # BB#0: -; SANDY-NEXT: vmovmskpd %xmm0, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovmskpd %xmm0, %eax # sched: [2:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movmskpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmovmskpd %xmm0, %eax # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movmskpd: ; BTVER2: # BB#0: @@ -2039,14 +2039,14 @@ define void @test_movntdqa(<2 x i64> %a0, <2 x i64> *%a1) { ; SANDY-LABEL: test_movntdqa: ; SANDY: # BB#0: ; SANDY-NEXT: vpaddq %xmm0, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vmovntdq %xmm0, (%rdi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovntdq %xmm0, (%rdi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movntdqa: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpaddq %xmm0, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vmovntdq %xmm0, (%rdi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovntdq %xmm0, (%rdi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movntdqa: ; BTVER2: # BB#0: @@ -2080,14 +2080,14 @@ define void @test_movntpd(<2 x double> %a0, <2 x double> *%a1) { ; SANDY-LABEL: test_movntpd: ; SANDY: # BB#0: ; SANDY-NEXT: vaddpd %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmovntpd %xmm0, (%rdi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovntpd %xmm0, (%rdi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movntpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vaddpd %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vmovntpd %xmm0, (%rdi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovntpd %xmm0, (%rdi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movntpd: ; BTVER2: # BB#0: @@ -2123,17 +2123,17 @@ define <2 x i64> @test_movq_mem(<2 x i64> %a0, i64 *%a1) { ; ; SANDY-LABEL: test_movq_mem: ; SANDY: # BB#0: -; SANDY-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero sched: [4:0.50] +; SANDY-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero sched: [6:0.50] ; SANDY-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vmovq %xmm0, (%rdi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovq %xmm0, (%rdi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movq_mem: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero sched: [4:0.50] +; HASWELL-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero sched: [?:5.000000e-01] ; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vmovq %xmm0, (%rdi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovq %xmm0, (%rdi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movq_mem: ; BTVER2: # BB#0: @@ -2174,13 +2174,13 @@ define <2 x i64> @test_movq_reg(<2 x i64> %a0, <2 x i64> %a1) { ; SANDY: # BB#0: ; SANDY-NEXT: vmovq {{.*#+}} xmm0 = xmm0[0],zero sched: [1:0.33] ; SANDY-NEXT: vpaddq %xmm0, %xmm1, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movq_reg: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmovq {{.*#+}} xmm0 = xmm0[0],zero sched: [1:0.33] ; HASWELL-NEXT: vpaddq %xmm0, %xmm1, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movq_reg: ; BTVER2: # BB#0: @@ -2216,17 +2216,17 @@ define void @test_movsd_mem(double* %a0, double* %a1) { ; ; SANDY-LABEL: test_movsd_mem: ; SANDY: # BB#0: -; SANDY-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero sched: [4:0.50] +; SANDY-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero sched: [6:0.50] ; SANDY-NEXT: vaddsd %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmovsd %xmm0, (%rsi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovsd %xmm0, (%rsi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movsd_mem: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero sched: [4:0.50] +; HASWELL-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero sched: [?:5.000000e-01] ; HASWELL-NEXT: vaddsd %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vmovsd %xmm0, (%rsi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovsd %xmm0, (%rsi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movsd_mem: ; BTVER2: # BB#0: @@ -2266,12 +2266,12 @@ define <2 x double> @test_movsd_reg(<2 x double> %a0, <2 x double> %a1) { ; SANDY-LABEL: test_movsd_reg: ; SANDY: # BB#0: ; SANDY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0] sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movsd_reg: ; HASWELL: # BB#0: ; HASWELL-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0] sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movsd_reg: ; BTVER2: # BB#0: @@ -2305,17 +2305,17 @@ define void @test_movupd(<2 x double> *%a0, <2 x double> *%a1) { ; ; SANDY-LABEL: test_movupd: ; SANDY: # BB#0: -; SANDY-NEXT: vmovupd (%rdi), %xmm0 # sched: [4:0.50] +; SANDY-NEXT: vmovupd (%rdi), %xmm0 # sched: [6:0.50] ; SANDY-NEXT: vaddpd %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vmovupd %xmm0, (%rsi) # sched: [1:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovupd %xmm0, (%rsi) # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movupd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovupd (%rdi), %xmm0 # sched: [4:0.50] +; HASWELL-NEXT: vmovupd (%rdi), %xmm0 # sched: [?:5.000000e-01] ; HASWELL-NEXT: vaddpd %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vmovupd %xmm0, (%rsi) # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovupd %xmm0, (%rsi) # sched: [?:1.000000e+00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movupd: ; BTVER2: # BB#0: @@ -2351,14 +2351,14 @@ define <2 x double> @test_mulpd(<2 x double> %a0, <2 x double> %a1, <2 x double> ; SANDY-LABEL: test_mulpd: ; SANDY: # BB#0: ; SANDY-NEXT: vmulpd %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: vmulpd (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmulpd (%rdi), %xmm0, %xmm0 # sched: [11:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_mulpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmulpd %xmm1, %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: vmulpd (%rdi), %xmm0, %xmm0 # sched: [9:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmulpd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_mulpd: ; BTVER2: # BB#0: @@ -2393,14 +2393,14 @@ define double @test_mulsd(double %a0, double %a1, double *%a2) { ; SANDY-LABEL: test_mulsd: ; SANDY: # BB#0: ; SANDY-NEXT: vmulsd %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: vmulsd (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmulsd (%rdi), %xmm0, %xmm0 # sched: [11:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_mulsd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmulsd %xmm1, %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: vmulsd (%rdi), %xmm0, %xmm0 # sched: [9:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmulsd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_mulsd: ; BTVER2: # BB#0: @@ -2437,17 +2437,17 @@ define <2 x double> @test_orpd(<2 x double> %a0, <2 x double> %a1, <2 x double> ; ; SANDY-LABEL: test_orpd: ; SANDY: # BB#0: -; SANDY-NEXT: vorpd %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: vorpd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; SANDY-NEXT: vorpd %xmm1, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: vorpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] ; SANDY-NEXT: vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_orpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vorpd %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vorpd (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: vorpd (%rdi), %xmm0, %xmm0 # sched: [1:1.00] ; HASWELL-NEXT: vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_orpd: ; BTVER2: # BB#0: @@ -2496,14 +2496,14 @@ define <8 x i16> @test_packssdw(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; SANDY-LABEL: test_packssdw: ; SANDY: # BB#0: ; SANDY-NEXT: vpackssdw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpackssdw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpackssdw (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_packssdw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpackssdw %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpackssdw (%rdi), %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpackssdw (%rdi), %xmm0, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_packssdw: ; BTVER2: # BB#0: @@ -2548,14 +2548,14 @@ define <16 x i8> @test_packsswb(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_packsswb: ; SANDY: # BB#0: ; SANDY-NEXT: vpacksswb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpacksswb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpacksswb (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_packsswb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpacksswb %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpacksswb (%rdi), %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpacksswb (%rdi), %xmm0, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_packsswb: ; BTVER2: # BB#0: @@ -2600,14 +2600,14 @@ define <16 x i8> @test_packuswb(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_packuswb: ; SANDY: # BB#0: ; SANDY-NEXT: vpackuswb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpackuswb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpackuswb (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_packuswb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpackuswb %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpackuswb (%rdi), %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpackuswb (%rdi), %xmm0, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_packuswb: ; BTVER2: # BB#0: @@ -2648,14 +2648,14 @@ define <16 x i8> @test_paddb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_paddb: ; SANDY: # BB#0: ; SANDY-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpaddb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpaddb (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_paddb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpaddb (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_paddb: ; BTVER2: # BB#0: @@ -2694,14 +2694,14 @@ define <4 x i32> @test_paddd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; SANDY-LABEL: test_paddd: ; SANDY: # BB#0: ; SANDY-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpaddd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpaddd (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_paddd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpaddd (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_paddd: ; BTVER2: # BB#0: @@ -2736,14 +2736,14 @@ define <2 x i64> @test_paddq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) { ; SANDY-LABEL: test_paddq: ; SANDY: # BB#0: ; SANDY-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpaddq (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpaddq (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_paddq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddq (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpaddq (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_paddq: ; BTVER2: # BB#0: @@ -2782,14 +2782,14 @@ define <16 x i8> @test_paddsb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_paddsb: ; SANDY: # BB#0: ; SANDY-NEXT: vpaddsb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpaddsb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpaddsb (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_paddsb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpaddsb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddsb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpaddsb (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_paddsb: ; BTVER2: # BB#0: @@ -2829,14 +2829,14 @@ define <8 x i16> @test_paddsw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_paddsw: ; SANDY: # BB#0: ; SANDY-NEXT: vpaddsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpaddsw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpaddsw (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_paddsw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpaddsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddsw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpaddsw (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_paddsw: ; BTVER2: # BB#0: @@ -2876,14 +2876,14 @@ define <16 x i8> @test_paddusb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_paddusb: ; SANDY: # BB#0: ; SANDY-NEXT: vpaddusb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpaddusb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpaddusb (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_paddusb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpaddusb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddusb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpaddusb (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_paddusb: ; BTVER2: # BB#0: @@ -2923,14 +2923,14 @@ define <8 x i16> @test_paddusw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_paddusw: ; SANDY: # BB#0: ; SANDY-NEXT: vpaddusw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpaddusw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpaddusw (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_paddusw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpaddusw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddusw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpaddusw (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_paddusw: ; BTVER2: # BB#0: @@ -2970,14 +2970,14 @@ define <8 x i16> @test_paddw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_paddw: ; SANDY: # BB#0: ; SANDY-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpaddw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpaddw (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_paddw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpaddw (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_paddw: ; BTVER2: # BB#0: @@ -3015,16 +3015,16 @@ define <2 x i64> @test_pand(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) { ; SANDY-LABEL: test_pand: ; SANDY: # BB#0: ; SANDY-NEXT: vpand %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: vpand (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; SANDY-NEXT: vpand (%rdi), %xmm0, %xmm0 # sched: [7:0.50] ; SANDY-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pand: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpand %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; HASWELL-NEXT: vpand (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vpand (%rdi), %xmm0, %xmm0 # sched: [1:0.50] ; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pand: ; BTVER2: # BB#0: @@ -3070,16 +3070,16 @@ define <2 x i64> @test_pandn(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) { ; SANDY-LABEL: test_pandn: ; SANDY: # BB#0: ; SANDY-NEXT: vpandn %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: vpandn (%rdi), %xmm0, %xmm1 # sched: [5:0.50] +; SANDY-NEXT: vpandn (%rdi), %xmm0, %xmm1 # sched: [7:0.50] ; SANDY-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pandn: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpandn %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; HASWELL-NEXT: vpandn (%rdi), %xmm0, %xmm1 # sched: [5:0.50] +; HASWELL-NEXT: vpandn (%rdi), %xmm0, %xmm1 # sched: [1:0.50] ; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pandn: ; BTVER2: # BB#0: @@ -3122,14 +3122,14 @@ define <16 x i8> @test_pavgb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_pavgb: ; SANDY: # BB#0: ; SANDY-NEXT: vpavgb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpavgb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpavgb (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pavgb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpavgb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpavgb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpavgb (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pavgb: ; BTVER2: # BB#0: @@ -3169,14 +3169,14 @@ define <8 x i16> @test_pavgw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_pavgw: ; SANDY: # BB#0: ; SANDY-NEXT: vpavgw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpavgw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpavgw (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pavgw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpavgw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpavgw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpavgw (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pavgw: ; BTVER2: # BB#0: @@ -3217,16 +3217,16 @@ define <16 x i8> @test_pcmpeqb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_pcmpeqb: ; SANDY: # BB#0: ; SANDY-NEXT: vpcmpeqb %xmm1, %xmm0, %xmm1 # sched: [1:0.50] -; SANDY-NEXT: vpcmpeqb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; SANDY-NEXT: vpcmpeqb (%rdi), %xmm0, %xmm0 # sched: [7:0.50] ; SANDY-NEXT: vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pcmpeqb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpcmpeqb %xmm1, %xmm0, %xmm1 # sched: [1:0.50] -; HASWELL-NEXT: vpcmpeqb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vpcmpeqb (%rdi), %xmm0, %xmm0 # sched: [1:0.50] ; HASWELL-NEXT: vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pcmpeqb: ; BTVER2: # BB#0: @@ -3269,16 +3269,16 @@ define <4 x i32> @test_pcmpeqd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; SANDY-LABEL: test_pcmpeqd: ; SANDY: # BB#0: ; SANDY-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm1 # sched: [1:0.50] -; SANDY-NEXT: vpcmpeqd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; SANDY-NEXT: vpcmpeqd (%rdi), %xmm0, %xmm0 # sched: [7:0.50] ; SANDY-NEXT: vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pcmpeqd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm1 # sched: [1:0.50] -; HASWELL-NEXT: vpcmpeqd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vpcmpeqd (%rdi), %xmm0, %xmm0 # sched: [1:0.50] ; HASWELL-NEXT: vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pcmpeqd: ; BTVER2: # BB#0: @@ -3321,16 +3321,16 @@ define <8 x i16> @test_pcmpeqw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_pcmpeqw: ; SANDY: # BB#0: ; SANDY-NEXT: vpcmpeqw %xmm1, %xmm0, %xmm1 # sched: [1:0.50] -; SANDY-NEXT: vpcmpeqw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; SANDY-NEXT: vpcmpeqw (%rdi), %xmm0, %xmm0 # sched: [7:0.50] ; SANDY-NEXT: vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pcmpeqw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpcmpeqw %xmm1, %xmm0, %xmm1 # sched: [1:0.50] -; HASWELL-NEXT: vpcmpeqw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vpcmpeqw (%rdi), %xmm0, %xmm0 # sched: [1:0.50] ; HASWELL-NEXT: vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pcmpeqw: ; BTVER2: # BB#0: @@ -3374,16 +3374,16 @@ define <16 x i8> @test_pcmpgtb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_pcmpgtb: ; SANDY: # BB#0: ; SANDY-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm1 # sched: [1:0.50] -; SANDY-NEXT: vpcmpgtb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; SANDY-NEXT: vpcmpgtb (%rdi), %xmm0, %xmm0 # sched: [7:0.50] ; SANDY-NEXT: vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pcmpgtb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm1 # sched: [1:0.50] -; HASWELL-NEXT: vpcmpgtb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vpcmpgtb (%rdi), %xmm0, %xmm0 # sched: [1:0.50] ; HASWELL-NEXT: vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pcmpgtb: ; BTVER2: # BB#0: @@ -3427,16 +3427,16 @@ define <4 x i32> @test_pcmpgtd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; SANDY-LABEL: test_pcmpgtd: ; SANDY: # BB#0: ; SANDY-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm1 # sched: [1:0.50] -; SANDY-NEXT: vpcmpeqd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; SANDY-NEXT: vpcmpeqd (%rdi), %xmm0, %xmm0 # sched: [7:0.50] ; SANDY-NEXT: vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pcmpgtd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm1 # sched: [1:0.50] -; HASWELL-NEXT: vpcmpeqd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vpcmpeqd (%rdi), %xmm0, %xmm0 # sched: [1:0.50] ; HASWELL-NEXT: vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pcmpgtd: ; BTVER2: # BB#0: @@ -3480,16 +3480,16 @@ define <8 x i16> @test_pcmpgtw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_pcmpgtw: ; SANDY: # BB#0: ; SANDY-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm1 # sched: [1:0.50] -; SANDY-NEXT: vpcmpgtw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; SANDY-NEXT: vpcmpgtw (%rdi), %xmm0, %xmm0 # sched: [7:0.50] ; SANDY-NEXT: vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pcmpgtw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm1 # sched: [1:0.50] -; HASWELL-NEXT: vpcmpgtw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vpcmpgtw (%rdi), %xmm0, %xmm0 # sched: [1:0.50] ; HASWELL-NEXT: vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pcmpgtw: ; BTVER2: # BB#0: @@ -3526,15 +3526,15 @@ define i16 @test_pextrw(<8 x i16> %a0) { ; ; SANDY-LABEL: test_pextrw: ; SANDY: # BB#0: -; SANDY-NEXT: vpextrw $6, %xmm0, %eax # sched: [1:0.50] +; SANDY-NEXT: vpextrw $6, %xmm0, %eax # sched: [3:1.00] ; SANDY-NEXT: # kill: %AX<def> %AX<kill> %EAX<kill> -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pextrw: ; HASWELL: # BB#0: -; HASWELL-NEXT: vpextrw $6, %xmm0, %eax # sched: [1:1.00] +; HASWELL-NEXT: vpextrw $6, %xmm0, %eax # sched: [2:1.00] ; HASWELL-NEXT: # kill: %AX<def> %AX<kill> %EAX<kill> -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pextrw: ; BTVER2: # BB#0: @@ -3570,15 +3570,15 @@ define <8 x i16> @test_pinsrw(<8 x i16> %a0, i16 %a1, i16 *%a2) { ; ; SANDY-LABEL: test_pinsrw: ; SANDY: # BB#0: -; SANDY-NEXT: vpinsrw $1, %edi, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpinsrw $3, (%rsi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpinsrw $1, %edi, %xmm0, %xmm0 # sched: [2:1.00] +; SANDY-NEXT: vpinsrw $3, (%rsi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pinsrw: ; HASWELL: # BB#0: -; HASWELL-NEXT: vpinsrw $1, %edi, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpinsrw $3, (%rsi), %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpinsrw $1, %edi, %xmm0, %xmm0 # sched: [2:2.00] +; HASWELL-NEXT: vpinsrw $3, (%rsi), %xmm0, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pinsrw: ; BTVER2: # BB#0: @@ -3620,15 +3620,15 @@ define <4 x i32> @test_pmaddwd(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; ; SANDY-LABEL: test_pmaddwd: ; SANDY: # BB#0: -; SANDY-NEXT: vpmaddwd %xmm1, %xmm0, %xmm0 # sched: [5:1.00] +; SANDY-NEXT: vpmaddwd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vpmaddwd (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmaddwd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmaddwd %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: vpmaddwd (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpmaddwd (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmaddwd: ; BTVER2: # BB#0: @@ -3669,14 +3669,14 @@ define <8 x i16> @test_pmaxsw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_pmaxsw: ; SANDY: # BB#0: ; SANDY-NEXT: vpmaxsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpmaxsw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpmaxsw (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmaxsw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmaxsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpmaxsw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpmaxsw (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmaxsw: ; BTVER2: # BB#0: @@ -3716,14 +3716,14 @@ define <16 x i8> @test_pmaxub(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_pmaxub: ; SANDY: # BB#0: ; SANDY-NEXT: vpmaxub %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpmaxub (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpmaxub (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmaxub: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmaxub %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpmaxub (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpmaxub (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmaxub: ; BTVER2: # BB#0: @@ -3763,14 +3763,14 @@ define <8 x i16> @test_pminsw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_pminsw: ; SANDY: # BB#0: ; SANDY-NEXT: vpminsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpminsw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpminsw (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pminsw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpminsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpminsw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpminsw (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pminsw: ; BTVER2: # BB#0: @@ -3810,14 +3810,14 @@ define <16 x i8> @test_pminub(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_pminub: ; SANDY: # BB#0: ; SANDY-NEXT: vpminub %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpminub (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpminub (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pminub: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpminub %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpminub (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpminub (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pminub: ; BTVER2: # BB#0: @@ -3851,13 +3851,13 @@ define i32 @test_pmovmskb(<16 x i8> %a0) { ; ; SANDY-LABEL: test_pmovmskb: ; SANDY: # BB#0: -; SANDY-NEXT: vpmovmskb %xmm0, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpmovmskb %xmm0, %eax # sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmovmskb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmovmskb %xmm0, %eax # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmovmskb: ; BTVER2: # BB#0: @@ -3891,13 +3891,13 @@ define <8 x i16> @test_pmulhuw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY: # BB#0: ; SANDY-NEXT: vpmulhuw %xmm1, %xmm0, %xmm0 # sched: [5:1.00] ; SANDY-NEXT: vpmulhuw (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmulhuw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmulhuw %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: vpmulhuw (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpmulhuw (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmulhuw: ; BTVER2: # BB#0: @@ -3932,15 +3932,15 @@ define <8 x i16> @test_pmulhw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; ; SANDY-LABEL: test_pmulhw: ; SANDY: # BB#0: -; SANDY-NEXT: vpmulhw %xmm1, %xmm0, %xmm0 # sched: [5:1.00] +; SANDY-NEXT: vpmulhw %xmm1, %xmm0, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vpmulhw (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmulhw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmulhw %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: vpmulhw (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpmulhw (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmulhw: ; BTVER2: # BB#0: @@ -3975,15 +3975,15 @@ define <8 x i16> @test_pmullw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; ; SANDY-LABEL: test_pmullw: ; SANDY: # BB#0: -; SANDY-NEXT: vpmullw %xmm1, %xmm0, %xmm0 # sched: [5:1.00] +; SANDY-NEXT: vpmullw %xmm1, %xmm0, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vpmullw (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmullw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmullw %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: vpmullw (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpmullw (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmullw: ; BTVER2: # BB#0: @@ -4027,13 +4027,13 @@ define <2 x i64> @test_pmuludq(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; SANDY: # BB#0: ; SANDY-NEXT: vpmuludq %xmm1, %xmm0, %xmm0 # sched: [5:1.00] ; SANDY-NEXT: vpmuludq (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmuludq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmuludq %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: vpmuludq (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpmuludq (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmuludq: ; BTVER2: # BB#0: @@ -4073,16 +4073,16 @@ define <2 x i64> @test_por(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) { ; SANDY-LABEL: test_por: ; SANDY: # BB#0: ; SANDY-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: vpor (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; SANDY-NEXT: vpor (%rdi), %xmm0, %xmm0 # sched: [7:0.50] ; SANDY-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_por: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; HASWELL-NEXT: vpor (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vpor (%rdi), %xmm0, %xmm0 # sched: [1:0.50] ; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_por: ; BTVER2: # BB#0: @@ -4126,15 +4126,15 @@ define <2 x i64> @test_psadbw(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; ; SANDY-LABEL: test_psadbw: ; SANDY: # BB#0: -; SANDY-NEXT: vpsadbw %xmm1, %xmm0, %xmm0 # sched: [5:1.00] +; SANDY-NEXT: vpsadbw %xmm1, %xmm0, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vpsadbw (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psadbw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsadbw %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: vpsadbw (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpsadbw (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psadbw: ; BTVER2: # BB#0: @@ -4176,16 +4176,16 @@ define <4 x i32> @test_pshufd(<4 x i32> %a0, <4 x i32> *%a1) { ; SANDY-LABEL: test_pshufd: ; SANDY: # BB#0: ; SANDY-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,0,3,2] sched: [1:0.50] -; SANDY-NEXT: vpshufd {{.*#+}} xmm1 = mem[3,2,1,0] sched: [5:0.50] +; SANDY-NEXT: vpshufd {{.*#+}} xmm1 = mem[3,2,1,0] sched: [7:0.50] ; SANDY-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pshufd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,0,3,2] sched: [1:1.00] -; HASWELL-NEXT: vpshufd {{.*#+}} xmm1 = mem[3,2,1,0] sched: [5:1.00] +; HASWELL-NEXT: vpshufd {{.*#+}} xmm1 = mem[3,2,1,0] sched: [1:1.00] ; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pshufd: ; BTVER2: # BB#0: @@ -4227,16 +4227,16 @@ define <8 x i16> @test_pshufhw(<8 x i16> %a0, <8 x i16> *%a1) { ; SANDY-LABEL: test_pshufhw: ; SANDY: # BB#0: ; SANDY-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,6] sched: [1:0.50] -; SANDY-NEXT: vpshufhw {{.*#+}} xmm1 = mem[0,1,2,3,7,6,5,4] sched: [5:0.50] +; SANDY-NEXT: vpshufhw {{.*#+}} xmm1 = mem[0,1,2,3,7,6,5,4] sched: [7:0.50] ; SANDY-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pshufhw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,6] sched: [1:1.00] -; HASWELL-NEXT: vpshufhw {{.*#+}} xmm1 = mem[0,1,2,3,7,6,5,4] sched: [5:1.00] +; HASWELL-NEXT: vpshufhw {{.*#+}} xmm1 = mem[0,1,2,3,7,6,5,4] sched: [1:1.00] ; HASWELL-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pshufhw: ; BTVER2: # BB#0: @@ -4278,16 +4278,16 @@ define <8 x i16> @test_pshuflw(<8 x i16> %a0, <8 x i16> *%a1) { ; SANDY-LABEL: test_pshuflw: ; SANDY: # BB#0: ; SANDY-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7] sched: [1:0.50] -; SANDY-NEXT: vpshuflw {{.*#+}} xmm1 = mem[3,2,1,0,4,5,6,7] sched: [5:0.50] +; SANDY-NEXT: vpshuflw {{.*#+}} xmm1 = mem[3,2,1,0,4,5,6,7] sched: [7:0.50] ; SANDY-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pshuflw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7] sched: [1:1.00] -; HASWELL-NEXT: vpshuflw {{.