diff options
Diffstat (limited to 'llvm/test/CodeGen/X86/avx2-schedule.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/avx2-schedule.ll | 7111 |
1 files changed, 0 insertions, 7111 deletions
diff --git a/llvm/test/CodeGen/X86/avx2-schedule.ll b/llvm/test/CodeGen/X86/avx2-schedule.ll deleted file mode 100644 index 00c4fcb114f..00000000000 --- a/llvm/test/CodeGen/X86/avx2-schedule.ll +++ /dev/null @@ -1,7111 +0,0 @@ -; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=x86-64 -mattr=+avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=GENERIC -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=haswell -mattr=-avx512f | FileCheck %s --check-prefix=CHECK --check-prefix=HASWELL -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=broadwell -mattr=-avx512f | FileCheck %s --check-prefix=CHECK --check-prefix=BROADWELL -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=skylake -mattr=-avx512f | FileCheck %s --check-prefix=CHECK --check-prefix=SKYLAKE -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=skx -mattr=-avx512f | FileCheck %s --check-prefix=CHECK --check-prefix=SKX -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=znver1 -mattr=-avx512f | FileCheck %s --check-prefix=CHECK --check-prefix=ZNVER1 - -define <8 x i32> @test_broadcasti128(<8 x i32> %a0, <4 x i32> *%a1) { -; GENERIC-LABEL: test_broadcasti128: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vbroadcasti128 {{.*#+}} ymm1 = mem[0,1,0,1] sched: [6:1.00] -; GENERIC-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_broadcasti128: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vbroadcasti128 {{.*#+}} ymm1 = mem[0,1,0,1] sched: [7:0.50] -; HASWELL-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_broadcasti128: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vbroadcasti128 {{.*#+}} ymm1 = mem[0,1,0,1] sched: [6:0.50] -; BROADWELL-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_broadcasti128: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vbroadcasti128 {{.*#+}} ymm1 = mem[0,1,0,1] sched: [7:0.50] -; SKYLAKE-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_broadcasti128: -; SKX: # %bb.0: -; SKX-NEXT: vbroadcasti128 {{.*#+}} ymm1 = mem[0,1,0,1] sched: [7:0.50] -; SKX-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_broadcasti128: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vbroadcasti128 {{.*#+}} ymm1 = mem[0,1,0,1] sched: [8:0.50] -; ZNVER1-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = load <4 x i32>, <4 x i32> *%a1, align 16 - %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3> - %3 = add <8 x i32> %2, %a0 - ret <8 x i32> %3 -} - -define <4 x double> @test_broadcastsd_ymm(<2 x double> %a0) { -; GENERIC-LABEL: test_broadcastsd_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vbroadcastsd %xmm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_broadcastsd_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vbroadcastsd %xmm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_broadcastsd_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vbroadcastsd %xmm0, %ymm0 # sched: [3:1.00] -; BROADWELL-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_broadcastsd_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vbroadcastsd %xmm0, %ymm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [4:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_broadcastsd_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vbroadcastsd %xmm0, %ymm0 # sched: [3:1.00] -; SKX-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [4:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_broadcastsd_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vbroadcastsd %xmm0, %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <2 x double> %a0, <2 x double> undef, <4 x i32> zeroinitializer - %2 = fadd <4 x double> %1, %1 - ret <4 x double> %2 -} - -define <4 x float> @test_broadcastss(<4 x float> %a0) { -; GENERIC-LABEL: test_broadcastss: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vbroadcastss %xmm0, %xmm0 # sched: [1:1.00] -; GENERIC-NEXT: vaddps %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_broadcastss: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vbroadcastss %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vaddps %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_broadcastss: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vbroadcastss %xmm0, %xmm0 # sched: [1:1.00] -; BROADWELL-NEXT: vaddps %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_broadcastss: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vbroadcastss %xmm0, %xmm0 # sched: [1:1.00] -; SKYLAKE-NEXT: vaddps %xmm0, %xmm0, %xmm0 # sched: [4:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_broadcastss: -; SKX: # %bb.0: -; SKX-NEXT: vbroadcastss %xmm0, %xmm0 # sched: [1:1.00] -; SKX-NEXT: vaddps %xmm0, %xmm0, %xmm0 # sched: [4:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_broadcastss: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vbroadcastss %xmm0, %xmm0 # sched: [1:0.50] -; ZNVER1-NEXT: vaddps %xmm0, %xmm0, %xmm0 # sched: [3:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <4 x float> %a0, <4 x float> undef, <4 x i32> zeroinitializer - %2 = fadd <4 x float> %1, %1 - ret <4 x float> %2 -} - -define <8 x float> @test_broadcastss_ymm(<4 x float> %a0) { -; GENERIC-LABEL: test_broadcastss_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vbroadcastss %xmm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_broadcastss_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vbroadcastss %xmm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_broadcastss_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vbroadcastss %xmm0, %ymm0 # sched: [3:1.00] -; BROADWELL-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_broadcastss_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vbroadcastss %xmm0, %ymm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [4:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_broadcastss_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vbroadcastss %xmm0, %ymm0 # sched: [3:1.00] -; SKX-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [4:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_broadcastss_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vbroadcastss %xmm0, %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <4 x float> %a0, <4 x float> undef, <8 x i32> zeroinitializer - %2 = fadd <8 x float> %1, %1 - ret <8 x float> %2 -} - -define <4 x i32> @test_extracti128(<8 x i16> %a0, <4 x i32> *%a1) { -; GENERIC-LABEL: test_extracti128: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmovzxwd {{.*#+}} ymm1 = 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] -; GENERIC-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [1:1.00] -; GENERIC-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:1.00] -; GENERIC-NEXT: vzeroupper # sched: [1:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_extracti128: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmovzxwd {{.*#+}} ymm1 = 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: [3:1.00] -; HASWELL-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:1.00] -; HASWELL-NEXT: vzeroupper # sched: [0:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_extracti128: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmovzxwd {{.*#+}} ymm1 = 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: [3:1.00] -; BROADWELL-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [3:1.00] -; BROADWELL-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:1.00] -; BROADWELL-NEXT: vzeroupper # sched: [0:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_extracti128: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmovzxwd {{.*#+}} ymm1 = 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: [3:1.00] -; SKYLAKE-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:1.00] -; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_extracti128: -; SKX: # %bb.0: -; SKX-NEXT: vpmovzxwd {{.*#+}} ymm1 = 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: [3:1.00] -; SKX-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [3:1.00] -; SKX-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:1.00] -; SKX-NEXT: vzeroupper # sched: [0:0.67] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_extracti128: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmovzxwd {{.*#+}} ymm1 = 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] -; ZNVER1-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [2:0.25] -; ZNVER1-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:0.50] -; ZNVER1-NEXT: vzeroupper # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %z = zext <8 x i16> %a0 to <8 x i32> - %ext = shufflevector <8 x i32> %z, <8 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7> - store <4 x i32> %ext, <4 x i32> *%a1 - ret <4 x i32> %ext -} - -define <2 x double> @test_gatherdpd(<2 x double> %a0, i8* %a1, <4 x i32> %a2, <2 x double> %a3) { -; GENERIC-LABEL: test_gatherdpd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vgatherdpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_gatherdpd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vgatherdpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [26:2.67] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_gatherdpd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vgatherdpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [25:3.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_gatherdpd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vgatherdpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_gatherdpd: -; SKX: # %bb.0: -; SKX-NEXT: vgatherdpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_gatherdpd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vgatherdpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <2 x double> @llvm.x86.avx2.gather.d.pd(<2 x double> %a0, i8* %a1, <4 x i32> %a2, <2 x double> %a3, i8 2) - ret <2 x double> %1 -} -declare <2 x double> @llvm.x86.avx2.gather.d.pd(<2 x double>, i8*, <4 x i32>, <2 x double>, i8) nounwind readonly - -define <4 x double> @test_gatherdpd_ymm(<4 x double> %a0, i8* %a1, <4 x i32> %a2, <4 x double> %a3) { -; GENERIC-LABEL: test_gatherdpd_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vgatherdpd %ymm2, (%rdi,%xmm1,8), %ymm0 # sched: [5:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_gatherdpd_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vgatherdpd %ymm2, (%rdi,%xmm1,8), %ymm0 # sched: [27:4.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_gatherdpd_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vgatherdpd %ymm2, (%rdi,%xmm1,8), %ymm0 # sched: [26:5.