*#+}} xmm1 = mem[3,2,1,0,4,5,6,7] sched: [5:1.00] +; HASWELL-NEXT: vpshuflw {{.*#+}} xmm1 = mem[3,2,1,0,4,5,6,7] sched: [1:1.00] ; HASWELL-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pshuflw: ; BTVER2: # BB#0: @@ -4328,15 +4328,15 @@ define <4 x i32> @test_pslld(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; SANDY: # BB#0: ; SANDY-NEXT: vpslld %xmm1, %xmm0, %xmm0 # sched: [1:0.50] ; SANDY-NEXT: vpslld (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: vpslld $2, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpslld $2, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pslld: ; HASWELL: # BB#0: -; HASWELL-NEXT: vpslld %xmm1, %xmm0, %xmm0 # sched: [2:1.00] -; HASWELL-NEXT: vpslld (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: vpslld %xmm1, %xmm0, %xmm0 # sched: [4:1.00] +; HASWELL-NEXT: vpslld (%rdi), %xmm0, %xmm0 # sched: [1:1.00] ; HASWELL-NEXT: vpslld $2, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pslld: ; BTVER2: # BB#0: @@ -4378,12 +4378,12 @@ define <4 x i32> @test_pslldq(<4 x i32> %a0) { ; SANDY-LABEL: test_pslldq: ; SANDY: # BB#0: ; SANDY-NEXT: vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11] sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pslldq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11] sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pslldq: ; BTVER2: # BB#0: @@ -4419,15 +4419,15 @@ define <2 x i64> @test_psllq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) { ; SANDY: # BB#0: ; SANDY-NEXT: vpsllq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] ; SANDY-NEXT: vpsllq (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: vpsllq $2, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsllq $2, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psllq: ; HASWELL: # BB#0: -; HASWELL-NEXT: vpsllq %xmm1, %xmm0, %xmm0 # sched: [2:1.00] -; HASWELL-NEXT: vpsllq (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: vpsllq %xmm1, %xmm0, %xmm0 # sched: [4:1.00] +; HASWELL-NEXT: vpsllq (%rdi), %xmm0, %xmm0 # sched: [1:1.00] ; HASWELL-NEXT: vpsllq $2, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psllq: ; BTVER2: # BB#0: @@ -4470,15 +4470,15 @@ define <8 x i16> @test_psllw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY: # BB#0: ; SANDY-NEXT: vpsllw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] ; SANDY-NEXT: vpsllw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: vpsllw $2, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsllw $2, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psllw: ; HASWELL: # BB#0: -; HASWELL-NEXT: vpsllw %xmm1, %xmm0, %xmm0 # sched: [2:1.00] -; HASWELL-NEXT: vpsllw (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: vpsllw %xmm1, %xmm0, %xmm0 # sched: [4:1.00] +; HASWELL-NEXT: vpsllw (%rdi), %xmm0, %xmm0 # sched: [1:1.00] ; HASWELL-NEXT: vpsllw $2, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psllw: ; BTVER2: # BB#0: @@ -4519,17 +4519,17 @@ define <4 x i32> @test_psrad(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; ; SANDY-LABEL: test_psrad: ; SANDY: # BB#0: -; SANDY-NEXT: vpsrad %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpsrad (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: vpsrad $2, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsrad %xmm1, %xmm0, %xmm0 # sched: [2:1.00] +; SANDY-NEXT: vpsrad (%rdi), %xmm0, %xmm0 # sched: [8:1.00] +; SANDY-NEXT: vpsrad $2, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psrad: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsrad %xmm1, %xmm0, %xmm0 # sched: [2:1.00] -; HASWELL-NEXT: vpsrad (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: vpsrad (%rdi), %xmm0, %xmm0 # sched: [2:1.00] ; HASWELL-NEXT: vpsrad $2, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psrad: ; BTVER2: # BB#0: @@ -4570,17 +4570,17 @@ define <8 x i16> @test_psraw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; ; SANDY-LABEL: test_psraw: ; SANDY: # BB#0: -; SANDY-NEXT: vpsraw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpsraw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: vpsraw $2, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsraw %xmm1, %xmm0, %xmm0 # sched: [2:1.00] +; SANDY-NEXT: vpsraw (%rdi), %xmm0, %xmm0 # sched: [8:1.00] +; SANDY-NEXT: vpsraw $2, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psraw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsraw %xmm1, %xmm0, %xmm0 # sched: [2:1.00] -; HASWELL-NEXT: vpsraw (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: vpsraw (%rdi), %xmm0, %xmm0 # sched: [2:1.00] ; HASWELL-NEXT: vpsraw $2, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psraw: ; BTVER2: # BB#0: @@ -4621,17 +4621,17 @@ define <4 x i32> @test_psrld(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; ; SANDY-LABEL: test_psrld: ; SANDY: # BB#0: -; SANDY-NEXT: vpsrld %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpsrld (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: vpsrld $2, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsrld %xmm1, %xmm0, %xmm0 # sched: [2:1.00] +; SANDY-NEXT: vpsrld (%rdi), %xmm0, %xmm0 # sched: [8:1.00] +; SANDY-NEXT: vpsrld $2, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psrld: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsrld %xmm1, %xmm0, %xmm0 # sched: [2:1.00] -; HASWELL-NEXT: vpsrld (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: vpsrld (%rdi), %xmm0, %xmm0 # sched: [2:1.00] ; HASWELL-NEXT: vpsrld $2, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psrld: ; BTVER2: # BB#0: @@ -4673,12 +4673,12 @@ define <4 x i32> @test_psrldq(<4 x i32> %a0) { ; SANDY-LABEL: test_psrldq: ; SANDY: # BB#0: ; SANDY-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psrldq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psrldq: ; BTVER2: # BB#0: @@ -4712,17 +4712,17 @@ define <2 x i64> @test_psrlq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) { ; ; SANDY-LABEL: test_psrlq: ; SANDY: # BB#0: -; SANDY-NEXT: vpsrlq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpsrlq (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: vpsrlq $2, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsrlq %xmm1, %xmm0, %xmm0 # sched: [2:1.00] +; SANDY-NEXT: vpsrlq (%rdi), %xmm0, %xmm0 # sched: [8:1.00] +; SANDY-NEXT: vpsrlq $2, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psrlq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsrlq %xmm1, %xmm0, %xmm0 # sched: [2:1.00] -; HASWELL-NEXT: vpsrlq (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: vpsrlq (%rdi), %xmm0, %xmm0 # sched: [2:1.00] ; HASWELL-NEXT: vpsrlq $2, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psrlq: ; BTVER2: # BB#0: @@ -4763,17 +4763,17 @@ define <8 x i16> @test_psrlw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; ; SANDY-LABEL: test_psrlw: ; SANDY: # BB#0: -; SANDY-NEXT: vpsrlw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpsrlw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: vpsrlw $2, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsrlw %xmm1, %xmm0, %xmm0 # sched: [2:1.00] +; SANDY-NEXT: vpsrlw (%rdi), %xmm0, %xmm0 # sched: [8:1.00] +; SANDY-NEXT: vpsrlw $2, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psrlw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsrlw %xmm1, %xmm0, %xmm0 # sched: [2:1.00] -; HASWELL-NEXT: vpsrlw (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: vpsrlw (%rdi), %xmm0, %xmm0 # sched: [2:1.00] ; HASWELL-NEXT: vpsrlw $2, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psrlw: ; BTVER2: # BB#0: @@ -4816,14 +4816,14 @@ define <16 x i8> @test_psubb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_psubb: ; SANDY: # BB#0: ; SANDY-NEXT: vpsubb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpsubb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsubb (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psubb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsubb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpsubb (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psubb: ; BTVER2: # BB#0: @@ -4862,14 +4862,14 @@ define <4 x i32> @test_psubd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; SANDY-LABEL: test_psubd: ; SANDY: # BB#0: ; SANDY-NEXT: vpsubd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpsubd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsubd (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psubd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsubd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpsubd (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psubd: ; BTVER2: # BB#0: @@ -4904,14 +4904,14 @@ define <2 x i64> @test_psubq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) { ; SANDY-LABEL: test_psubq: ; SANDY: # BB#0: ; SANDY-NEXT: vpsubq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpsubq (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsubq (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psubq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsubq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubq (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpsubq (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psubq: ; BTVER2: # BB#0: @@ -4950,14 +4950,14 @@ define <16 x i8> @test_psubsb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_psubsb: ; SANDY: # BB#0: ; SANDY-NEXT: vpsubsb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpsubsb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsubsb (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psubsb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsubsb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubsb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpsubsb (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psubsb: ; BTVER2: # BB#0: @@ -4997,14 +4997,14 @@ define <8 x i16> @test_psubsw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_psubsw: ; SANDY: # BB#0: ; SANDY-NEXT: vpsubsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpsubsw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsubsw (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psubsw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsubsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubsw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpsubsw (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psubsw: ; BTVER2: # BB#0: @@ -5044,14 +5044,14 @@ define <16 x i8> @test_psubusb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_psubusb: ; SANDY: # BB#0: ; SANDY-NEXT: vpsubusb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpsubusb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsubusb (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psubusb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsubusb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubusb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpsubusb (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psubusb: ; BTVER2: # BB#0: @@ -5091,14 +5091,14 @@ define <8 x i16> @test_psubusw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_psubusw: ; SANDY: # BB#0: ; SANDY-NEXT: vpsubusw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpsubusw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsubusw (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psubusw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsubusw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubusw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpsubusw (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psubusw: ; BTVER2: # BB#0: @@ -5138,14 +5138,14 @@ define <8 x i16> @test_psubw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_psubw: ; SANDY: # BB#0: ; SANDY-NEXT: vpsubw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpsubw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsubw (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psubw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsubw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpsubw (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psubw: ; BTVER2: # BB#0: @@ -5184,14 +5184,14 @@ define <16 x i8> @test_punpckhbw(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_punpckhbw: ; SANDY: # BB#0: ; SANDY-NEXT: vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15] sched: [1:0.50] -; SANDY-NEXT: vpunpckhbw {{.*#+}} xmm0 = xmm0[8],mem[8],xmm0[9],mem[9],xmm0[10],mem[10],xmm0[11],mem[11],xmm0[12],mem[12],xmm0[13],mem[13],xmm0[14],mem[14],xmm0[15],mem[15] sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpunpckhbw {{.*#+}} xmm0 = xmm0[8],mem[8],xmm0[9],mem[9],xmm0[10],mem[10],xmm0[11],mem[11],xmm0[12],mem[12],xmm0[13],mem[13],xmm0[14],mem[14],xmm0[15],mem[15] sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_punpckhbw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15] sched: [1:1.00] -; HASWELL-NEXT: vpunpckhbw {{.*#+}} xmm0 = xmm0[8],mem[8],xmm0[9],mem[9],xmm0[10],mem[10],xmm0[11],mem[11],xmm0[12],mem[12],xmm0[13],mem[13],xmm0[14],mem[14],xmm0[15],mem[15] sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpunpckhbw {{.*#+}} xmm0 = xmm0[8],mem[8],xmm0[9],mem[9],xmm0[10],mem[10],xmm0[11],mem[11],xmm0[12],mem[12],xmm0[13],mem[13],xmm0[14],mem[14],xmm0[15],mem[15] sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_punpckhbw: ; BTVER2: # BB#0: @@ -5231,16 +5231,16 @@ define <4 x i32> @test_punpckhdq(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; SANDY-LABEL: test_punpckhdq: ; SANDY: # BB#0: ; SANDY-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3] sched: [1:0.50] -; SANDY-NEXT: vpunpckhdq {{.*#+}} xmm1 = xmm1[2],mem[2],xmm1[3],mem[3] sched: [5:0.50] +; SANDY-NEXT: vpunpckhdq {{.*#+}} xmm1 = xmm1[2],mem[2],xmm1[3],mem[3] sched: [7:0.50] ; SANDY-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_punpckhdq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3] sched: [1:1.00] -; HASWELL-NEXT: vpunpckhdq {{.*#+}} xmm1 = xmm1[2],mem[2],xmm1[3],mem[3] sched: [5:1.00] +; HASWELL-NEXT: vpunpckhdq {{.