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_gatherdpd_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vgatherdpd %ymm2, (%rdi,%xmm1,8), %ymm0 # sched: [25:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_gatherdpd_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vgatherdpd %ymm2, (%rdi,%xmm1,8), %ymm0 # sched: [25:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_gatherdpd_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vgatherdpd %ymm2, (%rdi,%xmm1,8), %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x double> @llvm.x86.avx2.gather.d.pd.256(<4 x double> %a0, i8* %a1, <4 x i32> %a2, <4 x double> %a3, i8 8) - ret <4 x double> %1 -} -declare <4 x double> @llvm.x86.avx2.gather.d.pd.256(<4 x double>, i8*, <4 x i32>, <4 x double>, i8) nounwind readonly - -define <4 x float> @test_gatherdps(<4 x float> %a0, i8* %a1, <4 x i32> %a2, <4 x float> %a3) { -; GENERIC-LABEL: test_gatherdps: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vgatherdps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_gatherdps: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vgatherdps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [25:3.67] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_gatherdps: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vgatherdps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [25:3.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_gatherdps: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vgatherdps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_gatherdps: -; SKX: # %bb.0: -; SKX-NEXT: vgatherdps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_gatherdps: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vgatherdps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x float> @llvm.x86.avx2.gather.d.ps(<4 x float> %a0, i8* %a1, <4 x i32> %a2, <4 x float> %a3, i8 2) - ret <4 x float> %1 -} -declare <4 x float> @llvm.x86.avx2.gather.d.ps(<4 x float>, i8*, <4 x i32>, <4 x float>, i8) nounwind readonly - -define <8 x float> @test_gatherdps_ymm(<8 x float> %a0, i8* %a1, <8 x i32> %a2, <8 x float> %a3) { -; GENERIC-LABEL: test_gatherdps_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vgatherdps %ymm2, (%rdi,%ymm1,4), %ymm0 # sched: [5:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_gatherdps_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vgatherdps %ymm2, (%rdi,%ymm1,4), %ymm0 # sched: [27:6.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_gatherdps_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vgatherdps %ymm2, (%rdi,%ymm1,4), %ymm0 # sched: [26:4.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_gatherdps_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vgatherdps %ymm2, (%rdi,%ymm1,4), %ymm0 # sched: [25:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_gatherdps_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vgatherdps %ymm2, (%rdi,%ymm1,4), %ymm0 # sched: [25:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_gatherdps_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vgatherdps %ymm2, (%rdi,%ymm1,4), %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x float> @llvm.x86.avx2.gather.d.ps.256(<8 x float> %a0, i8* %a1, <8 x i32> %a2, <8 x float> %a3, i8 4) - ret <8 x float> %1 -} -declare <8 x float> @llvm.x86.avx2.gather.d.ps.256(<8 x float>, i8*, <8 x i32>, <8 x float>, i8) nounwind readonly - -define <2 x double> @test_gatherqpd(<2 x double> %a0, i8* %a1, <2 x i64> %a2, <2 x double> %a3) { -; GENERIC-LABEL: test_gatherqpd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vgatherqpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_gatherqpd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vgatherqpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [23:3.33] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_gatherqpd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vgatherqpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:3.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_gatherqpd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vgatherqpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_gatherqpd: -; SKX: # %bb.0: -; SKX-NEXT: vgatherqpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_gatherqpd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vgatherqpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <2 x double> @llvm.x86.avx2.gather.q.pd(<2 x double> %a0, i8* %a1, <2 x i64> %a2, <2 x double> %a3, i8 2) - ret <2 x double> %1 -} -declare <2 x double> @llvm.x86.avx2.gather.q.pd(<2 x double>, i8*, <2 x i64>, <2 x double>, i8) nounwind readonly - -define <4 x double> @test_gatherqpd_ymm(<4 x double> %a0, i8* %a1, <4 x i64> %a2, <4 x double> %a3) { -; GENERIC-LABEL: test_gatherqpd_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vgatherqpd %ymm2, (%rdi,%ymm1,8), %ymm0 # sched: [5:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_gatherqpd_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vgatherqpd %ymm2, (%rdi,%ymm1,8), %ymm0 # sched: [24:5.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_gatherqpd_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vgatherqpd %ymm2, (%rdi,%ymm1,8), %ymm0 # sched: [23:3.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_gatherqpd_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vgatherqpd %ymm2, (%rdi,%ymm1,8), %ymm0 # sched: [25:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_gatherqpd_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vgatherqpd %ymm2, (%rdi,%ymm1,8), %ymm0 # sched: [25:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_gatherqpd_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vgatherqpd %ymm2, (%rdi,%ymm1,8), %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x double> @llvm.x86.avx2.gather.q.pd.256(<4 x double> %a0, i8* %a1, <4 x i64> %a2, <4 x double> %a3, i8 8) - ret <4 x double> %1 -} -declare <4 x double> @llvm.x86.avx2.gather.q.pd.256(<4 x double>, i8*, <4 x i64>, <4 x double>, i8) nounwind readonly - -define <4 x float> @test_gatherqps(<4 x float> %a0, i8* %a1, <2 x i64> %a2, <4 x float> %a3) { -; GENERIC-LABEL: test_gatherqps: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vgatherqps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_gatherqps: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vgatherqps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [25:3.67] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_gatherqps: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vgatherqps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [27:5.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_gatherqps: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vgatherqps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_gatherqps: -; SKX: # %bb.0: -; SKX-NEXT: vgatherqps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_gatherqps: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vgatherqps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x float> @llvm.x86.avx2.gather.q.ps(<4 x float> %a0, i8* %a1, <2 x i64> %a2, <4 x float> %a3, i8 2) - ret <4 x float> %1 -} -declare <4 x float> @llvm.x86.avx2.gather.q.ps(<4 x float>, i8*, <2 x i64>, <4 x float>, i8) nounwind readonly - -define <4 x float> @test_gatherqps_ymm(<4 x float> %a0, i8* %a1, <4 x i64> %a2, <4 x float> %a3) { -; GENERIC-LABEL: test_gatherqps_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [5:0.50] -; GENERIC-NEXT: vzeroupper # sched: [1:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_gatherqps_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [28:3.67] -; HASWELL-NEXT: vzeroupper # sched: [0:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_gatherqps_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [24:5.00] -; BROADWELL-NEXT: vzeroupper # sched: [0:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_gatherqps_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [25:1.00] -; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_gatherqps_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [25:1.00] -; SKX-NEXT: vzeroupper # sched: [0:0.67] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_gatherqps_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [100:0.25] -; ZNVER1-NEXT: vzeroupper # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x float> @llvm.x86.avx2.gather.q.ps.256(<4 x float> %a0, i8* %a1, <4 x i64> %a2, <4 x float> %a3, i8 4) - ret <4 x float> %1 -} -declare <4 x float> @llvm.x86.avx2.gather.q.ps.256(<4 x float>, i8*, <4 x i64>, <4 x float>, i8) nounwind readonly - -define <8 x i32> @test_inserti128(<8 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { -; GENERIC-LABEL: test_inserti128: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 # sched: [1:1.00] -; GENERIC-NEXT: vinserti128 $1, (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; GENERIC-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_inserti128: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 # sched: [3:1.00] -; HASWELL-NEXT: vinserti128 $1, (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; HASWELL-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_inserti128: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 # sched: [3:1.00] -; BROADWELL-NEXT: vinserti128 $1, (%rdi), %ymm0, %ymm0 # sched: [6:0.50] -; BROADWELL-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_inserti128: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 # sched: [3:1.00] -; SKYLAKE-NEXT: vinserti128 $1, (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; SKYLAKE-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_inserti128: -; SKX: # %bb.0: -; SKX-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 # sched: [3:1.00] -; SKX-NEXT: vinserti128 $1, (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; SKX-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_inserti128: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 # sched: [2:0.25] -; ZNVER1-NEXT: vinserti128 $1, (%rdi), %ymm0, %ymm0 # sched: [9:0.50] -; ZNVER1-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <4 x i32> %a1, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef> - %2 = shufflevector <8 x i32> %a0, <8 x i32> %1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11> - %3 = load <4 x i32>, <4 x i32> *%a2, align 16 - %4 = shufflevector <4 x i32> %3, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef> - %5 = shufflevector <8 x i32> %a0, <8 x i32> %4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11> - %6 = add <8 x i32> %2, %5 - ret <8 x i32> %6 -} - -define <4 x i64> @test_movntdqa(i8* %a0) { -; GENERIC-LABEL: test_movntdqa: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vmovntdqa (%rdi), %ymm0 # sched: [7:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_movntdqa: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vmovntdqa (%rdi), %ymm0 # sched: [7:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_movntdqa: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vmovntdqa (%rdi), %ymm0 # sched: [6:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_movntdqa: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vmovntdqa (%rdi), %ymm0 # sched: [7:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_movntdqa: -; SKX: # %bb.0: -; SKX-NEXT: vmovntdqa (%rdi), %ymm0 # sched: [7:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_movntdqa: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vmovntdqa (%rdi), %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i64> @llvm.x86.avx2.movntdqa(i8* %a0) - ret <4 x i64> %1 -} -declare <4 x i64> @llvm.x86.avx2.movntdqa(i8*) nounwind readonly - -define <16 x i16> @test_mpsadbw(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_mpsadbw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vmpsadbw $7, %ymm1, %ymm0, %ymm0 # sched: [7:1.00] -; GENERIC-NEXT: vmpsadbw $7, (%rdi), %ymm0, %ymm0 # sched: [14:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_mpsadbw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vmpsadbw $7, %ymm1, %ymm0, %ymm0 # sched: [7:2.00] -; HASWELL-NEXT: vmpsadbw $7, (%rdi), %ymm0, %ymm0 # sched: [14:2.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_mpsadbw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vmpsadbw $7, %ymm1, %ymm0, %ymm0 # sched: [7:2.00] -; BROADWELL-NEXT: vmpsadbw $7, (%rdi), %ymm0, %ymm0 # sched: [13:2.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_mpsadbw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vmpsadbw $7, %ymm1, %ymm0, %ymm0 # sched: [4:2.00] -; SKYLAKE-NEXT: vmpsadbw $7, (%rdi), %ymm0, %ymm0 # sched: [11:2.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_mpsadbw: -; SKX: # %bb.0: -; SKX-NEXT: vmpsadbw $7, %ymm1, %ymm0, %ymm0 # sched: [4:2.00] -; SKX-NEXT: vmpsadbw $7, (%rdi), %ymm0, %ymm0 # sched: [11:2.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_mpsadbw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vmpsadbw $7, %ymm1, %ymm0, %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: vmpsadbw $7, (%rdi), %ymm0, %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.mpsadbw(<32 x i8> %a0, <32 x i8> %a1, i8 7) - %2 = bitcast <16 x i16> %1 to <32 x i8> - %3 = load <32 x i8>, <32 x i8> *%a2, align 32 - %4 = call <16 x i16> @llvm.x86.avx2.mpsadbw(<32 x i8> %2, <32 x i8> %3, i8 7) - ret <16 x i16> %4 -} -declare <16 x i16> @llvm.x86.avx2.mpsadbw(<32 x i8>, <32 x i8>, i8) nounwind readnone - -define <32 x i8> @test_pabsb(<32 x i8> %a0, <32 x i8> *%a1) { -; GENERIC-LABEL: test_pabsb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpabsb %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpabsb (%rdi), %ymm1 # sched: [8:0.50] -; GENERIC-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pabsb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpabsb %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpabsb (%rdi), %ymm1 # sched: [8:0.50] -; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pabsb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpabsb %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpabsb (%rdi), %ymm1 # sched: [7:0.50] -; BROADWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pabsb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpabsb %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpabsb (%rdi), %ymm1 # sched: [8:0.50] -; SKYLAKE-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pabsb: -; SKX: # %bb.0: -; SKX-NEXT: vpabsb %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpabsb (%rdi), %ymm1 # sched: [8:0.50] -; SKX-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pabsb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpabsb (%rdi), %ymm1 # sched: [8:0.50] -; ZNVER1-NEXT: vpabsb %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <32 x i8> @llvm.x86.avx2.pabs.b(<32 x i8> %a0) - %2 = load <32 x i8>, <32 x i8> *%a1, align 32 - %3 = call <32 x i8> @llvm.x86.avx2.pabs.b(<32 x i8> %2) - %4 = or <32 x i8> %1, %3 - ret <32 x i8> %4 -} -declare <32 x i8> @llvm.x86.avx2.pabs.b(<32 x i8>) nounwind readnone - -define <8 x i32> @test_pabsd(<8 x i32> %a0, <8 x i32> *%a1) { -; GENERIC-LABEL: test_pabsd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpabsd %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpabsd (%rdi), %ymm1 # sched: [8:0.50] -; GENERIC-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pabsd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpabsd %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpabsd (%rdi), %ymm1 # sched: [8:0.50] -; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pabsd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpabsd %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpabsd (%rdi), %ymm1 # sched: [7:0.50] -; BROADWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pabsd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpabsd %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpabsd (%rdi), %ymm1 # sched: [8:0.50] -; SKYLAKE-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pabsd: -; SKX: # %bb.0: -; SKX-NEXT: vpabsd %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpabsd (%rdi), %ymm1 # sched: [8:0.50] -; SKX-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pabsd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpabsd (%rdi), %ymm1 # sched: [8:0.50] -; ZNVER1-NEXT: vpabsd %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.pabs.d(<8 x i32> %a0) - %2 = load <8 x i32>, <8 x i32> *%a1, align 32 - %3 = call <8 x i32> @llvm.x86.avx2.pabs.d(<8 x i32> %2) - %4 = or <8 x i32> %1, %3 - ret <8 x i32> %4 -} -declare <8 x i32> @llvm.x86.avx2.pabs.d(<8 x i32>) nounwind readnone - -define <16 x i16> @test_pabsw(<16 x i16> %a0, <16 x i16> *%a1) { -; GENERIC-LABEL: test_pabsw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpabsw %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpabsw (%rdi), %ymm1 # sched: [8:0.50] -; GENERIC-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pabsw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpabsw %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpabsw (%rdi), %ymm1 # sched: [8:0.50] -; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pabsw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpabsw %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpabsw (%rdi), %ymm1 # sched: [7:0.50] -; BROADWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pabsw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpabsw %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpabsw (%rdi), %ymm1 # sched: [8:0.50] -; SKYLAKE-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pabsw: -; SKX: # %bb.0: -; SKX-NEXT: vpabsw %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpabsw (%rdi), %ymm1 # sched: [8:0.50] -; SKX-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pabsw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpabsw (%rdi), %ymm1 # sched: [8:0.50] -; ZNVER1-NEXT: vpabsw %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.pabs.w(<16 x i16> %a0) - %2 = load <16 x i16>, <16 x i16> *%a1, align 32 - %3 = call <16 x i16> @llvm.x86.avx2.pabs.w(<16 x i16> %2) - %4 = or <16 x i16> %1, %3 - ret <16 x i16> %4 -} -declare <16 x i16> @llvm.x86.avx2.pabs.w(<16 x i16>) nounwind readnone - -define <16 x i16> @test_packssdw(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_packssdw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_packssdw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_packssdw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; BROADWELL-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_packssdw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; SKYLAKE-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_packssdw: -; SKX: # %bb.0: -; SKX-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; SKX-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_packssdw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a0, <8 x i32> %a1) - %2 = bitcast <16 x i16> %1 to <8 x i32> - %3 = load <8 x i32>, <8 x i32> *%a2, align 32 - %4 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %2, <8 x i32> %3) - ret <16 x i16> %4 -} -declare <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32>, <8 x i32>) nounwind readnone - -define <32 x i8> @test_packsswb(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_packsswb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_packsswb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_packsswb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; BROADWELL-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_packsswb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; SKYLAKE-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_packsswb: -; SKX: # %bb.0: -; SKX-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; SKX-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_packsswb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %a0, <16 x i16> %a1) - %2 = bitcast <32 x i8> %1 to <16 x i16> - %3 = load <16 x i16>, <16 x i16> *%a2, align 32 - %4 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %2, <16 x i16> %3) - ret <32 x i8> %4 -} -declare <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16>, <16 x i16>) nounwind readnone - -define <16 x i16> @test_packusdw(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_packusdw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_packusdw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_packusdw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; BROADWELL-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_packusdw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; SKYLAKE-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_packusdw: -; SKX: # %bb.0: -; SKX-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; SKX-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_packusdw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a0, <8 x i32> %a1) - %2 = bitcast <16 x i16> %1 to <8 x i32> - %3 = load <8 x i32>, <8 x i32> *%a2, align 32 - %4 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %2, <8 x i32> %3) - ret <16 x i16> %4 -} -declare <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32>, <8 x i32>) nounwind readnone - -define <32 x i8> @test_packuswb(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_packuswb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_packuswb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_packuswb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; BROADWELL-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_packuswb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; SKYLAKE-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_packuswb: -; SKX: # %bb.0: -; SKX-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; SKX-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_packuswb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %a0, <16 x i16> %a1) - %2 = bitcast <32 x i8> %1 to <16 x i16> - %3 = load <16 x i16>, <16 x i16> *%a2, align 32 - %4 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %2, <16 x i16> %3) - ret <32 x i8> %4 -} -declare <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16>, <16 x i16>) nounwind readnone - -define <32 x i8> @test_paddb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_paddb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpaddb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_paddb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_paddb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpaddb (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_paddb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: vpaddb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_paddb: -; SKX: # %bb.0: -; SKX-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: vpaddb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_paddb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpaddb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = add <32 x i8> %a0, %a1 - %2 = load <32 x i8>, <32 x i8> *%a2, align 32 - %3 = add <32 x i8> %1, %2 - ret <32 x i8> %3 -} - -define <8 x i32> @test_paddd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_paddd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpaddd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_paddd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_paddd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpaddd (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_paddd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: vpaddd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_paddd: -; SKX: # %bb.0: -; SKX-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: vpaddd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_paddd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpaddd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = add <8 x i32> %a0, %a1 - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = add <8 x i32> %1, %2 - ret <8 x i32> %3 -} - -define <4 x i64> @test_paddq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { -; GENERIC-LABEL: test_paddq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpaddq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_paddq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_paddq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpaddq (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_paddq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: vpaddq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_paddq: -; SKX: # %bb.0: -; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: vpaddq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_paddq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpaddq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = add <4 x i64> %a0, %a1 - %2 = load <4 x i64>, <4 x i64> *%a2, align 32 - %3 = add <4 x i64> %1, %2 - ret <4 x i64> %3 -} - -define <32 x i8> @test_paddsb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_paddsb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_paddsb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_paddsb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_paddsb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_paddsb: -; SKX: # %bb.0: -; SKX-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_paddsb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <32 x i8> @llvm.sadd.sat.v32i8(<32 x i8> %a0, <32 x i8> %a1) - %2 = load <32 x i8>, <32 x i8> *%a2, align 32 - %3 = call <32 x i8> @llvm.sadd.sat.v32i8(<32 x i8> %1, <32 x i8> %2) - ret <32 x i8> %3 -} -declare <32 x i8> @llvm.sadd.sat.v32i8(<32 x i8>, <32 x i8>) nounwind readnone - -define <16 x i16> @test_paddsw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_paddsw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_paddsw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_paddsw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_paddsw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_paddsw: -; SKX: # %bb.0: -; SKX-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_paddsw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.sadd.sat.v16i16(<16 x i16> %a0, <16 x i16> %a1) - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = call <16 x i16> @llvm.sadd.sat.v16i16(<16 x i16> %1, <16 x i16> %2) - ret <16 x i16> %3 -} -declare <16 x i16> @llvm.sadd.sat.v16i16(<16 x i16>, <16 x i16>) nounwind readnone - -define <32 x i8> @test_paddusb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_paddusb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpaddusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_paddusb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_paddusb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpaddusb (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_paddusb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpaddusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_paddusb: -; SKX: # %bb.0: -; SKX-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpaddusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_paddusb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpaddusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <32 x i8> @llvm.uadd.sat.v32i8(<32 x i8> %a0, <32 x i8> %a1) - %2 = load <32 x i8>, <32 x i8> *%a2, align 32 - %3 = call <32 x i8> @llvm.uadd.sat.v32i8(<32 x i8> %1, <32 x i8> %2) - ret <32 x i8> %3 -} -declare <32 x i8> @llvm.uadd.sat.v32i8(<32 x i8>, <32 x i8>) nounwind readnone - -define <16 x i16> @test_paddusw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_paddusw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpaddusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_paddusw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_paddusw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpaddusw (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_paddusw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpaddusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_paddusw: -; SKX: # %bb.0: -; SKX-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpaddusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_paddusw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpaddusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.uadd.sat.v16i16(<16 x i16> %a0, <16 x i16> %a1) - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = call <16 x i16> @llvm.uadd.sat.v16i16(<16 x i16> %1, <16 x i16> %2) - ret <16 x i16> %3 -} -declare <16 x i16> @llvm.uadd.sat.v16i16(<16 x i16>, <16 x i16>) nounwind readnone - -define <16 x i16> @test_paddw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_paddw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpaddw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_paddw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpaddw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_paddw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpaddw (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_paddw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: vpaddw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_paddw: -; SKX: # %bb.0: -; SKX-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: vpaddw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_paddw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpaddw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = add <16 x i16> %a0, %a1 - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = add <16 x i16> %1, %2 - ret <16 x i16> %3 -} - -define <32 x i8> @test_palignr(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_palignr: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0],ymm1[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16] sched: [1:1.00] -; GENERIC-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm1[0],ymm0[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm1[16] sched: [1:1.00] -; GENERIC-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_palignr: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0],ymm1[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16] sched: [1:1.00] -; HASWELL-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm1[0],ymm0[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm1[16] sched: [1:1.00] -; HASWELL-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_palignr: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0],ymm1[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16] sched: [1:1.00] -; BROADWELL-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm1[0],ymm0[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm1[16] sched: [1:1.00] -; BROADWELL-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_palignr: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0],ymm1[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16] sched: [1:1.00] -; SKYLAKE-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm1[0],ymm0[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm1[16] sched: [1:1.00] -; SKYLAKE-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_palignr: -; SKX: # %bb.0: -; SKX-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0],ymm1[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16] sched: [1:1.00] -; SKX-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm1[0],ymm0[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm1[16] sched: [1:1.00] -; SKX-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_palignr: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0],ymm1[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16] sched: [1:0.25] -; ZNVER1-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm1[0],ymm0[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm1[16] sched: [1:0.25] -; ZNVER1-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <32 x i8> %a1, <32 x i8> %a0, <32 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 32, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 48> - %2 = load <32 x i8>, <32 x i8> *%a2, align 32 - %3 = shufflevector <32 x i8> %a0, <32 x i8> %1, <32 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 32, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 48> - %4 = add <32 x i8> %1, %3 - ret <32 x i8> %4 -} - -define <4 x i64> @test_pand(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { -; GENERIC-LABEL: test_pand: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpand %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; GENERIC-NEXT: vpand (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pand: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpand %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: vpand (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pand: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpand %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; BROADWELL-NEXT: vpand (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pand: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpand %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: vpand (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pand: -; SKX: # %bb.0: -; SKX-NEXT: vpand %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: vpand (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pand: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpand %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpand (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = and <4 x i64> %a0, %a1 - %2 = load <4 x i64>, <4 x i64> *%a2, align 32 - %3 = and <4 x i64> %1, %2 - %4 = add <4 x i64> %3, %a1 - ret <4 x i64> %4 -} - -define <4 x i64> @test_pandn(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { -; GENERIC-LABEL: test_pandn: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpandn %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; GENERIC-NEXT: vpandn (%rdi), %ymm0, %ymm1 # sched: [8:0.50] -; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pandn: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpandn %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: vpandn (%rdi), %ymm0, %ymm1 # sched: [8:0.50] -; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pandn: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpandn %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; BROADWELL-NEXT: vpandn (%rdi), %ymm0, %ymm1 # sched: [7:0.50] -; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pandn: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpandn %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: vpandn (%rdi), %ymm0, %ymm1 # sched: [8:0.50] -; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pandn: -; SKX: # %bb.0: -; SKX-NEXT: vpandn %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: vpandn (%rdi), %ymm0, %ymm1 # sched: [8:0.50] -; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pandn: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpandn %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpandn (%rdi), %ymm0, %ymm1 # sched: [8:0.50] -; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = xor <4 x i64> %a0, <i64 -1, i64 -1, i64 -1, i64 -1> - %2 = and <4 x i64> %a1, %1 - %3 = load <4 x i64>, <4 x i64> *%a2, align 32 - %4 = xor <4 x i64> %2, <i64 -1, i64 -1, i64 -1, i64 -1> - %5 = and <4 x i64> %3, %4 - %6 = add <4 x i64> %2, %5 - ret <4 x i64> %6 -} - -define <32 x i8> @test_pavgb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_pavgb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpavgb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpavgb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pavgb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpavgb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpavgb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pavgb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpavgb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpavgb (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pavgb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpavgb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpavgb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pavgb: -; SKX: # %bb.0: -; SKX-NEXT: vpavgb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpavgb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pavgb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpavgb %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpavgb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = zext <32 x i8> %a0 to <32 x i16> - %2 = zext <32 x i8> %a1 to <32 x i16> - %3 = add <32 x i16> %1, %2 - %4 = add <32 x i16> %3, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> - %5 = lshr <32 x i16> %4, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> - %6 = trunc <32 x i16> %5 to <32 x i8> - %7 = load <32 x i8>, <32 x i8> *%a2, align 32 - %8 = zext <32 x i8> %6 to <32 x i16> - %9 = zext <32 x i8> %7 to <32 x i16> - %10 = add <32 x i16> %8, %9 - %11 = add <32 x i16> %10, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> - %12 = lshr <32 x i16> %11, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> - %13 = trunc <32 x i16> %12 to <32 x i8> - ret <32 x i8> %13 -} - -define <16 x i16> @test_pavgw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_pavgw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpavgw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpavgw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pavgw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpavgw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpavgw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pavgw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpavgw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpavgw (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pavgw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpavgw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpavgw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pavgw: -; SKX: # %bb.0: -; SKX-NEXT: vpavgw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpavgw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pavgw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpavgw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpavgw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = zext <16 x i16> %a0 to <16 x i32> - %2 = zext <16 x i16> %a1 to <16 x i32> - %3 = add <16 x i32> %1, %2 - %4 = add <16 x i32> %3, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> - %5 = lshr <16 x i32> %4, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> - %6 = trunc <16 x i32> %5 to <16 x i16> - %7 = load <16 x i16>, <16 x i16> *%a2, align 32 - %8 = zext <16 x i16> %6 to <16 x i32> - %9 = zext <16 x i16> %7 to <16 x i32> - %10 = add <16 x i32> %8, %9 - %11 = add <16 x i32> %10, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> - %12 = lshr <16 x i32> %11, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> - %13 = trunc <16 x i32> %12 to <16 x i16> - ret <16 x i16> %13 -} - -define <4 x i32> @test_pblendd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { -; GENERIC-LABEL: test_pblendd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3] sched: [1:0.50] -; GENERIC-NEXT: vpblendd {{.*#+}} xmm1 = mem[0],xmm1[1],mem[2],xmm1[3] sched: [7:0.50] -; GENERIC-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pblendd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3] sched: [1:0.33] -; HASWELL-NEXT: vpblendd {{.*#+}} xmm1 = mem[0],xmm1[1],mem[2],xmm1[3] sched: [7:0.50] -; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pblendd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3] sched: [1:0.33] -; BROADWELL-NEXT: vpblendd {{.*#+}} xmm1 = mem[0],xmm1[1],mem[2],xmm1[3] sched: [6:0.50] -; BROADWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pblendd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3] sched: [1:0.33] -; SKYLAKE-NEXT: vpblendd {{.*#+}} xmm1 = mem[0],xmm1[1],mem[2],xmm1[3] sched: [7:0.50] -; SKYLAKE-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pblendd: -; SKX: # %bb.0: -; SKX-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3] sched: [1:0.33] -; SKX-NEXT: vpblendd {{.*#+}} xmm1 = mem[0],xmm1[1],mem[2],xmm1[3] sched: [7:0.50] -; SKX-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pblendd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3] sched: [1:0.50] -; ZNVER1-NEXT: vpblendd {{.*#+}} xmm1 = mem[0],xmm1[1],mem[2],xmm1[3] sched: [8:1.00] -; ZNVER1-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <4 x i32> %a0, <4 x i32> %a1, <4 x i32> <i32 4, i32 5, i32 6, i32 3> - %2 = load <4 x i32>, <4 x i32> *%a2, align 16 - %3 = shufflevector <4 x i32> %a1, <4 x i32> %2, <4 x i32> <i32 4, i32 1, i32 6, i32 3> - %4 = add <4 x i32> %1, %3 - ret <4 x i32> %4 -} - -define <8 x i32> @test_pblendd_ymm(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_pblendd_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7] sched: [1:0.50] -; GENERIC-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0],mem[1,2],ymm1[3,4,5,6,7] sched: [8:0.50] -; GENERIC-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pblendd_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7] sched: [1:0.33] -; HASWELL-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0],mem[1,2],ymm1[3,4,5,6,7] sched: [8:0.50] -; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pblendd_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7] sched: [1:0.33] -; BROADWELL-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0],mem[1,2],ymm1[3,4,5,6,7] sched: [7:0.50] -; BROADWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pblendd_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7] sched: [1:0.33] -; SKYLAKE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0],mem[1,2],ymm1[3,4,5,6,7] sched: [8:0.50] -; SKYLAKE-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pblendd_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7] sched: [1:0.33] -; SKX-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0],mem[1,2],ymm1[3,4,5,6,7] sched: [8:0.50] -; SKX-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pblendd_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7] sched: [1:0.50] -; ZNVER1-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0],mem[1,2],ymm1[3,4,5,6,7] sched: [9:1.50] -; ZNVER1-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <8 x i32> %a0, <8 x i32> %a1, <8 x i32> <i32 8, i32 9, i32 10, i32 3, i32 4, i32 5, i32 6, i32 15> - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = shufflevector <8 x i32> %a1, <8 x i32> %2, <8 x i32> <i32 0, i32 9, i32 10, i32 3, i32 4, i32 5, i32 6, i32 7> - %4 = add <8 x i32> %1, %3 - ret <8 x i32> %4 -} - -define <32 x i8> @test_pblendvb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> %a2, <32 x i8> *%a3, <32 x i8> %a4) { -; GENERIC-LABEL: test_pblendvb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpblendvb %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:1.00] -; GENERIC-NEXT: vpblendvb %ymm3, (%rdi), %ymm0, %ymm0 # sched: [9:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pblendvb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpblendvb %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:2.00] -; HASWELL-NEXT: vpblendvb %ymm3, (%rdi), %ymm0, %ymm0 # sched: [9:2.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pblendvb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpblendvb %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:2.00] -; BROADWELL-NEXT: vpblendvb %ymm3, (%rdi), %ymm0, %ymm0 # sched: [8:2.