*#+}} xmm1 = xmm1[2],mem[2],xmm1[3],mem[3] sched: [1:1.00] ; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_punpckhdq: ; BTVER2: # BB#0: @@ -5280,16 +5280,16 @@ define <2 x i64> @test_punpckhqdq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) ; SANDY-LABEL: test_punpckhqdq: ; SANDY: # BB#0: ; SANDY-NEXT: vpunpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1] sched: [1:0.50] -; SANDY-NEXT: vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],mem[1] sched: [5:0.50] +; SANDY-NEXT: vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],mem[1] sched: [7:0.50] ; SANDY-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_punpckhqdq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpunpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1] sched: [1:1.00] -; HASWELL-NEXT: vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],mem[1] sched: [5:1.00] +; HASWELL-NEXT: vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],mem[1] sched: [1:1.00] ; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_punpckhqdq: ; BTVER2: # BB#0: @@ -5330,14 +5330,14 @@ define <8 x i16> @test_punpckhwd(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_punpckhwd: ; SANDY: # BB#0: ; SANDY-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7] sched: [1:0.50] -; SANDY-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7] sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7] sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_punpckhwd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7] sched: [1:1.00] -; HASWELL-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7] sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7] sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_punpckhwd: ; BTVER2: # BB#0: @@ -5376,14 +5376,14 @@ define <16 x i8> @test_punpcklbw(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_punpcklbw: ; SANDY: # BB#0: ; SANDY-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7] sched: [1:0.50] -; SANDY-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3],xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7] sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3],xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7] sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_punpcklbw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7] sched: [1:1.00] -; HASWELL-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3],xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7] sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3],xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7] sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_punpcklbw: ; BTVER2: # BB#0: @@ -5423,16 +5423,16 @@ define <4 x i32> @test_punpckldq(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; SANDY-LABEL: test_punpckldq: ; SANDY: # BB#0: ; SANDY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] sched: [1:0.50] -; SANDY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1] sched: [5:0.50] +; SANDY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1] sched: [7:0.50] ; SANDY-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_punpckldq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] sched: [1:1.00] -; HASWELL-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1] sched: [5:1.00] +; HASWELL-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1] sched: [1:1.00] ; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_punpckldq: ; BTVER2: # BB#0: @@ -5472,16 +5472,16 @@ define <2 x i64> @test_punpcklqdq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) ; SANDY-LABEL: test_punpcklqdq: ; SANDY: # BB#0: ; SANDY-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] sched: [1:0.50] -; SANDY-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [5:0.50] +; SANDY-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [7:0.50] ; SANDY-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_punpcklqdq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] sched: [1:1.00] -; HASWELL-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [5:1.00] +; HASWELL-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [1:1.00] ; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_punpcklqdq: ; BTVER2: # BB#0: @@ -5522,14 +5522,14 @@ define <8 x i16> @test_punpcklwd(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_punpcklwd: ; SANDY: # BB#0: ; SANDY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] sched: [1:0.50] -; SANDY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3] sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3] sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_punpcklwd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] sched: [1:1.00] -; HASWELL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3] sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3] sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_punpcklwd: ; BTVER2: # BB#0: @@ -5567,16 +5567,16 @@ define <2 x i64> @test_pxor(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) { ; SANDY-LABEL: test_pxor: ; SANDY: # BB#0: ; SANDY-NEXT: vpxor %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: vpxor (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; SANDY-NEXT: vpxor (%rdi), %xmm0, %xmm0 # sched: [7:0.50] ; SANDY-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pxor: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpxor %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; HASWELL-NEXT: vpxor (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; HASWELL-NEXT: vpxor (%rdi), %xmm0, %xmm0 # sched: [1:0.50] ; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pxor: ; BTVER2: # BB#0: @@ -5616,16 +5616,16 @@ define <2 x double> @test_shufpd(<2 x double> %a0, <2 x double> %a1, <2 x double ; SANDY-LABEL: test_shufpd: ; SANDY: # BB#0: ; SANDY-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1],xmm1[0] sched: [1:1.00] -; SANDY-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1],mem[0] sched: [5:1.00] +; SANDY-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1],mem[0] sched: [7:1.00] ; SANDY-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_shufpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1],xmm1[0] sched: [1:1.00] -; HASWELL-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1],mem[0] sched: [5:1.00] +; HASWELL-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1],mem[0] sched: [1:1.00] ; HASWELL-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_shufpd: ; BTVER2: # BB#0: @@ -5665,17 +5665,17 @@ define <2 x double> @test_sqrtpd(<2 x double> %a0, <2 x double> *%a1) { ; ; SANDY-LABEL: test_sqrtpd: ; SANDY: # BB#0: -; SANDY-NEXT: vsqrtpd %xmm0, %xmm0 # sched: [15:1.00] -; SANDY-NEXT: vsqrtpd (%rdi), %xmm1 # sched: [19:1.00] +; SANDY-NEXT: vsqrtpd %xmm0, %xmm0 # sched: [22:1.00] +; SANDY-NEXT: vsqrtpd (%rdi), %xmm1 # sched: [28:1.00] ; SANDY-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_sqrtpd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vsqrtpd %xmm0, %xmm0 # sched: [15:1.00] -; HASWELL-NEXT: vsqrtpd (%rdi), %xmm1 # sched: [19:1.00] +; HASWELL-NEXT: vsqrtpd %xmm0, %xmm0 # sched: [21:1.00] +; HASWELL-NEXT: vsqrtpd (%rdi), %xmm1 # sched: [21:1.00] ; HASWELL-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_sqrtpd: ; BTVER2: # BB#0: @@ -5720,19 +5720,19 @@ define <2 x double> @test_sqrtsd(<2 x double> %a0, <2 x double> *%a1) { ; ; SANDY-LABEL: test_sqrtsd: ; SANDY: # BB#0: -; SANDY-NEXT: vsqrtsd %xmm0, %xmm0, %xmm0 # sched: [19:1.00] -; SANDY-NEXT: vmovapd (%rdi), %xmm1 # sched: [4:0.50] -; SANDY-NEXT: vsqrtsd %xmm1, %xmm1, %xmm1 # sched: [19:1.00] +; SANDY-NEXT: vsqrtsd %xmm0, %xmm0, %xmm0 # sched: [21:1.00] +; SANDY-NEXT: vmovapd (%rdi), %xmm1 # sched: [6:0.50] +; SANDY-NEXT: vsqrtsd %xmm1, %xmm1, %xmm1 # sched: [21:1.00] ; SANDY-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_sqrtsd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vsqrtsd %xmm0, %xmm0, %xmm0 # sched: [19:1.00] -; HASWELL-NEXT: vmovapd (%rdi), %xmm1 # sched: [4:0.50] -; HASWELL-NEXT: vsqrtsd %xmm1, %xmm1, %xmm1 # sched: [19:1.00] +; HASWELL-NEXT: vsqrtsd %xmm0, %xmm0, %xmm0 # sched: [21:1.00] +; HASWELL-NEXT: vmovapd (%rdi), %xmm1 # sched: [?:5.000000e-01] +; HASWELL-NEXT: vsqrtsd %xmm1, %xmm1, %xmm1 # sched: [21:1.00] ; HASWELL-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_sqrtsd: ; BTVER2: # BB#0: @@ -5771,14 +5771,14 @@ define <2 x double> @test_subpd(<2 x double> %a0, <2 x double> %a1, <2 x double> ; SANDY-LABEL: test_subpd: ; SANDY: # BB#0: ; SANDY-NEXT: vsubpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vsubpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vsubpd (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_subpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vsubpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vsubpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vsubpd (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_subpd: ; BTVER2: # BB#0: @@ -5813,14 +5813,14 @@ define double @test_subsd(double %a0, double %a1, double *%a2) { ; SANDY-LABEL: test_subsd: ; SANDY: # BB#0: ; SANDY-NEXT: vsubsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vsubsd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vsubsd (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_subsd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vsubsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vsubsd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vsubsd (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_subsd: ; BTVER2: # BB#0: @@ -5879,30 +5879,30 @@ define i32 @test_ucomisd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) ; SANDY-LABEL: test_ucomisd: ; SANDY: # BB#0: ; SANDY-NEXT: vucomisd %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: setnp %al # sched: [1:0.33] -; SANDY-NEXT: sete %cl # sched: [1:0.33] +; SANDY-NEXT: setnp %al # sched: [1:1.00] +; SANDY-NEXT: sete %cl # sched: [1:1.00] ; SANDY-NEXT: andb %al, %cl # sched: [1:0.33] ; SANDY-NEXT: vucomisd (%rdi), %xmm0 # sched: [7:1.00] -; SANDY-NEXT: setnp %al # sched: [1:0.33] -; SANDY-NEXT: sete %dl # sched: [1:0.33] +; SANDY-NEXT: setnp %al # sched: [1:1.00] +; SANDY-NEXT: sete %dl # sched: [1:1.00] ; SANDY-NEXT: andb %al, %dl # sched: [1:0.33] ; SANDY-NEXT: orb %cl, %dl # sched: [1:0.33] ; SANDY-NEXT: movzbl %dl, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_ucomisd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vucomisd %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: setnp %al # sched: [1:0.50] -; HASWELL-NEXT: sete %cl # sched: [1:0.50] +; HASWELL-NEXT: setnp %al # sched: [1:1.00] +; HASWELL-NEXT: sete %cl # sched: [1:1.00] ; HASWELL-NEXT: andb %al, %cl # sched: [1:0.25] ; HASWELL-NEXT: vucomisd (%rdi), %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: setnp %al # sched: [1:0.50] -; HASWELL-NEXT: sete %dl # sched: [1:0.50] +; HASWELL-NEXT: setnp %al # sched: [1:1.00] +; HASWELL-NEXT: sete %dl # sched: [1:1.00] ; HASWELL-NEXT: andb %al, %dl # sched: [1:0.25] ; HASWELL-NEXT: orb %cl, %dl # sched: [1:0.25] ; HASWELL-NEXT: movzbl %dl, %eax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_ucomisd: ; BTVER2: # BB#0: @@ -5950,16 +5950,16 @@ define <2 x double> @test_unpckhpd(<2 x double> %a0, <2 x double> %a1, <2 x doub ; SANDY-LABEL: test_unpckhpd: ; SANDY: # BB#0: ; SANDY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1] sched: [1:1.00] -; SANDY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1] sched: [5:1.00] +; SANDY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1] sched: [7:1.00] ; SANDY-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_unpckhpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1] sched: [1:1.00] -; HASWELL-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1] sched: [5:1.00] +; HASWELL-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1] sched: [1:1.00] ; HASWELL-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_unpckhpd: ; BTVER2: # BB#0: @@ -6005,16 +6005,16 @@ define <2 x double> @test_unpcklpd(<2 x double> %a0, <2 x double> %a1, <2 x doub ; SANDY-LABEL: test_unpcklpd: ; SANDY: # BB#0: ; SANDY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0] sched: [1:1.00] -; SANDY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0] sched: [5:1.00] +; SANDY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0] sched: [7:1.00] ; SANDY-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_unpcklpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0] sched: [1:1.00] -; HASWELL-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0] sched: [5:1.00] +; HASWELL-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0] sched: [1:1.00] ; HASWELL-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_unpcklpd: ; BTVER2: # BB#0: @@ -6053,17 +6053,17 @@ define <2 x double> @test_xorpd(<2 x double> %a0, <2 x double> %a1, <2 x double> ; ; SANDY-LABEL: test_xorpd: ; SANDY: # BB#0: -; SANDY-NEXT: vxorpd %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: vxorpd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] +; SANDY-NEXT: vxorpd %xmm1, %xmm0, %xmm0 # sched: [1:1.00] +; SANDY-NEXT: vxorpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] ; SANDY-NEXT: vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_xorpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vxorpd %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vxorpd (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: vxorpd (%rdi), %xmm0, %xmm0 # sched: [1:1.00] ; HASWELL-NEXT: vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_xorpd: ; BTVER2: # BB#0: diff --git a/llvm/test/CodeGen/X86/sse3-schedule.ll b/llvm/test/CodeGen/X86/sse3-schedule.ll index 482b2fcab64..672ae838a04 100644 --- a/llvm/test/CodeGen/X86/sse3-schedule.ll +++ b/llvm/test/CodeGen/X86/sse3-schedule.ll @@ -31,14 +31,14 @@ define <2 x double> @test_addsubpd(<2 x double> %a0, <2 x double> %a1, <2 x doub ; SANDY-LABEL: test_addsubpd: ; SANDY: # BB#0: ; SANDY-NEXT: vaddsubpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vaddsubpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vaddsubpd (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_addsubpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vaddsubpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vaddsubpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vaddsubpd (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_addsubpd: ; BTVER2: # BB#0: @@ -74,14 +74,14 @@ define <4 x float> @test_addsubps(<4 x float> %a0, <4 x float> %a1, <4 x float> ; SANDY-LABEL: test_addsubps: ; SANDY: # BB#0: ; SANDY-NEXT: vaddsubps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vaddsubps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vaddsubps (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_addsubps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vaddsubps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vaddsubps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vaddsubps (%rdi), %xmm0, %xmm0 # sched: [3:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_addsubps: ; BTVER2: # BB#0: @@ -116,15 +116,15 @@ define <2 x double> @test_haddpd(<2 x double> %a0, <2 x double> %a1, <2 x double ; ; SANDY-LABEL: test_haddpd: ; SANDY: # BB#0: -; SANDY-NEXT: vhaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vhaddpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vhaddpd %xmm1, %xmm0, %xmm0 # sched: [5:2.00] +; SANDY-NEXT: vhaddpd (%rdi), %xmm0, %xmm0 # sched: [11:2.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_haddpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vhaddpd %xmm1, %xmm0, %xmm0 # sched: [5:2.00] -; HASWELL-NEXT: vhaddpd (%rdi), %xmm0, %xmm0 # sched: [9:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vhaddpd (%rdi), %xmm0, %xmm0 # sched: [5:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_haddpd: ; BTVER2: # BB#0: @@ -159,15 +159,15 @@ define <4 x float> @test_haddps(<4 x float> %a0, <4 x float> %a1, <4 x float> *% ; ; SANDY-LABEL: test_haddps: ; SANDY: # BB#0: -; SANDY-NEXT: vhaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vhaddps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vhaddps %xmm1, %xmm0, %xmm0 # sched: [5:2.00] +; SANDY-NEXT: vhaddps (%rdi), %xmm0, %xmm0 # sched: [11:2.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_haddps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vhaddps %xmm1, %xmm0, %xmm0 # sched: [5:2.00] -; HASWELL-NEXT: vhaddps (%rdi), %xmm0, %xmm0 # sched: [9:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vhaddps (%rdi), %xmm0, %xmm0 # sched: [5:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_haddps: ; BTVER2: # BB#0: @@ -202,15 +202,15 @@ define <2 x double> @test_hsubpd(<2 x double> %a0, <2 x double> %a1, <2 x double ; ; SANDY-LABEL: test_hsubpd: ; SANDY: # BB#0: -; SANDY-NEXT: vhsubpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vhsubpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vhsubpd %xmm1, %xmm0, %xmm0 # sched: [5:2.00] +; SANDY-NEXT: vhsubpd (%rdi), %xmm0, %xmm0 # sched: [11:2.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_hsubpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vhsubpd %xmm1, %xmm0, %xmm0 # sched: [5:2.00] -; HASWELL-NEXT: vhsubpd (%rdi), %xmm0, %xmm0 # sched: [9:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vhsubpd (%rdi), %xmm0, %xmm0 # sched: [5:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_hsubpd: ; BTVER2: # BB#0: @@ -245,15 +245,15 @@ define <4 x float> @test_hsubps(<4 x float> %a0, <4 x float> %a1, <4 x float> *% ; ; SANDY-LABEL: test_hsubps: ; SANDY: # BB#0: -; SANDY-NEXT: vhsubps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vhsubps (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vhsubps %xmm1, %xmm0, %xmm0 # sched: [5:2.00] +; SANDY-NEXT: vhsubps (%rdi), %xmm0, %xmm0 # sched: [11:2.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_hsubps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vhsubps %xmm1, %xmm0, %xmm0 # sched: [5:2.00] -; HASWELL-NEXT: vhsubps (%rdi), %xmm0, %xmm0 # sched: [9:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vhsubps (%rdi), %xmm0, %xmm0 # sched: [5:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_hsubps: ; BTVER2: # BB#0: @@ -287,13 +287,13 @@ define <16 x i8> @test_lddqu(i8* %a0) { ; ; SANDY-LABEL: test_lddqu: ; SANDY: # BB#0: -; SANDY-NEXT: vlddqu (%rdi), %xmm0 # sched: [4:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vlddqu (%rdi), %xmm0 # sched: [6:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_lddqu: ; HASWELL: # BB#0: -; HASWELL-NEXT: vlddqu (%rdi), %xmm0 # sched: [4:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vlddqu (%rdi), %xmm0 # sched: [?:5.000000e-01] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_lddqu: ; BTVER2: # BB#0: @@ -330,16 +330,16 @@ define <2 x double> @test_movddup(<2 x double> %a0, <2 x double> *%a1) { ; SANDY-LABEL: test_movddup: ; SANDY: # BB#0: ; SANDY-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0] sched: [1:1.00] -; SANDY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0] sched: [4:0.50] +; SANDY-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0] sched: [6:0.50] ; SANDY-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movddup: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0] sched: [1:1.00] -; HASWELL-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0] sched: [4:0.50] +; HASWELL-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0] sched: [?:5.000000e-01] ; HASWELL-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movddup: ; BTVER2: # BB#0: @@ -380,16 +380,16 @@ define <4 x float> @test_movshdup(<4 x float> %a0, <4 x float> *%a1) { ; SANDY-LABEL: test_movshdup: ; SANDY: # BB#0: ; SANDY-NEXT: vmovshdup {{.*#+}} xmm0 = xmm0[1,1,3,3] sched: [1:1.00] -; SANDY-NEXT: vmovshdup {{.*#+}} xmm1 = mem[1,1,3,3] sched: [4:0.50] +; SANDY-NEXT: vmovshdup {{.*#+}} xmm1 = mem[1,1,3,3] sched: [6:0.50] ; SANDY-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movshdup: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmovshdup {{.*#+}} xmm0 = xmm0[1,1,3,3] sched: [1:1.00] -; HASWELL-NEXT: vmovshdup {{.*#+}} xmm1 = mem[1,1,3,3] sched: [4:0.50] +; HASWELL-NEXT: vmovshdup {{.*#+}} xmm1 = mem[1,1,3,3] sched: [?:5.000000e-01] ; HASWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movshdup: ; BTVER2: # BB#0: @@ -430,16 +430,16 @@ define <4 x float> @test_movsldup(<4 x float> %a0, <4 x float> *%a1) { ; SANDY-LABEL: test_movsldup: ; SANDY: # BB#0: ; SANDY-NEXT: vmovsldup {{.*#+}} xmm0 = xmm0[0,0,2,2] sched: [1:1.00] -; SANDY-NEXT: vmovsldup {{.*#+}} xmm1 = mem[0,0,2,2] sched: [4:0.50] +; SANDY-NEXT: vmovsldup {{.*#+}} xmm1 = mem[0,0,2,2] sched: [6:0.50] ; SANDY-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movsldup: ; HASWELL: # BB#0: ; HASWELL-NEXT: vmovsldup {{.*#+}} xmm0 = xmm0[0,0,2,2] sched: [1:1.00] -; HASWELL-NEXT: vmovsldup {{.*#+}} xmm1 = mem[0,0,2,2] sched: [4:0.50] +; HASWELL-NEXT: vmovsldup {{.*#+}} xmm1 = mem[0,0,2,2] sched: [?:5.000000e-01] ; HASWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movsldup: ; BTVER2: # BB#0: diff --git a/llvm/test/CodeGen/X86/sse41-schedule.ll b/llvm/test/CodeGen/X86/sse41-schedule.ll index 340b9abe887..b30d8164157 100644 --- a/llvm/test/CodeGen/X86/sse41-schedule.ll +++ b/llvm/test/CodeGen/X86/sse41-schedule.ll @@ -25,17 +25,17 @@ define <2 x double> @test_blendpd(<2 x double> %a0, <2 x double> %a1, <2 x doubl ; ; SANDY-LABEL: test_blendpd: ; SANDY: # BB#0: -; SANDY-NEXT: vblendpd {{.*#+}} xmm0 = xmm0[0],xmm1[1] sched: [1:0.50] +; SANDY-NEXT: vblendpd {{.*#+}} xmm0 = xmm0[0],xmm1[1] sched: [1:1.00] ; SANDY-NEXT: vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vblendpd {{.*#+}} xmm0 = xmm0[0],mem[1] sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vblendpd {{.*#+}} xmm0 = xmm0[0],mem[1] sched: [7:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_blendpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vblendpd {{.*#+}} xmm0 = xmm0[0],xmm1[1] sched: [1:0.33] ; HASWELL-NEXT: vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vblendpd {{.*#+}} xmm0 = xmm0[0],mem[1] sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vblendpd {{.*#+}} xmm0 = xmm0[0],mem[1] sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_blendpd: ; BTVER2: # BB#0: @@ -65,15 +65,15 @@ define <4 x float> @test_blendps(<4 x float> %a0, <4 x float> %a1, <4 x float> * ; ; SANDY-LABEL: test_blendps: ; SANDY: # BB#0: -; SANDY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2],xmm0[3] sched: [1:0.50] -; SANDY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],mem[1],xmm0[2,3] sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2],xmm0[3] sched: [1:1.00] +; SANDY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],mem[1],xmm0[2,3] sched: [7:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_blendps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2],xmm0[3] sched: [1:0.33] -; HASWELL-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],mem[1],xmm0[2,3] sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],mem[1],xmm0[2,3] sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_blendps: ; BTVER2: # BB#0: @@ -107,15 +107,15 @@ define <2 x double> @test_blendvpd(<2 x double> %a0, <2 x double> %a1, <2 x doub ; ; SANDY-LABEL: test_blendvpd: ; SANDY: # BB#0: -; SANDY-NEXT: vblendvpd %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:1.00] -; SANDY-NEXT: vblendvpd %xmm2, (%rdi), %xmm0, %xmm0 # sched: [6:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vblendvpd %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:2.00] +; SANDY-NEXT: vblendvpd %xmm2, (%rdi), %xmm0, %xmm0 # sched: [8:2.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_blendvpd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vblendvpd %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:2.00] -; HASWELL-NEXT: vblendvpd %xmm2, (%rdi), %xmm0, %xmm0 # sched: [6:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vblendvpd %xmm2, (%rdi), %xmm0, %xmm0 # sched: [2:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_blendvpd: ; BTVER2: # BB#0: @@ -150,15 +150,15 @@ define <4 x float> @test_blendvps(<4 x float> %a0, <4 x float> %a1, <4 x float> ; ; SANDY-LABEL: test_blendvps: ; SANDY: # BB#0: -; SANDY-NEXT: vblendvps %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:1.00] -; SANDY-NEXT: vblendvps %xmm2, (%rdi), %xmm0, %xmm0 # sched: [6:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vblendvps %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:2.00] +; SANDY-NEXT: vblendvps %xmm2, (%rdi), %xmm0, %xmm0 # sched: [8:2.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_blendvps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vblendvps %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:2.00] -; HASWELL-NEXT: vblendvps %xmm2, (%rdi), %xmm0, %xmm0 # sched: [6:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vblendvps %xmm2, (%rdi), %xmm0, %xmm0 # sched: [2:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_blendvps: ; BTVER2: # BB#0: @@ -187,15 +187,15 @@ define <2 x double> @test_dppd(<2 x double> %a0, <2 x double> %a1, <2 x double> ; ; SANDY-LABEL: test_dppd: ; SANDY: # BB#0: -; SANDY-NEXT: vdppd $7, %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vdppd $7, (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vdppd $7, %xmm1, %xmm0, %xmm0 # sched: [9:1.00] +; SANDY-NEXT: vdppd $7, (%rdi), %xmm0, %xmm0 # sched: [15:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_dppd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vdppd $7, %xmm1, %xmm0, %xmm0 # sched: [9:1.00] -; HASWELL-NEXT: vdppd $7, (%rdi), %xmm0, %xmm0 # sched: [13:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vdppd $7, (%rdi), %xmm0, %xmm0 # sched: [9:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_dppd: ; BTVER2: # BB#0: @@ -224,15 +224,15 @@ define <4 x float> @test_dpps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2 ; ; SANDY-LABEL: test_dpps: ; SANDY: # BB#0: -; SANDY-NEXT: vdpps $7, %xmm1, %xmm0, %xmm0 # sched: [3:1.00] +; SANDY-NEXT: vdpps $7, %xmm1, %xmm0, %xmm0 # sched: [12:2.00] ; SANDY-NEXT: vdpps $7, (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_dpps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vdpps $7, %xmm1, %xmm0, %xmm0 # sched: [14:2.00] -; HASWELL-NEXT: vdpps $7, (%rdi), %xmm0, %xmm0 # sched: [18:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vdpps $7, (%rdi), %xmm0, %xmm0 # sched: [14:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_dpps: ; BTVER2: # BB#0: @@ -262,14 +262,14 @@ define <4 x float> @test_insertps(<4 x float> %a0, <4 x float> %a1, float *%a2) ; SANDY-LABEL: test_insertps: ; SANDY: # BB#0: ; SANDY-NEXT: vinsertps {{.*#+}} xmm0 = zero,xmm1[0],xmm0[2,3] sched: [1:1.00] -; SANDY-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0] sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0] sched: [7:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_insertps: ; HASWELL: # BB#0: ; HASWELL-NEXT: vinsertps {{.*#+}} xmm0 = zero,xmm1[0],xmm0[2,3] sched: [1:1.00] -; HASWELL-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0] sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0] sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_insertps: ; BTVER2: # BB#0: @@ -296,13 +296,13 @@ define <2 x i64> @test_movntdqa(i8* %a0) { ; ; SANDY-LABEL: test_movntdqa: ; SANDY: # BB#0: -; SANDY-NEXT: vmovntdqa (%rdi), %xmm0 # sched: [4:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmovntdqa (%rdi), %xmm0 # sched: [6:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_movntdqa: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmovntdqa (%rdi), %xmm0 # sched: [4:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmovntdqa (%rdi), %xmm0 # sched: [?:5.000000e-01] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_movntdqa: ; BTVER2: # BB#0: @@ -328,15 +328,15 @@ define <8 x i16> @test_mpsadbw(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; ; SANDY-LABEL: test_mpsadbw: ; SANDY: # BB#0: -; SANDY-NEXT: vmpsadbw $7, %xmm1, %xmm0, %xmm0 # sched: [6:1.00] -; SANDY-NEXT: vmpsadbw $7, (%rdi), %xmm0, %xmm0 # sched: [6:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vmpsadbw $7, %xmm1, %xmm0, %xmm0 # sched: [5:1.00] +; SANDY-NEXT: vmpsadbw $7, (%rdi), %xmm0, %xmm0 # sched: [11:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_mpsadbw: ; HASWELL: # BB#0: -; HASWELL-NEXT: vmpsadbw $7, %xmm1, %xmm0, %xmm0 # sched: [6:2.00] -; HASWELL-NEXT: vmpsadbw $7, (%rdi), %xmm0, %xmm0 # sched: [6:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vmpsadbw $7, %xmm1, %xmm0, %xmm0 # sched: [7:2.00] +; HASWELL-NEXT: vmpsadbw $7, (%rdi), %xmm0, %xmm0 # sched: [7:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_mpsadbw: ; BTVER2: # BB#0: @@ -367,14 +367,14 @@ define <8 x i16> @test_packusdw(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; SANDY-LABEL: test_packusdw: ; SANDY: # BB#0: ; SANDY-NEXT: vpackusdw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpackusdw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpackusdw (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_packusdw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpackusdw %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpackusdw (%rdi), %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpackusdw (%rdi), %xmm0, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_packusdw: ; BTVER2: # BB#0: @@ -411,14 +411,14 @@ define <16 x i8> @test_pblendvb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> %a2, <16 ; SANDY-LABEL: test_pblendvb: ; SANDY: # BB#0: ; SANDY-NEXT: vpblendvb %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:1.00] -; SANDY-NEXT: vpblendvb %xmm2, (%rdi), %xmm0, %xmm0 # sched: [6:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpblendvb %xmm2, (%rdi), %xmm0, %xmm0 # sched: [8:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pblendvb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpblendvb %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:2.00] -; HASWELL-NEXT: vpblendvb %xmm2, (%rdi), %xmm0, %xmm0 # sched: [6:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpblendvb %xmm2, (%rdi), %xmm0, %xmm0 # sched: [2:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pblendvb: ; BTVER2: # BB#0: @@ -448,14 +448,14 @@ define <8 x i16> @test_pblendw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_pblendw: ; SANDY: # BB#0: ; SANDY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] sched: [1:0.50] -; SANDY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],mem[2,3],xmm0[4,5,6],mem[7] sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],mem[2,3],xmm0[4,5,6],mem[7] sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pblendw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] sched: [1:1.00] -; HASWELL-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],mem[2,3],xmm0[4,5,6],mem[7] sched: [4:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],mem[2,3],xmm0[4,5,6],mem[7] sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pblendw: ; BTVER2: # BB#0: @@ -484,14 +484,14 @@ define <2 x i64> @test_pcmpeqq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) { ; SANDY-LABEL: test_pcmpeqq: ; SANDY: # BB#0: ; SANDY-NEXT: vpcmpeqq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpcmpeqq (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpcmpeqq (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pcmpeqq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpcmpeqq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpcmpeqq (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpcmpeqq (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pcmpeqq: ; BTVER2: # BB#0: @@ -521,15 +521,15 @@ define i32 @test_pextrb(<16 x i8> %a0, i8 *%a1) { ; ; SANDY-LABEL: test_pextrb: ; SANDY: # BB#0: -; SANDY-NEXT: vpextrb $3, %xmm0, %eax # sched: [1:0.50] +; SANDY-NEXT: vpextrb $3, %xmm0, %eax # sched: [3:1.00] ; SANDY-NEXT: vpextrb $1, %xmm0, (%rdi) # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pextrb: ; HASWELL: # BB#0: -; HASWELL-NEXT: vpextrb $3, %xmm0, %eax # sched: [1:1.00] -; HASWELL-NEXT: vpextrb $1, %xmm0, (%rdi) # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpextrb $3, %xmm0, %eax # sched: [2:1.00] +; HASWELL-NEXT: vpextrb $1, %xmm0, (%rdi) # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pextrb: ; BTVER2: # BB#0: @@ -558,15 +558,15 @@ define i32 @test_pextrd(<4 x i32> %a0, i32 *%a1) { ; ; SANDY-LABEL: test_pextrd: ; SANDY: # BB#0: -; SANDY-NEXT: vpextrd $3, %xmm0, %eax # sched: [1:0.50] +; SANDY-NEXT: vpextrd $3, %xmm0, %eax # sched: [3:1.00] ; SANDY-NEXT: vpextrd $1, %xmm0, (%rdi) # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pextrd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vpextrd $3, %xmm0, %eax # sched: [1:1.00] -; HASWELL-NEXT: vpextrd $1, %xmm0, (%rdi) # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpextrd $3, %xmm0, %eax # sched: [2:1.00] +; HASWELL-NEXT: vpextrd $1, %xmm0, (%rdi) # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pextrd: ; BTVER2: # BB#0: @@ -594,15 +594,15 @@ define i64 @test_pextrq(<2 x i64> %a0, <2 x i64> %a1, i64 *%a2) { ; ; SANDY-LABEL: test_pextrq: ; SANDY: # BB#0: -; SANDY-NEXT: vpextrq $1, %xmm0, %rax # sched: [1:0.50] +; SANDY-NEXT: vpextrq $1, %xmm0, %rax # sched: [3:1.00] ; SANDY-NEXT: vpextrq $1, %xmm0, (%rdi) # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pextrq: ; HASWELL: # BB#0: -; HASWELL-NEXT: vpextrq $1, %xmm0, %rax # sched: [1:1.00] -; HASWELL-NEXT: vpextrq $1, %xmm0, (%rdi) # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpextrq $1, %xmm0, %rax # sched: [2:1.00] +; HASWELL-NEXT: vpextrq $1, %xmm0, (%rdi) # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pextrq: ; BTVER2: # BB#0: @@ -630,15 +630,15 @@ define i32 @test_pextrw(<8 x i16> %a0, i16 *%a1) { ; ; SANDY-LABEL: test_pextrw: ; SANDY: # BB#0: -; SANDY-NEXT: vpextrw $3, %xmm0, %eax # sched: [1:0.50] +; SANDY-NEXT: vpextrw $3, %xmm0, %eax # sched: [3:1.00] ; SANDY-NEXT: vpextrw $1, %xmm0, (%rdi) # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pextrw: ; HASWELL: # BB#0: -; HASWELL-NEXT: vpextrw $3, %xmm0, %eax # sched: [1:1.00] -; HASWELL-NEXT: vpextrw $1, %xmm0, (%rdi) # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpextrw $3, %xmm0, %eax # sched: [2:1.00] +; HASWELL-NEXT: vpextrw $1, %xmm0, (%rdi) # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pextrw: ; BTVER2: # BB#0: @@ -667,15 +667,15 @@ define <8 x i16> @test_phminposuw(<8 x i16> *%a0) { ; ; SANDY-LABEL: test_phminposuw: ; SANDY: # BB#0: -; SANDY-NEXT: vphminposuw (%rdi), %xmm0 # sched: [9:1.00] +; SANDY-NEXT: vphminposuw (%rdi), %xmm0 # sched: [11:1.00] ; SANDY-NEXT: vphminposuw %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_phminposuw: ; HASWELL: # BB#0: -; HASWELL-NEXT: vphminposuw (%rdi), %xmm0 # sched: [9:1.00] +; HASWELL-NEXT: vphminposuw (%rdi), %xmm0 # sched: [5:1.00] ; HASWELL-NEXT: vphminposuw %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_phminposuw: ; BTVER2: # BB#0: @@ -704,15 +704,15 @@ define <16 x i8> @test_pinsrb(<16 x i8> %a0, i8 %a1, i8 *%a2) { ; ; SANDY-LABEL: test_pinsrb: ; SANDY: # BB#0: -; SANDY-NEXT: vpinsrb $1, %edi, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpinsrb $3, (%rsi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpinsrb $1, %edi, %xmm0, %xmm0 # sched: [2:1.00] +; SANDY-NEXT: vpinsrb $3, (%rsi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pinsrb: ; HASWELL: # BB#0: -; HASWELL-NEXT: vpinsrb $1, %edi, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpinsrb $3, (%rsi), %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpinsrb $1, %edi, %xmm0, %xmm0 # sched: [2:2.00] +; HASWELL-NEXT: vpinsrb $3, (%rsi), %xmm0, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pinsrb: ; BTVER2: # BB#0: @@ -740,15 +740,15 @@ define <4 x i32> @test_pinsrd(<4 x i32> %a0, i32 %a1, i32 *%a2) { ; ; SANDY-LABEL: test_pinsrd: ; SANDY: # BB#0: -; SANDY-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpinsrd $3, (%rsi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 # sched: [2:1.00] +; SANDY-NEXT: vpinsrd $3, (%rsi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pinsrd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpinsrd $3, (%rsi), %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 # sched: [2:2.00] +; HASWELL-NEXT: vpinsrd $3, (%rsi), %xmm0, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pinsrd: ; BTVER2: # BB#0: @@ -778,17 +778,17 @@ define <2 x i64> @test_pinsrq(<2 x i64> %a0, <2 x i64> %a1, i64 %a2, i64 *%a3) { ; ; SANDY-LABEL: test_pinsrq: ; SANDY: # BB#0: -; SANDY-NEXT: vpinsrq $1, %rdi, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpinsrq $1, (%rsi), %xmm1, %xmm1 # sched: [5:0.50] +; SANDY-NEXT: vpinsrq $1, %rdi, %xmm0, %xmm0 # sched: [2:1.00] +; SANDY-NEXT: vpinsrq $1, (%rsi), %xmm1, %xmm1 # sched: [7:0.50] ; SANDY-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pinsrq: ; HASWELL: # BB#0: -; HASWELL-NEXT: vpinsrq $1, %rdi, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpinsrq $1, (%rsi), %xmm1, %xmm1 # sched: [5:1.00] +; HASWELL-NEXT: vpinsrq $1, %rdi, %xmm0, %xmm0 # sched: [2:2.00] +; HASWELL-NEXT: vpinsrq $1, (%rsi), %xmm1, %xmm1 # sched: [1:1.00] ; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pinsrq: ; BTVER2: # BB#0: @@ -819,14 +819,14 @@ define <16 x i8> @test_pmaxsb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_pmaxsb: ; SANDY: # BB#0: ; SANDY-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpmaxsb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpmaxsb (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmaxsb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpmaxsb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpmaxsb (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmaxsb: ; BTVER2: # BB#0: @@ -856,14 +856,14 @@ define <4 x i32> @test_pmaxsd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; SANDY-LABEL: test_pmaxsd: ; SANDY: # BB#0: ; SANDY-NEXT: vpmaxsd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpmaxsd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpmaxsd (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmaxsd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmaxsd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpmaxsd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpmaxsd (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmaxsd: ; BTVER2: # BB#0: @@ -893,14 +893,14 @@ define <4 x i32> @test_pmaxud(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; SANDY-LABEL: test_pmaxud: ; SANDY: # BB#0: ; SANDY-NEXT: vpmaxud %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpmaxud (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpmaxud (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmaxud: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmaxud %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpmaxud (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpmaxud (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmaxud: ; BTVER2: # BB#0: @@ -930,14 +930,14 @@ define <8 x i16> @test_pmaxuw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_pmaxuw: ; SANDY: # BB#0: ; SANDY-NEXT: vpmaxuw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpmaxuw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpmaxuw (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmaxuw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmaxuw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpmaxuw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpmaxuw (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmaxuw: ; BTVER2: # BB#0: @@ -967,14 +967,14 @@ define <16 x i8> @test_pminsb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_pminsb: ; SANDY: # BB#0: ; SANDY-NEXT: vpminsb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpminsb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpminsb (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pminsb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpminsb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpminsb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpminsb (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pminsb: ; BTVER2: # BB#0: @@ -1004,14 +1004,14 @@ define <4 x i32> @test_pminsd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; SANDY-LABEL: test_pminsd: ; SANDY: # BB#0: ; SANDY-NEXT: vpminsd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpminsd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpminsd (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pminsd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpminsd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpminsd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpminsd (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pminsd: ; BTVER2: # BB#0: @@ -1041,14 +1041,14 @@ define <4 x i32> @test_pminud(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; SANDY-LABEL: test_pminud: ; SANDY: # BB#0: ; SANDY-NEXT: vpminud %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpminud (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpminud (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pminud: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpminud %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpminud (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpminud (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pminud: ; BTVER2: # BB#0: @@ -1078,14 +1078,14 @@ define <8 x i16> @test_pminuw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_pminuw: ; SANDY: # BB#0: ; SANDY-NEXT: vpminuw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpminuw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpminuw (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pminuw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpminuw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpminuw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpminuw (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pminuw: ; BTVER2: # BB#0: @@ -1118,16 +1118,16 @@ define <8 x i16> @test_pmovsxbw(<16 x i8> %a0, <8 x i8> *%a1) { ; SANDY-LABEL: test_pmovsxbw: ; SANDY: # BB#0: ; SANDY-NEXT: vpmovsxbw %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpmovsxbw (%rdi), %xmm1 # sched: [5:0.50] +; SANDY-NEXT: vpmovsxbw (%rdi), %xmm1 # sched: [7:0.50] ; SANDY-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmovsxbw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmovsxbw %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpmovsxbw (%rdi), %xmm1 # sched: [5:1.00] +; HASWELL-NEXT: vpmovsxbw (%rdi), %xmm1 # sched: [1:1.00] ; HASWELL-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmovsxbw: ; BTVER2: # BB#0: @@ -1162,16 +1162,16 @@ define <4 x i32> @test_pmovsxbd(<16 x i8> %a0, <4 x i8> *%a1) { ; SANDY-LABEL: test_pmovsxbd: ; SANDY: # BB#0: ; SANDY-NEXT: vpmovsxbd %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpmovsxbd (%rdi), %xmm1 # sched: [5:0.50] +; SANDY-NEXT: vpmovsxbd (%rdi), %xmm1 # sched: [7:0.50] ; SANDY-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmovsxbd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmovsxbd %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpmovsxbd (%rdi), %xmm1 # sched: [5:1.00] +; HASWELL-NEXT: vpmovsxbd (%rdi), %xmm1 # sched: [1:1.00] ; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmovsxbd: ; BTVER2: # BB#0: @@ -1206,16 +1206,16 @@ define <2 x i64> @test_pmovsxbq(<16 x i8> %a0, <2 x i8> *%a1) { ; SANDY-LABEL: test_pmovsxbq: ; SANDY: # BB#0: ; SANDY-NEXT: vpmovsxbq %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpmovsxbq (%rdi), %xmm1 # sched: [5:0.50] +; SANDY-NEXT: vpmovsxbq (%rdi), %xmm1 # sched: [7:0.50] ; SANDY-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmovsxbq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmovsxbq %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpmovsxbq (%rdi), %xmm1 # sched: [5:1.00] +; HASWELL-NEXT: vpmovsxbq (%rdi), %xmm1 # sched: [1:1.00] ; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmovsxbq: ; BTVER2: # BB#0: @@ -1250,16 +1250,16 @@ define <2 x i64> @test_pmovsxdq(<4 x i32> %a0, <2 x i32> *%a1) { ; SANDY-LABEL: test_pmovsxdq: ; SANDY: # BB#0: ; SANDY-NEXT: vpmovsxdq %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpmovsxdq (%rdi), %xmm1 # sched: [5:0.50] +; SANDY-NEXT: vpmovsxdq (%rdi), %xmm1 # sched: [7:0.50] ; SANDY-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmovsxdq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmovsxdq %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpmovsxdq (%rdi), %xmm1 # sched: [5:1.00] +; HASWELL-NEXT: vpmovsxdq (%rdi), %xmm1 # sched: [1:1.00] ; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmovsxdq: ; BTVER2: # BB#0: @@ -1294,16 +1294,16 @@ define <4 x i32> @test_pmovsxwd(<8 x i16> %a0, <4 x i16> *%a1) { ; SANDY-LABEL: test_pmovsxwd: ; SANDY: # BB#0: ; SANDY-NEXT: vpmovsxwd %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpmovsxwd (%rdi), %xmm1 # sched: [5:0.50] +; SANDY-NEXT: vpmovsxwd (%rdi), %xmm1 # sched: [7:0.50] ; SANDY-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmovsxwd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmovsxwd %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpmovsxwd (%rdi), %xmm1 # sched: [5:1.00] +; HASWELL-NEXT: vpmovsxwd (%rdi), %xmm1 # sched: [1:1.00] ; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmovsxwd: ; BTVER2: # BB#0: @@ -1338,16 +1338,16 @@ define <2 x i64> @test_pmovsxwq(<8 x i16> %a0, <2 x i16> *%a1) { ; SANDY-LABEL: test_pmovsxwq: ; SANDY: # BB#0: ; SANDY-NEXT: vpmovsxwq %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpmovsxwq (%rdi), %xmm1 # sched: [5:0.50] +; SANDY-NEXT: vpmovsxwq (%rdi), %xmm1 # sched: [7:0.50] ; SANDY-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmovsxwq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmovsxwq %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpmovsxwq (%rdi), %xmm1 # sched: [5:1.00] +; HASWELL-NEXT: vpmovsxwq (%rdi), %xmm1 # sched: [1:1.00] ; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmovsxwq: ; BTVER2: # BB#0: @@ -1382,16 +1382,16 @@ define <8 x i16> @test_pmovzxbw(<16 x i8> %a0, <8 x i8> *%a1) { ; SANDY-LABEL: test_pmovzxbw: ; SANDY: # BB#0: ; SANDY-NEXT: vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [1:0.50] -; SANDY-NEXT: vpmovzxbw {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero sched: [5:0.50] +; SANDY-NEXT: vpmovzxbw {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero sched: [7:0.50] ; SANDY-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmovzxbw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [1:1.00] -; HASWELL-NEXT: vpmovzxbw {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero sched: [5:1.00] +; HASWELL-NEXT: vpmovzxbw {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero sched: [1:1.00] ; HASWELL-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmovzxbw: ; BTVER2: # BB#0: @@ -1426,16 +1426,16 @@ define <4 x i32> @test_pmovzxbd(<16 x i8> %a0, <4 x i8> *%a1) { ; SANDY-LABEL: test_pmovzxbd: ; SANDY: # BB#0: ; SANDY-NEXT: vpmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [1:0.50] -; SANDY-NEXT: vpmovzxbd {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [5:0.50] +; SANDY-NEXT: vpmovzxbd {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [7:0.50] ; SANDY-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmovzxbd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [1:1.00] -; HASWELL-NEXT: vpmovzxbd {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [5:1.00] +; HASWELL-NEXT: vpmovzxbd {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [1:1.00] ; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmovzxbd: ; BTVER2: # BB#0: @@ -1470,16 +1470,16 @@ define <2 x i64> @test_pmovzxbq(<16 x i8> %a0, <2 x i8> *%a1) { ; SANDY-LABEL: test_pmovzxbq: ; SANDY: # BB#0: ; SANDY-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero sched: [1:0.50] -; SANDY-NEXT: vpmovzxbq {{.*#+}} xmm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero sched: [5:0.50] +; SANDY-NEXT: vpmovzxbq {{.*#+}} xmm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero sched: [7:0.50] ; SANDY-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmovzxbq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero sched: [1:1.00] -; HASWELL-NEXT: vpmovzxbq {{.