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pblendvb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpblendvb %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:0.67] -; SKYLAKE-NEXT: vpblendvb %ymm3, (%rdi), %ymm0, %ymm0 # sched: [8:0.67] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pblendvb: -; SKX: # %bb.0: -; SKX-NEXT: vpblendvb %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:0.67] -; SKX-NEXT: vpblendvb %ymm3, (%rdi), %ymm0, %ymm0 # sched: [8:0.67] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pblendvb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpblendvb %ymm2, %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; ZNVER1-NEXT: vpblendvb %ymm3, (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <32 x i8> @llvm.x86.avx2.pblendvb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> %a2) - %2 = load <32 x i8>, <32 x i8> *%a3, align 32 - %3 = call <32 x i8> @llvm.x86.avx2.pblendvb(<32 x i8> %1, <32 x i8> %2, <32 x i8> %a4) - ret <32 x i8> %3 -} -declare <32 x i8> @llvm.x86.avx2.pblendvb(<32 x i8>, <32 x i8>, <32 x i8>) nounwind readnone - -define <16 x i16> @test_pblendw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_pblendw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7,8,9],ymm1[10,11,12],ymm0[13,14,15] sched: [1:0.50] -; GENERIC-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm1[1],mem[2],ymm1[3],mem[4],ymm1[5],mem[6],ymm1[7],mem[8],ymm1[9],mem[10],ymm1[11],mem[12],ymm1[13],mem[14],ymm1[15] sched: [8:0.50] -; GENERIC-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pblendw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7,8,9],ymm1[10,11,12],ymm0[13,14,15] sched: [1:1.00] -; HASWELL-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm1[1],mem[2],ymm1[3],mem[4],ymm1[5],mem[6],ymm1[7],mem[8],ymm1[9],mem[10],ymm1[11],mem[12],ymm1[13],mem[14],ymm1[15] sched: [8:1.00] -; HASWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pblendw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7,8,9],ymm1[10,11,12],ymm0[13,14,15] sched: [1:1.00] -; BROADWELL-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm1[1],mem[2],ymm1[3],mem[4],ymm1[5],mem[6],ymm1[7],mem[8],ymm1[9],mem[10],ymm1[11],mem[12],ymm1[13],mem[14],ymm1[15] sched: [7:1.00] -; BROADWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pblendw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7,8,9],ymm1[10,11,12],ymm0[13,14,15] sched: [1:1.00] -; SKYLAKE-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm1[1],mem[2],ymm1[3],mem[4],ymm1[5],mem[6],ymm1[7],mem[8],ymm1[9],mem[10],ymm1[11],mem[12],ymm1[13],mem[14],ymm1[15] sched: [8:1.00] -; SKYLAKE-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pblendw: -; SKX: # %bb.0: -; SKX-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7,8,9],ymm1[10,11,12],ymm0[13,14,15] sched: [1:1.00] -; SKX-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm1[1],mem[2],ymm1[3],mem[4],ymm1[5],mem[6],ymm1[7],mem[8],ymm1[9],mem[10],ymm1[11],mem[12],ymm1[13],mem[14],ymm1[15] sched: [8:1.00] -; SKX-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pblendw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7,8,9],ymm1[10,11,12],ymm0[13,14,15] sched: [2:0.33] -; ZNVER1-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm1[1],mem[2],ymm1[3],mem[4],ymm1[5],mem[6],ymm1[7],mem[8],ymm1[9],mem[10],ymm1[11],mem[12],ymm1[13],mem[14],ymm1[15] sched: [9:0.50] -; ZNVER1-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <16 x i16> %a0, <16 x i16> %a1, <16 x i32> <i32 0, i32 1, i32 18, i32 19, i32 20, i32 5, i32 6, i32 7, i32 8, i32 9, i32 26, i32 27, i32 28, i32 13, i32 14, i32 15> - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = shufflevector <16 x i16> %a1, <16 x i16> %2, <16 x i32> <i32 16, i32 1, i32 18, i32 3, i32 20, i32 5, i32 22, i32 7, i32 24, i32 9, i32 26, i32 11, i32 28, i32 13, i32 30, i32 15> - %4 = add <16 x i16> %1, %3 - ret <16 x i16> %4 -} - -define <16 x i8> @test_pbroadcastb(<16 x i8> %a0, <16 x i8> *%a1) { -; GENERIC-LABEL: test_pbroadcastb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpbroadcastb %xmm0, %xmm0 # sched: [1:0.50] -; GENERIC-NEXT: vpbroadcastb (%rdi), %xmm1 # sched: [7:0.50] -; GENERIC-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pbroadcastb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpbroadcastb %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vpbroadcastb (%rdi), %xmm1 # sched: [9:1.00] -; HASWELL-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pbroadcastb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpbroadcastb (%rdi), %xmm1 # sched: [9:1.00] -; BROADWELL-NEXT: vpbroadcastb %xmm0, %xmm0 # sched: [3:1.00] -; BROADWELL-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pbroadcastb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpbroadcastb %xmm0, %xmm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vpbroadcastb (%rdi), %xmm1 # sched: [7:1.00] -; SKYLAKE-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pbroadcastb: -; SKX: # %bb.0: -; SKX-NEXT: vpbroadcastb %xmm0, %xmm0 # sched: [3:1.00] -; SKX-NEXT: vpbroadcastb (%rdi), %xmm1 # sched: [7:1.00] -; SKX-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pbroadcastb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpbroadcastb (%rdi), %xmm1 # sched: [8:1.00] -; ZNVER1-NEXT: vpbroadcastb %xmm0, %xmm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <16 x i32> zeroinitializer - %2 = load <16 x i8>, <16 x i8> *%a1, align 16 - %3 = shufflevector <16 x i8> %2, <16 x i8> undef, <16 x i32> zeroinitializer - %4 = add <16 x i8> %1, %3 - ret <16 x i8> %4 -} - -define <32 x i8> @test_pbroadcastb_ymm(<32 x i8> %a0, <32 x i8> *%a1) { -; GENERIC-LABEL: test_pbroadcastb_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpbroadcastb %xmm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpbroadcastb (%rdi), %ymm1 # sched: [7:0.50] -; GENERIC-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pbroadcastb_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpbroadcastb %xmm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vpbroadcastb (%rdi), %ymm1 # sched: [9:1.00] -; HASWELL-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pbroadcastb_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpbroadcastb (%rdi), %ymm1 # sched: [9:1.00] -; BROADWELL-NEXT: vpbroadcastb %xmm0, %ymm0 # sched: [3:1.00] -; BROADWELL-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pbroadcastb_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpbroadcastb %xmm0, %ymm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vpbroadcastb (%rdi), %ymm1 # sched: [8:1.00] -; SKYLAKE-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pbroadcastb_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vpbroadcastb %xmm0, %ymm0 # sched: [3:1.00] -; SKX-NEXT: vpbroadcastb (%rdi), %ymm1 # sched: [8:1.00] -; SKX-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pbroadcastb_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpbroadcastb (%rdi), %ymm1 # sched: [8:2.00] -; ZNVER1-NEXT: vpbroadcastb %xmm0, %ymm0 # sched: [2:0.25] -; ZNVER1-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <32 x i8> %a0, <32 x i8> undef, <32 x i32> zeroinitializer - %2 = load <32 x i8>, <32 x i8> *%a1, align 32 - %3 = shufflevector <32 x i8> %2, <32 x i8> undef, <32 x i32> zeroinitializer - %4 = add <32 x i8> %1, %3 - ret <32 x i8> %4 -} - -define <4 x i32> @test_pbroadcastd(<4 x i32> %a0, <4 x i32> *%a1) { -; GENERIC-LABEL: test_pbroadcastd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpbroadcastd %xmm0, %xmm0 # sched: [1:0.50] -; GENERIC-NEXT: vpbroadcastd (%rdi), %xmm1 # sched: [7:0.50] -; GENERIC-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pbroadcastd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpbroadcastd %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpbroadcastd (%rdi), %xmm1 # sched: [6:0.50] -; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pbroadcastd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpbroadcastd %xmm0, %xmm0 # sched: [1:1.00] -; BROADWELL-NEXT: vpbroadcastd (%rdi), %xmm1 # sched: [5:0.50] -; BROADWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pbroadcastd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpbroadcastd %xmm0, %xmm0 # sched: [1:1.00] -; SKYLAKE-NEXT: vpbroadcastd (%rdi), %xmm1 # sched: [6:0.50] -; SKYLAKE-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pbroadcastd: -; SKX: # %bb.0: -; SKX-NEXT: vpbroadcastd %xmm0, %xmm0 # sched: [1:1.00] -; SKX-NEXT: vpbroadcastd (%rdi), %xmm1 # sched: [6:0.50] -; SKX-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pbroadcastd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpbroadcastd (%rdi), %xmm1 # sched: [8:0.50] -; ZNVER1-NEXT: vpbroadcastd %xmm0, %xmm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <4 x i32> %a0, <4 x i32> undef, <4 x i32> zeroinitializer - %2 = load <4 x i32>, <4 x i32> *%a1, align 16 - %3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> zeroinitializer - %4 = add <4 x i32> %1, %3 - ret <4 x i32> %4 -} - -define <8 x i32> @test_pbroadcastd_ymm(<8 x i32> %a0, <8 x i32> *%a1) { -; GENERIC-LABEL: test_pbroadcastd_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpbroadcastd %xmm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpbroadcastd (%rdi), %ymm1 # sched: [7:0.50] -; GENERIC-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pbroadcastd_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpbroadcastd %xmm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vpbroadcastd (%rdi), %ymm1 # sched: [7:0.50] -; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pbroadcastd_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpbroadcastd %xmm0, %ymm0 # sched: [3:1.00] -; BROADWELL-NEXT: vpbroadcastd (%rdi), %ymm1 # sched: [6:0.50] -; BROADWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pbroadcastd_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpbroadcastd %xmm0, %ymm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vpbroadcastd (%rdi), %ymm1 # sched: [7:0.50] -; SKYLAKE-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pbroadcastd_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vpbroadcastd %xmm0, %ymm0 # sched: [3:1.00] -; SKX-NEXT: vpbroadcastd (%rdi), %ymm1 # sched: [7:0.50] -; SKX-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pbroadcastd_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpbroadcastd (%rdi), %ymm1 # sched: [8:0.50] -; ZNVER1-NEXT: vpbroadcastd %xmm0, %ymm0 # sched: [2:0.25] -; ZNVER1-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <8 x i32> %a0, <8 x i32> undef, <8 x i32> zeroinitializer - %2 = load <8 x i32>, <8 x i32> *%a1, align 32 - %3 = shufflevector <8 x i32> %2, <8 x i32> undef, <8 x i32> zeroinitializer - %4 = add <8 x i32> %1, %3 - ret <8 x i32> %4 -} - -define <2 x i64> @test_pbroadcastq(<2 x i64> %a0, <2 x i64> *%a1) { -; GENERIC-LABEL: test_pbroadcastq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpbroadcastq %xmm0, %xmm0 # sched: [1:0.50] -; GENERIC-NEXT: vpbroadcastq (%rdi), %xmm1 # sched: [7:0.50] -; GENERIC-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pbroadcastq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpbroadcastq %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpbroadcastq (%rdi), %xmm1 # sched: [6:0.50] -; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pbroadcastq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpbroadcastq %xmm0, %xmm0 # sched: [1:1.00] -; BROADWELL-NEXT: vpbroadcastq (%rdi), %xmm1 # sched: [5:0.50] -; BROADWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pbroadcastq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpbroadcastq %xmm0, %xmm0 # sched: [1:1.00] -; SKYLAKE-NEXT: vpbroadcastq (%rdi), %xmm1 # sched: [6:0.50] -; SKYLAKE-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pbroadcastq: -; SKX: # %bb.0: -; SKX-NEXT: vpbroadcastq %xmm0, %xmm0 # sched: [1:1.00] -; SKX-NEXT: vpbroadcastq (%rdi), %xmm1 # sched: [6:0.50] -; SKX-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pbroadcastq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpbroadcastq (%rdi), %xmm1 # sched: [8:0.50] -; ZNVER1-NEXT: vpbroadcastq %xmm0, %xmm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <2 x i64> %a0, <2 x i64> undef, <2 x i32> zeroinitializer - %2 = load <2 x i64>, <2 x i64> *%a1, align 16 - %3 = shufflevector <2 x i64> %2, <2 x i64> undef, <2 x i32> zeroinitializer - %4 = add <2 x i64> %1, %3 - ret <2 x i64> %4 -} - -define <4 x i64> @test_pbroadcastq_ymm(<4 x i64> %a0, <4 x i64> *%a1) { -; GENERIC-LABEL: test_pbroadcastq_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpbroadcastq %xmm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpbroadcastq (%rdi), %ymm1 # sched: [7:0.50] -; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pbroadcastq_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpbroadcastq %xmm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vpbroadcastq (%rdi), %ymm1 # sched: [7:0.50] -; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pbroadcastq_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpbroadcastq %xmm0, %ymm0 # sched: [3:1.00] -; BROADWELL-NEXT: vpbroadcastq (%rdi), %ymm1 # sched: [6:0.50] -; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pbroadcastq_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpbroadcastq %xmm0, %ymm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vpbroadcastq (%rdi), %ymm1 # sched: [7:0.50] -; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pbroadcastq_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vpbroadcastq %xmm0, %ymm0 # sched: [3:1.00] -; SKX-NEXT: vpbroadcastq (%rdi), %ymm1 # sched: [7:0.50] -; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pbroadcastq_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpbroadcastq (%rdi), %ymm1 # sched: [8:0.50] -; ZNVER1-NEXT: vpbroadcastq %xmm0, %ymm0 # sched: [2:0.25] -; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <4 x i64> %a0, <4 x i64> undef, <4 x i32> zeroinitializer - %2 = load <4 x i64>, <4 x i64> *%a1, align 32 - %3 = shufflevector <4 x i64> %2, <4 x i64> undef, <4 x i32> zeroinitializer - %4 = add <4 x i64> %1, %3 - ret <4 x i64> %4 -} - -define <8 x i16> @test_pbroadcastw(<8 x i16> %a0, <8 x i16> *%a1) { -; GENERIC-LABEL: test_pbroadcastw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpbroadcastw %xmm0, %xmm0 # sched: [1:0.50] -; GENERIC-NEXT: vpbroadcastw (%rdi), %xmm1 # sched: [7:0.50] -; GENERIC-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pbroadcastw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpbroadcastw %xmm0, %xmm0 # sched: [3:1.00] -; HASWELL-NEXT: vpbroadcastw (%rdi), %xmm1 # sched: [9:1.00] -; HASWELL-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pbroadcastw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpbroadcastw (%rdi), %xmm1 # sched: [9:1.00] -; BROADWELL-NEXT: vpbroadcastw %xmm0, %xmm0 # sched: [3:1.00] -; BROADWELL-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pbroadcastw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpbroadcastw %xmm0, %xmm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vpbroadcastw (%rdi), %xmm1 # sched: [7:1.00] -; SKYLAKE-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pbroadcastw: -; SKX: # %bb.0: -; SKX-NEXT: vpbroadcastw %xmm0, %xmm0 # sched: [3:1.00] -; SKX-NEXT: vpbroadcastw (%rdi), %xmm1 # sched: [7:1.00] -; SKX-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pbroadcastw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpbroadcastw (%rdi), %xmm1 # sched: [8:1.00] -; ZNVER1-NEXT: vpbroadcastw %xmm0, %xmm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <8 x i16> %a0, <8 x i16> undef, <8 x i32> zeroinitializer - %2 = load <8 x i16>, <8 x i16> *%a1, align 16 - %3 = shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> zeroinitializer - %4 = add <8 x i16> %1, %3 - ret <8 x i16> %4 -} - -define <16 x i16> @test_pbroadcastw_ymm(<16 x i16> %a0, <16 x i16> *%a1) { -; GENERIC-LABEL: test_pbroadcastw_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpbroadcastw %xmm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpbroadcastw (%rdi), %ymm1 # sched: [7:0.50] -; GENERIC-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pbroadcastw_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpbroadcastw %xmm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vpbroadcastw (%rdi), %ymm1 # sched: [9:1.00] -; HASWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pbroadcastw_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpbroadcastw (%rdi), %ymm1 # sched: [9:1.00] -; BROADWELL-NEXT: vpbroadcastw %xmm0, %ymm0 # sched: [3:1.00] -; BROADWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pbroadcastw_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpbroadcastw %xmm0, %ymm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vpbroadcastw (%rdi), %ymm1 # sched: [8:1.00] -; SKYLAKE-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pbroadcastw_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vpbroadcastw %xmm0, %ymm0 # sched: [3:1.00] -; SKX-NEXT: vpbroadcastw (%rdi), %ymm1 # sched: [8:1.00] -; SKX-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pbroadcastw_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpbroadcastw (%rdi), %ymm1 # sched: [8:2.00] -; ZNVER1-NEXT: vpbroadcastw %xmm0, %ymm0 # sched: [2:0.25] -; ZNVER1-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <16 x i16> %a0, <16 x i16> undef, <16 x i32> zeroinitializer - %2 = load <16 x i16>, <16 x i16> *%a1, align 32 - %3 = shufflevector <16 x i16> %2, <16 x i16> undef, <16 x i32> zeroinitializer - %4 = add <16 x i16> %1, %3 - ret <16 x i16> %4 -} - -define <32 x i8> @test_pcmpeqb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_pcmpeqb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpcmpeqb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pcmpeqb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpcmpeqb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pcmpeqb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpcmpeqb (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pcmpeqb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpcmpeqb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pcmpeqb: -; SKX: # %bb.0: -; SKX-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpcmpeqb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pcmpeqb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpcmpeqb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = icmp eq <32 x i8> %a0, %a1 - %2 = sext <32 x i1> %1 to <32 x i8> - %3 = load <32 x i8>, <32 x i8> *%a2, align 32 - %4 = icmp eq <32 x i8> %2, %3 - %5 = sext <32 x i1> %4 to <32 x i8> - ret <32 x i8> %5 -} - -define <8 x i32> @test_pcmpeqd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_pcmpeqd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpcmpeqd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpcmpeqd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pcmpeqd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpcmpeqd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpcmpeqd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pcmpeqd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpcmpeqd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpcmpeqd (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pcmpeqd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpcmpeqd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpcmpeqd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pcmpeqd: -; SKX: # %bb.0: -; SKX-NEXT: vpcmpeqd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpcmpeqd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pcmpeqd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpcmpeqd %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpcmpeqd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = icmp eq <8 x i32> %a0, %a1 - %2 = sext <8 x i1> %1 to <8 x i32> - %3 = load <8 x i32>, <8 x i32> *%a2, align 32 - %4 = icmp eq <8 x i32> %2, %3 - %5 = sext <8 x i1> %4 to <8 x i32> - ret <8 x i32> %5 -} - -define <4 x i64> @test_pcmpeqq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { -; GENERIC-LABEL: test_pcmpeqq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpcmpeqq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpcmpeqq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pcmpeqq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpcmpeqq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpcmpeqq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pcmpeqq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpcmpeqq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpcmpeqq (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pcmpeqq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpcmpeqq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpcmpeqq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pcmpeqq: -; SKX: # %bb.0: -; SKX-NEXT: vpcmpeqq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpcmpeqq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pcmpeqq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpcmpeqq %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpcmpeqq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = icmp eq <4 x i64> %a0, %a1 - %2 = sext <4 x i1> %1 to <4 x i64> - %3 = load <4 x i64>, <4 x i64> *%a2, align 32 - %4 = icmp eq <4 x i64> %2, %3 - %5 = sext <4 x i1> %4 to <4 x i64> - ret <4 x i64> %5 -} - -define <16 x i16> @test_pcmpeqw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_pcmpeqw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpcmpeqw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pcmpeqw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpcmpeqw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pcmpeqw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpcmpeqw (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pcmpeqw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpcmpeqw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pcmpeqw: -; SKX: # %bb.0: -; SKX-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpcmpeqw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pcmpeqw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpcmpeqw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = icmp eq <16 x i16> %a0, %a1 - %2 = sext <16 x i1> %1 to <16 x i16> - %3 = load <16 x i16>, <16 x i16> *%a2, align 32 - %4 = icmp eq <16 x i16> %2, %3 - %5 = sext <16 x i1> %4 to <16 x i16> - ret <16 x i16> %5 -} - -define <32 x i8> @test_pcmpgtb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_pcmpgtb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpcmpgtb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pcmpgtb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpcmpgtb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pcmpgtb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpcmpgtb (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pcmpgtb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpcmpgtb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pcmpgtb: -; SKX: # %bb.0: -; SKX-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpcmpgtb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pcmpgtb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpcmpgtb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = icmp sgt <32 x i8> %a0, %a1 - %2 = sext <32 x i1> %1 