*#+}} xmm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero sched: [5:1.00] +; HASWELL-NEXT: vpmovzxbq {{.*#+}} xmm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero sched: [1:1.00] ; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmovzxbq: ; BTVER2: # BB#0: @@ -1514,16 +1514,16 @@ define <2 x i64> @test_pmovzxdq(<4 x i32> %a0, <2 x i32> *%a1) { ; SANDY-LABEL: test_pmovzxdq: ; SANDY: # BB#0: ; SANDY-NEXT: vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero sched: [1:0.50] -; SANDY-NEXT: vpmovzxdq {{.*#+}} xmm1 = mem[0],zero,mem[1],zero sched: [5:0.50] +; SANDY-NEXT: vpmovzxdq {{.*#+}} xmm1 = mem[0],zero,mem[1],zero sched: [7:0.50] ; SANDY-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmovzxdq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero sched: [1:1.00] -; HASWELL-NEXT: vpmovzxdq {{.*#+}} xmm1 = mem[0],zero,mem[1],zero sched: [5:1.00] +; HASWELL-NEXT: vpmovzxdq {{.*#+}} xmm1 = mem[0],zero,mem[1],zero sched: [1:1.00] ; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmovzxdq: ; BTVER2: # BB#0: @@ -1558,16 +1558,16 @@ define <4 x i32> @test_pmovzxwd(<8 x i16> %a0, <4 x i16> *%a1) { ; SANDY-LABEL: test_pmovzxwd: ; SANDY: # BB#0: ; SANDY-NEXT: vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [1:0.50] -; SANDY-NEXT: vpmovzxwd {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [5:0.50] +; SANDY-NEXT: vpmovzxwd {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [7:0.50] ; SANDY-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmovzxwd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [1:1.00] -; HASWELL-NEXT: vpmovzxwd {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [5:1.00] +; HASWELL-NEXT: vpmovzxwd {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [1:1.00] ; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmovzxwd: ; BTVER2: # BB#0: @@ -1602,16 +1602,16 @@ define <2 x i64> @test_pmovzxwq(<8 x i16> %a0, <2 x i16> *%a1) { ; SANDY-LABEL: test_pmovzxwq: ; SANDY: # BB#0: ; SANDY-NEXT: vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero sched: [1:0.50] -; SANDY-NEXT: vpmovzxwq {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero sched: [5:0.50] +; SANDY-NEXT: vpmovzxwq {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero sched: [7:0.50] ; SANDY-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmovzxwq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero sched: [1:1.00] -; HASWELL-NEXT: vpmovzxwq {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero sched: [5:1.00] +; HASWELL-NEXT: vpmovzxwq {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero sched: [1:1.00] ; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmovzxwq: ; BTVER2: # BB#0: @@ -1642,15 +1642,15 @@ define <2 x i64> @test_pmuldq(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; ; SANDY-LABEL: test_pmuldq: ; SANDY: # BB#0: -; SANDY-NEXT: vpmuldq %xmm1, %xmm0, %xmm0 # sched: [5:1.00] +; SANDY-NEXT: vpmuldq %xmm1, %xmm0, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vpmuldq (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmuldq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmuldq %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: vpmuldq (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpmuldq (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmuldq: ; BTVER2: # BB#0: @@ -1680,15 +1680,15 @@ define <4 x i32> @test_pmulld(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; ; SANDY-LABEL: test_pmulld: ; SANDY: # BB#0: -; SANDY-NEXT: vpmulld %xmm1, %xmm0, %xmm0 # sched: [5:1.00] +; SANDY-NEXT: vpmulld %xmm1, %xmm0, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vpmulld (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmulld: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmulld %xmm1, %xmm0, %xmm0 # sched: [10:2.00] ; HASWELL-NEXT: vpmulld (%rdi), %xmm0, %xmm0 # sched: [10:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmulld: ; BTVER2: # BB#0: @@ -1724,23 +1724,23 @@ define i32 @test_ptest(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) { ; ; SANDY-LABEL: test_ptest: ; SANDY: # BB#0: -; SANDY-NEXT: vptest %xmm1, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: setb %al # sched: [1:0.33] -; SANDY-NEXT: vptest (%rdi), %xmm0 # sched: [5:0.50] -; SANDY-NEXT: setb %cl # sched: [1:0.33] +; SANDY-NEXT: vptest %xmm1, %xmm0 # sched: [2:1.00] +; SANDY-NEXT: setb %al # sched: [1:1.00] +; SANDY-NEXT: vptest (%rdi), %xmm0 # sched: [8:1.00] +; SANDY-NEXT: setb %cl # sched: [1:1.00] ; SANDY-NEXT: andb %al, %cl # sched: [1:0.33] ; SANDY-NEXT: movzbl %cl, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_ptest: ; HASWELL: # BB#0: ; HASWELL-NEXT: vptest %xmm1, %xmm0 # sched: [2:1.00] -; HASWELL-NEXT: setb %al # sched: [1:0.50] +; HASWELL-NEXT: setb %al # sched: [1:1.00] ; HASWELL-NEXT: vptest (%rdi), %xmm0 # sched: [2:1.00] -; HASWELL-NEXT: setb %cl # sched: [1:0.50] +; HASWELL-NEXT: setb %cl # sched: [1:1.00] ; HASWELL-NEXT: andb %al, %cl # sched: [1:0.25] ; HASWELL-NEXT: movzbl %cl, %eax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_ptest: ; BTVER2: # BB#0: @@ -1778,16 +1778,16 @@ define <2 x double> @test_roundpd(<2 x double> %a0, <2 x double> *%a1) { ; SANDY-LABEL: test_roundpd: ; SANDY: # BB#0: ; SANDY-NEXT: vroundpd $7, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vroundpd $7, (%rdi), %xmm1 # sched: [7:1.00] +; SANDY-NEXT: vroundpd $7, (%rdi), %xmm1 # sched: [9:1.00] ; SANDY-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_roundpd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vroundpd $7, %xmm0, %xmm0 # sched: [6:2.00] -; HASWELL-NEXT: vroundpd $7, (%rdi), %xmm1 # sched: [10:2.00] +; HASWELL-NEXT: vroundpd $7, %xmm0, %xmm0 # sched: [5:2.00] +; HASWELL-NEXT: vroundpd $7, (%rdi), %xmm1 # sched: [6:1.00] ; HASWELL-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_roundpd: ; BTVER2: # BB#0: @@ -1822,16 +1822,16 @@ define <4 x float> @test_roundps(<4 x float> %a0, <4 x float> *%a1) { ; SANDY-LABEL: test_roundps: ; SANDY: # BB#0: ; SANDY-NEXT: vroundps $7, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: vroundps $7, (%rdi), %xmm1 # sched: [7:1.00] +; SANDY-NEXT: vroundps $7, (%rdi), %xmm1 # sched: [9:1.00] ; SANDY-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_roundps: ; HASWELL: # BB#0: -; HASWELL-NEXT: vroundps $7, %xmm0, %xmm0 # sched: [6:2.00] -; HASWELL-NEXT: vroundps $7, (%rdi), %xmm1 # sched: [10:2.00] +; HASWELL-NEXT: vroundps $7, %xmm0, %xmm0 # sched: [5:2.00] +; HASWELL-NEXT: vroundps $7, (%rdi), %xmm1 # sched: [6:1.00] ; HASWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_roundps: ; BTVER2: # BB#0: @@ -1867,16 +1867,16 @@ define <2 x double> @test_roundsd(<2 x double> %a0, <2 x double> %a1, <2 x doubl ; SANDY-LABEL: test_roundsd: ; SANDY: # BB#0: ; SANDY-NEXT: vroundsd $7, %xmm1, %xmm0, %xmm1 # sched: [3:1.00] -; SANDY-NEXT: vroundsd $7, (%rdi), %xmm0, %xmm0 # sched: [7:1.00] +; SANDY-NEXT: vroundsd $7, (%rdi), %xmm0, %xmm0 # sched: [9:1.00] ; SANDY-NEXT: vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_roundsd: ; HASWELL: # BB#0: -; HASWELL-NEXT: vroundsd $7, %xmm1, %xmm0, %xmm1 # sched: [6:2.00] -; HASWELL-NEXT: vroundsd $7, (%rdi), %xmm0, %xmm0 # sched: [10:2.00] +; HASWELL-NEXT: vroundsd $7, %xmm1, %xmm0, %xmm1 # sched: [5:2.00] +; HASWELL-NEXT: vroundsd $7, (%rdi), %xmm0, %xmm0 # sched: [6:1.00] ; HASWELL-NEXT: vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_roundsd: ; BTVER2: # BB#0: @@ -1912,16 +1912,16 @@ define <4 x float> @test_roundss(<4 x float> %a0, <4 x float> %a1, <4 x float> * ; SANDY-LABEL: test_roundss: ; SANDY: # BB#0: ; SANDY-NEXT: vroundss $7, %xmm1, %xmm0, %xmm1 # sched: [3:1.00] -; SANDY-NEXT: vroundss $7, (%rdi), %xmm0, %xmm0 # sched: [7:1.00] +; SANDY-NEXT: vroundss $7, (%rdi), %xmm0, %xmm0 # sched: [9:1.00] ; SANDY-NEXT: vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_roundss: ; HASWELL: # BB#0: -; HASWELL-NEXT: vroundss $7, %xmm1, %xmm0, %xmm1 # sched: [6:2.00] -; HASWELL-NEXT: vroundss $7, (%rdi), %xmm0, %xmm0 # sched: [10:2.00] +; HASWELL-NEXT: vroundss $7, %xmm1, %xmm0, %xmm1 # sched: [3:1.00] +; HASWELL-NEXT: vroundss $7, (%rdi), %xmm0, %xmm0 # sched: [6:1.00] ; HASWELL-NEXT: vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_roundss: ; BTVER2: # BB#0: diff --git a/llvm/test/CodeGen/X86/sse42-schedule.ll b/llvm/test/CodeGen/X86/sse42-schedule.ll index afc48bc57ee..c51c50a91bd 100644 --- a/llvm/test/CodeGen/X86/sse42-schedule.ll +++ b/llvm/test/CodeGen/X86/sse42-schedule.ll @@ -26,16 +26,16 @@ define i32 @crc32_32_8(i32 %a0, i8 %a1, i8 *%a2) { ; SANDY-LABEL: crc32_32_8: ; SANDY: # BB#0: ; SANDY-NEXT: crc32b %sil, %edi # sched: [3:1.00] -; SANDY-NEXT: crc32b (%rdx), %edi # sched: [7:1.00] +; SANDY-NEXT: crc32b (%rdx), %edi # sched: [8:1.00] ; SANDY-NEXT: movl %edi, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: crc32_32_8: ; HASWELL: # BB#0: ; HASWELL-NEXT: crc32b %sil, %edi # sched: [3:1.00] ; HASWELL-NEXT: crc32b (%rdx), %edi # sched: [7:1.00] ; HASWELL-NEXT: movl %edi, %eax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: crc32_32_8: ; BTVER2: # BB#0: @@ -68,16 +68,16 @@ define i32 @crc32_32_16(i32 %a0, i16 %a1, i16 *%a2) { ; SANDY-LABEL: crc32_32_16: ; SANDY: # BB#0: ; SANDY-NEXT: crc32w %si, %edi # sched: [3:1.00] -; SANDY-NEXT: crc32w (%rdx), %edi # sched: [7:1.00] +; SANDY-NEXT: crc32w (%rdx), %edi # sched: [8:1.00] ; SANDY-NEXT: movl %edi, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: crc32_32_16: ; HASWELL: # BB#0: ; HASWELL-NEXT: crc32w %si, %edi # sched: [3:1.00] ; HASWELL-NEXT: crc32w (%rdx), %edi # sched: [7:1.00] ; HASWELL-NEXT: movl %edi, %eax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: crc32_32_16: ; BTVER2: # BB#0: @@ -112,14 +112,14 @@ define i32 @crc32_32_32(i32 %a0, i32 %a1, i32 *%a2) { ; SANDY-NEXT: crc32l %esi, %edi # sched: [3:1.00] ; SANDY-NEXT: crc32l (%rdx), %edi # sched: [7:1.00] ; SANDY-NEXT: movl %edi, %eax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: crc32_32_32: ; HASWELL: # BB#0: ; HASWELL-NEXT: crc32l %esi, %edi # sched: [3:1.00] ; HASWELL-NEXT: crc32l (%rdx), %edi # sched: [7:1.00] ; HASWELL-NEXT: movl %edi, %eax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: crc32_32_32: ; BTVER2: # BB#0: @@ -152,16 +152,16 @@ define i64 @crc32_64_8(i64 %a0, i8 %a1, i8 *%a2) nounwind { ; SANDY-LABEL: crc32_64_8: ; SANDY: # BB#0: ; SANDY-NEXT: crc32b %sil, %edi # sched: [3:1.00] -; SANDY-NEXT: crc32b (%rdx), %edi # sched: [7:1.00] +; SANDY-NEXT: crc32b (%rdx), %edi # sched: [8:1.00] ; SANDY-NEXT: movq %rdi, %rax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: crc32_64_8: ; HASWELL: # BB#0: ; HASWELL-NEXT: crc32b %sil, %edi # sched: [3:1.00] ; HASWELL-NEXT: crc32b (%rdx), %edi # sched: [7:1.00] ; HASWELL-NEXT: movq %rdi, %rax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: crc32_64_8: ; BTVER2: # BB#0: @@ -196,14 +196,14 @@ define i64 @crc32_64_64(i64 %a0, i64 %a1, i64 *%a2) { ; SANDY-NEXT: crc32q %rsi, %rdi # sched: [3:1.00] ; SANDY-NEXT: crc32q (%rdx), %rdi # sched: [7:1.00] ; SANDY-NEXT: movq %rdi, %rax # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: crc32_64_64: ; HASWELL: # BB#0: ; HASWELL-NEXT: crc32q %rsi, %rdi # sched: [3:1.00] ; HASWELL-NEXT: crc32q (%rdx), %rdi # sched: [7:1.00] ; HASWELL-NEXT: movq %rdi, %rax # sched: [1:0.25] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: crc32_64_64: ; BTVER2: # BB#0: @@ -256,20 +256,20 @@ define i32 @test_pcmpestri(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-NEXT: vpcmpestri $7, (%rdi), %xmm0 # sched: [4:2.33] ; SANDY-NEXT: # kill: %ECX<def> %ECX<kill> %RCX<def> ; SANDY-NEXT: leal (%rcx,%rsi), %eax # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pcmpestri: ; HASWELL: # BB#0: ; HASWELL-NEXT: movl $7, %eax # sched: [1:0.25] ; HASWELL-NEXT: movl $7, %edx # sched: [1:0.25] -; HASWELL-NEXT: vpcmpestri $7, %xmm1, %xmm0 # sched: [11:3.00] +; HASWELL-NEXT: vpcmpestri $7, %xmm1, %xmm0 # sched: [18:4.00] ; HASWELL-NEXT: movl %ecx, %esi # sched: [1:0.25] ; HASWELL-NEXT: movl $7, %eax # sched: [1:0.25] ; HASWELL-NEXT: movl $7, %edx # sched: [1:0.25] -; HASWELL-NEXT: vpcmpestri $7, (%rdi), %xmm0 # sched: [11:3.00] +; HASWELL-NEXT: vpcmpestri $7, (%rdi), %xmm0 # sched: [18:4.00] ; HASWELL-NEXT: # kill: %ECX<def> %ECX<kill> %RCX<def> ; HASWELL-NEXT: leal (%rcx,%rsi), %eax # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pcmpestri: ; BTVER2: # BB#0: @@ -320,17 +320,17 @@ define <16 x i8> @test_pcmpestrm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-NEXT: movl $7, %eax # sched: [1:0.33] ; SANDY-NEXT: movl $7, %edx # sched: [1:0.33] ; SANDY-NEXT: vpcmpestrm $7, (%rdi), %xmm0 # sched: [11:2.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pcmpestrm: ; HASWELL: # BB#0: ; HASWELL-NEXT: movl $7, %eax # sched: [1:0.25] ; HASWELL-NEXT: movl $7, %edx # sched: [1:0.25] -; HASWELL-NEXT: vpcmpestrm $7, %xmm1, %xmm0 # sched: [10:4.00] +; HASWELL-NEXT: vpcmpestrm $7, %xmm1, %xmm0 # sched: [19:4.00] ; HASWELL-NEXT: movl $7, %eax # sched: [1:0.25] ; HASWELL-NEXT: movl $7, %edx # sched: [1:0.25] -; HASWELL-NEXT: vpcmpestrm $7, (%rdi), %xmm0 # sched: [10:3.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpcmpestrm $7, (%rdi), %xmm0 # sched: [19:4.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pcmpestrm: ; BTVER2: # BB#0: @@ -369,12 +369,12 @@ define i32 @test_pcmpistri(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; ; SANDY-LABEL: test_pcmpistri: ; SANDY: # BB#0: -; SANDY-NEXT: vpcmpistri $7, %xmm1, %xmm0 # sched: [3:1.00] +; SANDY-NEXT: vpcmpistri $7, %xmm1, %xmm0 # sched: [11:3.00] ; SANDY-NEXT: movl %ecx, %eax # sched: [1:0.33] -; SANDY-NEXT: vpcmpistri $7, (%rdi), %xmm0 # sched: [3:1.00] +; SANDY-NEXT: vpcmpistri $7, (%rdi), %xmm0 # sched: [17:3.00] ; SANDY-NEXT: # kill: %ECX<def> %ECX<kill> %RCX<def> ; SANDY-NEXT: leal (%rcx,%rax), %eax # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pcmpistri: ; HASWELL: # BB#0: @@ -383,7 +383,7 @@ define i32 @test_pcmpistri(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; HASWELL-NEXT: vpcmpistri $7, (%rdi), %xmm0 # sched: [11:3.00] ; HASWELL-NEXT: # kill: %ECX<def> %ECX<kill> %RCX<def> ; HASWELL-NEXT: leal (%rcx,%rax), %eax # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pcmpistri: ; BTVER2: # BB#0: @@ -416,15 +416,15 @@ define <16 x i8> @test_pcmpistrm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; ; SANDY-LABEL: test_pcmpistrm: ; SANDY: # BB#0: -; SANDY-NEXT: vpcmpistrm $7, %xmm1, %xmm0 # sched: [11:1.00] -; SANDY-NEXT: vpcmpistrm $7, (%rdi), %xmm0 # sched: [11:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpcmpistrm $7, %xmm1, %xmm0 # sched: [11:3.00] +; SANDY-NEXT: vpcmpistrm $7, (%rdi), %xmm0 # sched: [17:3.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pcmpistrm: ; HASWELL: # BB#0: -; HASWELL-NEXT: vpcmpistrm $7, %xmm1, %xmm0 # sched: [10:3.00] -; HASWELL-NEXT: vpcmpistrm $7, (%rdi), %xmm0 # sched: [10:3.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpcmpistrm $7, %xmm1, %xmm0 # sched: [11:3.00] +; HASWELL-NEXT: vpcmpistrm $7, (%rdi), %xmm0 # sched: [11:3.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pcmpistrm: ; BTVER2: # BB#0: @@ -453,15 +453,15 @@ define <2 x i64> @test_pcmpgtq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) { ; ; SANDY-LABEL: test_pcmpgtq: ; SANDY: # BB#0: -; SANDY-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpcmpgtq (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm0 # sched: [5:1.00] +; SANDY-NEXT: vpcmpgtq (%rdi), %xmm0, %xmm0 # sched: [11:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pcmpgtq: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm0 # sched: [5:1.00] ; HASWELL-NEXT: vpcmpgtq (%rdi), %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pcmpgtq: ; BTVER2: # BB#0: diff --git a/llvm/test/CodeGen/X86/ssse3-schedule.ll b/llvm/test/CodeGen/X86/ssse3-schedule.ll index 8b7a0c0ec02..353784e6dc4 100644 --- a/llvm/test/CodeGen/X86/ssse3-schedule.ll +++ b/llvm/test/CodeGen/X86/ssse3-schedule.ll @@ -35,16 +35,16 @@ define <16 x i8> @test_pabsb(<16 x i8> %a0, <16 x i8> *%a1) { ; SANDY-LABEL: test_pabsb: ; SANDY: # BB#0: ; SANDY-NEXT: vpabsb %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpabsb (%rdi), %xmm1 # sched: [5:0.50] +; SANDY-NEXT: vpabsb (%rdi), %xmm1 # sched: [7:0.50] ; SANDY-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pabsb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpabsb %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpabsb (%rdi), %xmm1 # sched: [5:0.50] +; HASWELL-NEXT: vpabsb (%rdi), %xmm1 # sched: [1:0.50] ; HASWELL-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pabsb: ; BTVER2: # BB#0: @@ -86,16 +86,16 @@ define <4 x i32> @test_pabsd(<4 x i32> %a0, <4 x i32> *%a1) { ; SANDY-LABEL: test_pabsd: ; SANDY: # BB#0: ; SANDY-NEXT: vpabsd %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpabsd (%rdi), %xmm1 # sched: [5:0.50] +; SANDY-NEXT: vpabsd (%rdi), %xmm1 # sched: [7:0.50] ; SANDY-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pabsd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpabsd %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpabsd (%rdi), %xmm1 # sched: [5:0.50] +; HASWELL-NEXT: vpabsd (%rdi), %xmm1 # sched: [1:0.50] ; HASWELL-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pabsd: ; BTVER2: # BB#0: @@ -136,12 +136,12 @@ define <8 x i16> @test_pabsw(<8 x i16> %a0, <8 x i16> *%a1) { ; SANDY-LABEL: test_pabsw: ; SANDY: # BB#0: ; SANDY-NEXT: vpabsw %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pabsw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpabsw %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pabsw: ; BTVER2: # BB#0: @@ -182,14 +182,14 @@ define <8 x i16> @test_palignr(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_palignr: ; SANDY: # BB#0: ; SANDY-NEXT: vpalignr {{.*#+}} xmm0 = xmm0[6,7,8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5] sched: [1:0.50] -; SANDY-NEXT: vpalignr {{.*#+}} xmm0 = mem[14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13] sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpalignr {{.*#+}} xmm0 = mem[14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13] sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_palignr: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpalignr {{.*#+}} xmm0 = xmm0[6,7,8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5] sched: [1:1.00] -; HASWELL-NEXT: vpalignr {{.*#+}} xmm0 = mem[14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13] sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpalignr {{.*#+}} xmm0 = mem[14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13] sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_palignr: ; BTVER2: # BB#0: @@ -223,15 +223,15 @@ define <4 x i32> @test_phaddd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; ; SANDY-LABEL: test_phaddd: ; SANDY: # BB#0: -; SANDY-NEXT: vphaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vphaddd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vphaddd %xmm1, %xmm0, %xmm0 # sched: [3:1.50] +; SANDY-NEXT: vphaddd (%rdi), %xmm0, %xmm0 # sched: [9:1.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_phaddd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vphaddd %xmm1, %xmm0, %xmm0 # sched: [3:2.00] -; HASWELL-NEXT: vphaddd (%rdi), %xmm0, %xmm0 # sched: [6:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vphaddd (%rdi), %xmm0, %xmm0 # sched: [3:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_phaddd: ; BTVER2: # BB#0: @@ -274,15 +274,15 @@ define <8 x i16> @test_phaddsw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; ; SANDY-LABEL: test_phaddsw: ; SANDY: # BB#0: -; SANDY-NEXT: vphaddsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vphaddsw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vphaddsw %xmm1, %xmm0, %xmm0 # sched: [3:1.50] +; SANDY-NEXT: vphaddsw (%rdi), %xmm0, %xmm0 # sched: [9:1.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_phaddsw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vphaddsw %xmm1, %xmm0, %xmm0 # sched: [3:2.00] -; HASWELL-NEXT: vphaddsw (%rdi), %xmm0, %xmm0 # sched: [6:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vphaddsw (%rdi), %xmm0, %xmm0 # sched: [3:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_phaddsw: ; BTVER2: # BB#0: @@ -317,15 +317,15 @@ define <8 x i16> @test_phaddw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; ; SANDY-LABEL: test_phaddw: ; SANDY: # BB#0: -; SANDY-NEXT: vphaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vphaddw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vphaddw %xmm1, %xmm0, %xmm0 # sched: [3:1.50] +; SANDY-NEXT: vphaddw (%rdi), %xmm0, %xmm0 # sched: [9:1.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_phaddw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vphaddw %xmm1, %xmm0, %xmm0 # sched: [3:2.00] -; HASWELL-NEXT: vphaddw (%rdi), %xmm0, %xmm0 # sched: [6:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vphaddw (%rdi), %xmm0, %xmm0 # sched: [3:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_phaddw: ; BTVER2: # BB#0: @@ -360,15 +360,15 @@ define <4 x i32> @test_phsubd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; ; SANDY-LABEL: test_phsubd: ; SANDY: # BB#0: -; SANDY-NEXT: vphsubd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vphsubd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vphsubd %xmm1, %xmm0, %xmm0 # sched: [3:1.50] +; SANDY-NEXT: vphsubd (%rdi), %xmm0, %xmm0 # sched: [9:1.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_phsubd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vphsubd %xmm1, %xmm0, %xmm0 # sched: [3:2.00] -; HASWELL-NEXT: vphsubd (%rdi), %xmm0, %xmm0 # sched: [6:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vphsubd (%rdi), %xmm0, %xmm0 # sched: [3:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_phsubd: ; BTVER2: # BB#0: @@ -411,15 +411,15 @@ define <8 x i16> @test_phsubsw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; ; SANDY-LABEL: test_phsubsw: ; SANDY: # BB#0: -; SANDY-NEXT: vphsubsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vphsubsw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vphsubsw %xmm1, %xmm0, %xmm0 # sched: [3:1.50] +; SANDY-NEXT: vphsubsw (%rdi), %xmm0, %xmm0 # sched: [9:1.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_phsubsw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vphsubsw %xmm1, %xmm0, %xmm0 # sched: [3:2.00] -; HASWELL-NEXT: vphsubsw (%rdi), %xmm0, %xmm0 # sched: [6:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vphsubsw (%rdi), %xmm0, %xmm0 # sched: [3:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_phsubsw: ; BTVER2: # BB#0: @@ -454,15 +454,15 @@ define <8 x i16> @test_phsubw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; ; SANDY-LABEL: test_phsubw: ; SANDY: # BB#0: -; SANDY-NEXT: vphsubw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vphsubw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vphsubw %xmm1, %xmm0, %xmm0 # sched: [3:1.50] +; SANDY-NEXT: vphsubw (%rdi), %xmm0, %xmm0 # sched: [9:1.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_phsubw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vphsubw %xmm1, %xmm0, %xmm0 # sched: [3:2.00] -; HASWELL-NEXT: vphsubw (%rdi), %xmm0, %xmm0 # sched: [6:2.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vphsubw (%rdi), %xmm0, %xmm0 # sched: [3:2.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_phsubw: ; BTVER2: # BB#0: @@ -497,15 +497,15 @@ define <8 x i16> @test_pmaddubsw(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; ; SANDY-LABEL: test_pmaddubsw: ; SANDY: # BB#0: -; SANDY-NEXT: vpmaddubsw %xmm1, %xmm0, %xmm0 # sched: [5:1.00] +; SANDY-NEXT: vpmaddubsw %xmm1, %xmm0, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vpmaddubsw (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmaddubsw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmaddubsw %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: vpmaddubsw (%rdi), %xmm0, %xmm0 # sched: [9:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpmaddubsw (%rdi), %xmm0, %xmm0 # sched: [5:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmaddubsw: ; BTVER2: # BB#0: @@ -538,13 +538,13 @@ define <8 x i16> @test_pmulhrsw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; ; SANDY-LABEL: test_pmulhrsw: ; SANDY: # BB#0: -; SANDY-NEXT: vpmulhrsw %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpmulhrsw %xmm1, %xmm0, %xmm0 # sched: [3:1.00] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pmulhrsw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpmulhrsw %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pmulhrsw: ; BTVER2: # BB#0: @@ -579,14 +579,14 @@ define <16 x i8> @test_pshufb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_pshufb: ; SANDY: # BB#0: ; SANDY-NEXT: vpshufb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpshufb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpshufb (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_pshufb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpshufb %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpshufb (%rdi), %xmm0, %xmm0 # sched: [5:1.00] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpshufb (%rdi), %xmm0, %xmm0 # sched: [1:1.00] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_pshufb: ; BTVER2: # BB#0: @@ -630,14 +630,14 @@ define <16 x i8> @test_psignb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) { ; SANDY-LABEL: test_psignb: ; SANDY: # BB#0: ; SANDY-NEXT: vpsignb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpsignb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsignb (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psignb: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsignb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsignb (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpsignb (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psignb: ; BTVER2: # BB#0: @@ -681,14 +681,14 @@ define <4 x i32> @test_psignd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { ; SANDY-LABEL: test_psignd: ; SANDY: # BB#0: ; SANDY-NEXT: vpsignd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpsignd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsignd (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psignd: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsignd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsignd (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpsignd (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psignd: ; BTVER2: # BB#0: @@ -732,14 +732,14 @@ define <8 x i16> @test_psignw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { ; SANDY-LABEL: test_psignw: ; SANDY: # BB#0: ; SANDY-NEXT: vpsignw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SANDY-NEXT: vpsignw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; SANDY-NEXT: retq # sched: [5:1.00] +; SANDY-NEXT: vpsignw (%rdi), %xmm0, %xmm0 # sched: [7:0.50] +; SANDY-NEXT: retq # sched: [1:1.00] ; ; HASWELL-LABEL: test_psignw: ; HASWELL: # BB#0: ; HASWELL-NEXT: vpsignw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsignw (%rdi), %xmm0, %xmm0 # sched: [5:0.50] -; HASWELL-NEXT: retq # sched: [1:1.00] +; HASWELL-NEXT: vpsignw (%rdi), %xmm0, %xmm0 # sched: [1:0.50] +; HASWELL-NEXT: retq # sched: [2:1.00] ; ; BTVER2-LABEL: test_psignw: ; BTVER2: # BB#0: diff --git a/llvm/test/CodeGen/X86/vector-shift-ashr-512.ll b/llvm/test/CodeGen/X86/vector-shift-ashr-512.ll index 4d4b7f4e822..a07d2dd90be 100644 --- a/llvm/test/CodeGen/X86/vector-shift-ashr-512.ll +++ b/llvm/test/CodeGen/X86/vector-shift-ashr-512.ll @@ -201,14 +201,14 @@ define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, <64 x i8> %b) nounwind { ; AVX512DQ-NEXT: vpsraw $2, %ymm0, %ymm5 ; AVX512DQ-NEXT: vpaddw %ymm2, %ymm2, %ymm8 ; AVX512DQ-NEXT: vpblendvb %ymm8, %ymm5, %ymm0, %ymm0 +; AVX512DQ-NEXT: vpunpckhbw {{.*#+}} ymm5 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31] +; AVX512DQ-NEXT: vpsraw $4, %ymm5, %ymm9 +; AVX512DQ-NEXT: vpblendvb %ymm3, %ymm9, %ymm5, %ymm3 ; AVX512DQ-NEXT: vpsraw $1, %ymm0, %ymm5 ; AVX512DQ-NEXT: vpaddw %ymm8, %ymm8, %ymm9 ; AVX512DQ-NEXT: vpblendvb %ymm9, %ymm5, %ymm0, %ymm0 ; AVX512DQ-NEXT: vpsrlw $8, %ymm0, %ymm0 ; AVX512DQ-NEXT: vpackuswb %ymm4, %ymm0, %ymm0 -; AVX512DQ-NEXT: vpunpckhbw {{.*#+}} ymm4 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31] -; AVX512DQ-NEXT: vpsraw $4, %ymm4, %ymm5 -; AVX512DQ-NEXT: vpblendvb %ymm3, %ymm5, %ymm4, %ymm3 ; AVX512DQ-NEXT: vpsraw $2, %ymm3, %ymm4 ; AVX512DQ-NEXT: vpblendvb %ymm6, %ymm4, %ymm3, %ymm3 ; AVX512DQ-NEXT: vpsraw $1, %ymm3, %ymm4 @@ -328,14 +328,14 @@ define <64 x i8> @constant_shift_v64i8(<64 x i8> %a) nounwind { ; AVX512DQ-NEXT: vpsraw $2, %ymm0, %ymm5 ; AVX512DQ-NEXT: vpaddw %ymm2, %ymm2, %ymm8 ; AVX512DQ-NEXT: vpblendvb %ymm8, %ymm5, %ymm0, %ymm0 +; AVX512DQ-NEXT: vpunpckhbw {{.*#+}} ymm5 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31] +; AVX512DQ-NEXT: vpsraw $4, %ymm5, %ymm9 +; AVX512DQ-NEXT: vpblendvb %ymm3, %ymm9, %ymm5, %ymm3 ; AVX512DQ-NEXT: vpsraw $1, %ymm0, %ymm5 ; AVX512DQ-NEXT: vpaddw %ymm8, %ymm8, %ymm9 ; AVX512DQ-NEXT: vpblendvb %ymm9, %ymm5, %ymm0, %ymm0 ; AVX512DQ-NEXT: vpsrlw $8, %ymm0, %ymm0 ; AVX512DQ-NEXT: vpackuswb %ymm4, %ymm0, %ymm0 -; AVX512DQ-NEXT: vpunpckhbw {{.*#+}} ymm4 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31] -; AVX512DQ-NEXT: vpsraw $4, %ymm4, %ymm5 -; AVX512DQ-NEXT: vpblendvb %ymm3, %ymm5, %ymm4, %ymm3 ; AVX512DQ-NEXT: vpsraw $2, %ymm3, %ymm4 ; AVX512DQ-NEXT: vpblendvb %ymm6, %ymm4, %ymm3, %ymm3 ; AVX512DQ-NEXT: vpsraw $1, %ymm3, %ymm4 diff --git a/llvm/test/CodeGen/X86/vector-shuffle-512-v32.ll b/llvm/test/CodeGen/X86/vector-shuffle-512-v32.ll index 7a5c992bb82..d6c4ff1255e 100644 --- a/llvm/test/CodeGen/X86/vector-shuffle-512-v32.ll +++ b/llvm/test/CodeGen/X86/vector-shuffle-512-v32.ll @@ -68,13 +68,13 @@ define <32 x i16> @shuffle_v32i16_0f_1f_0e_16_0d_1d_04_1e_0b_1b_0a_1a_09_19_08_1 ; KNL-NEXT: vpshufb {{.*#+}} xmm4 = xmm1[8,9,12,13,12,13,10,11,0,1,4,5,4,5,0,1] ; KNL-NEXT: vpshufd {{.*#+}} xmm1 = xmm2[0,1,0,3] ; KNL-NEXT: vpshuflw {{.*#+}} xmm2 = xmm1[0,3,2,2,4,5,6,7] -; KNL-NEXT: vinserti128 $1, %xmm2, %ymm4, %ymm1 -; KNL-NEXT: vextracti128 $1, %ymm0, %xmm5 -; KNL-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7] +; KNL-NEXT: vextracti128 $1, %ymm0, %xmm1 +; KNL-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] ; KNL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[12,13,10,11,8,9,14,15,4,5,2,3,2,3,6,7] -; KNL-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[6,7,2,3,4,5,6,7,2,3,2,3,0,1,14,15] -; KNL-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm0 -; KNL-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15] +; KNL-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[6,7,2,3,4,5,6,7,2,3,2,3,0,1,14,15] +; KNL-NEXT: vinserti128 $1, %xmm2, %ymm4, %ymm5 +; KNL-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 +; KNL-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0],ymm5[1],ymm0[2],ymm5[3],ymm0[4],ymm5[5],ymm0[6],ymm5[7],ymm0[8],ymm5[9],ymm0[10],ymm5[11],ymm0[12],ymm5[13],ymm0[14],ymm5[15] ; KNL-NEXT: vextracti128 $1, %ymm3, %xmm3 ; KNL-NEXT: vpbroadcastw %xmm3, %ymm3 ; KNL-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0] |