to <32 x i8> - %3 = load <32 x i8>, <32 x i8> *%a2, align 32 - %4 = icmp sgt <32 x i8> %2, %3 - %5 = sext <32 x i1> %4 to <32 x i8> - ret <32 x i8> %5 -} - -define <8 x i32> @test_pcmpgtd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_pcmpgtd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpcmpgtd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pcmpgtd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpcmpgtd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pcmpgtd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpcmpgtd (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pcmpgtd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpcmpgtd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pcmpgtd: -; SKX: # %bb.0: -; SKX-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpcmpgtd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pcmpgtd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpcmpgtd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = icmp sgt <8 x i32> %a0, %a1 - %2 = sext <8 x i1> %1 to <8 x i32> - %3 = load <8 x i32>, <8 x i32> *%a2, align 32 - %4 = icmp sgt <8 x i32> %2, %3 - %5 = sext <8 x i1> %4 to <8 x i32> - ret <8 x i32> %5 -} - -define <4 x i64> @test_pcmpgtq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { -; GENERIC-LABEL: test_pcmpgtq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpcmpgtq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pcmpgtq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: vpcmpgtq (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pcmpgtq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; BROADWELL-NEXT: vpcmpgtq (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pcmpgtq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vpcmpgtq (%rdi), %ymm0, %ymm0 # sched: [10:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pcmpgtq: -; SKX: # %bb.0: -; SKX-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SKX-NEXT: vpcmpgtq (%rdi), %ymm0, %ymm0 # sched: [10:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pcmpgtq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpcmpgtq (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = icmp sgt <4 x i64> %a0, %a1 - %2 = sext <4 x i1> %1 to <4 x i64> - %3 = load <4 x i64>, <4 x i64> *%a2, align 32 - %4 = icmp sgt <4 x i64> %2, %3 - %5 = sext <4 x i1> %4 to <4 x i64> - ret <4 x i64> %5 -} - -define <16 x i16> @test_pcmpgtw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_pcmpgtw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpcmpgtw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pcmpgtw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpcmpgtw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pcmpgtw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpcmpgtw (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pcmpgtw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpcmpgtw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pcmpgtw: -; SKX: # %bb.0: -; SKX-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpcmpgtw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pcmpgtw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpcmpgtw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = icmp sgt <16 x i16> %a0, %a1 - %2 = sext <16 x i1> %1 to <16 x i16> - %3 = load <16 x i16>, <16 x i16> *%a2, align 32 - %4 = icmp sgt <16 x i16> %2, %3 - %5 = sext <16 x i1> %4 to <16 x i16> - ret <16 x i16> %5 -} - -define <4 x i64> @test_perm2i128(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { -; GENERIC-LABEL: test_perm2i128: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [1:1.00] -; GENERIC-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [8:1.00] -; GENERIC-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_perm2i128: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [3:1.00] -; HASWELL-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [10:1.00] -; HASWELL-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_perm2i128: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [3:1.00] -; BROADWELL-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [9:1.00] -; BROADWELL-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_perm2i128: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [3:1.00] -; SKYLAKE-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [10:1.00] -; SKYLAKE-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_perm2i128: -; SKX: # %bb.0: -; SKX-NEXT: vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [3:1.00] -; SKX-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [10:1.00] -; SKX-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_perm2i128: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [2:0.25] -; ZNVER1-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [9:0.50] -; ZNVER1-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <4 x i64> %a0, <4 x i64> %a1, <4 x i32> <i32 2, i32 3, i32 4, i32 5> - %2 = load <4 x i64>, <4 x i64> *%a2, align 32 - %3 = shufflevector <4 x i64> %a0, <4 x i64> %2, <4 x i32> <i32 2, i32 3, i32 4, i32 5> - %4 = add <4 x i64> %1, %3 - ret <4 x i64> %4 -} - -define <8 x i32> @test_permd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_permd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpermd %ymm1, %ymm0, %ymm1 # sched: [1:1.00] -; GENERIC-NEXT: vpermd (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; GENERIC-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_permd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpermd %ymm1, %ymm0, %ymm1 # sched: [3:1.00] -; HASWELL-NEXT: vpermd (%rdi), %ymm0, %ymm0 # sched: [10:1.00] -; HASWELL-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_permd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpermd %ymm1, %ymm0, %ymm1 # sched: [3:1.00] -; BROADWELL-NEXT: vpermd (%rdi), %ymm0, %ymm0 # sched: [9:1.00] -; BROADWELL-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_permd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpermd %ymm1, %ymm0, %ymm1 # sched: [3:1.00] -; SKYLAKE-NEXT: vpermd (%rdi), %ymm0, %ymm0 # sched: [10:1.00] -; SKYLAKE-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_permd: -; SKX: # %bb.0: -; SKX-NEXT: vpermd %ymm1, %ymm0, %ymm1 # sched: [3:1.00] -; SKX-NEXT: vpermd (%rdi), %ymm0, %ymm0 # sched: [10:1.00] -; SKX-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_permd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpermd %ymm1, %ymm0, %ymm1 # sched: [2:0.25] -; ZNVER1-NEXT: vpermd (%rdi), %ymm0, %ymm0 # sched: [9:0.50] -; ZNVER1-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.permd(<8 x i32> %a1, <8 x i32> %a0) - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = call <8 x i32> @llvm.x86.avx2.permd(<8 x i32> %2, <8 x i32> %a0) - %4 = add <8 x i32> %1, %3 - ret <8 x i32> %4 -} -declare <8 x i32> @llvm.x86.avx2.permd(<8 x i32>, <8 x i32>) nounwind readonly - -define <4 x double> @test_permpd(<4 x double> %a0, <4 x double> *%a1) { -; GENERIC-LABEL: test_permpd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [1:1.00] -; GENERIC-NEXT: vpermpd {{.*#+}} ymm1 = mem[0,2,2,3] sched: [8:1.00] -; GENERIC-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_permpd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [3:1.00] -; HASWELL-NEXT: vpermpd {{.*#+}} ymm1 = mem[0,2,2,3] sched: [10:1.00] -; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_permpd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [3:1.00] -; BROADWELL-NEXT: vpermpd {{.*#+}} ymm1 = mem[0,2,2,3] sched: [9:1.00] -; BROADWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_permpd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [3:1.00] -; SKYLAKE-NEXT: vpermpd {{.*#+}} ymm1 = mem[0,2,2,3] sched: [10:1.00] -; SKYLAKE-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_permpd: -; SKX: # %bb.0: -; SKX-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [3:1.00] -; SKX-NEXT: vpermpd {{.*#+}} ymm1 = mem[0,2,2,3] sched: [10:1.00] -; SKX-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_permpd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpermpd {{.*#+}} ymm1 = mem[0,2,2,3] sched: [107:0.50] -; ZNVER1-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [100:0.25] -; ZNVER1-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <4 x double> %a0, <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 2, i32 3> - %2 = load <4 x double>, <4 x double> *%a1, align 32 - %3 = shufflevector <4 x double> %2, <4 x double> undef, <4 x i32> <i32 0, i32 2, i32 2, i32 3> - %4 = fadd <4 x double> %1, %3 - ret <4 x double> %4 -} - -define <8 x float> @test_permps(<8 x i32> %a0, <8 x float> %a1, <8 x float> *%a2) { -; GENERIC-LABEL: test_permps: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpermps %ymm1, %ymm0, %ymm1 # sched: [1:1.00] -; GENERIC-NEXT: vpermps (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; GENERIC-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_permps: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpermps %ymm1, %ymm0, %ymm1 # sched: [3:1.00] -; HASWELL-NEXT: vpermps (%rdi), %ymm0, %ymm0 # sched: [10:1.00] -; HASWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_permps: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpermps %ymm1, %ymm0, %ymm1 # sched: [3:1.00] -; BROADWELL-NEXT: vpermps (%rdi), %ymm0, %ymm0 # sched: [9:1.00] -; BROADWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_permps: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpermps %ymm1, %ymm0, %ymm1 # sched: [3:1.00] -; SKYLAKE-NEXT: vpermps (%rdi), %ymm0, %ymm0 # sched: [10:1.00] -; SKYLAKE-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [4:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_permps: -; SKX: # %bb.0: -; SKX-NEXT: vpermps %ymm1, %ymm0, %ymm1 # sched: [3:1.00] -; SKX-NEXT: vpermps (%rdi), %ymm0, %ymm0 # sched: [10:1.00] -; SKX-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [4:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_permps: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpermps %ymm1, %ymm0, %ymm1 # sched: [100:0.25] -; ZNVER1-NEXT: vpermps (%rdi), %ymm0, %ymm0 # sched: [107:0.50] -; ZNVER1-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x float> @llvm.x86.avx2.permps(<8 x float> %a1, <8 x i32> %a0) - %2 = load <8 x float>, <8 x float> *%a2, align 32 - %3 = call <8 x float> @llvm.x86.avx2.permps(<8 x float> %2, <8 x i32> %a0) - %4 = fadd <8 x float> %1, %3 - ret <8 x float> %4 -} -declare <8 x float> @llvm.x86.avx2.permps(<8 x float>, <8 x i32>) nounwind readonly - -define <4 x i64> @test_permq(<4 x i64> %a0, <4 x i64> *%a1) { -; GENERIC-LABEL: test_permq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [1:1.00] -; GENERIC-NEXT: vpermq {{.*#+}} ymm1 = mem[0,2,2,3] sched: [8:1.00] -; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_permq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [3:1.00] -; HASWELL-NEXT: vpermq {{.*#+}} ymm1 = mem[0,2,2,3] sched: [10:1.00] -; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_permq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [3:1.00] -; BROADWELL-NEXT: vpermq {{.*#+}} ymm1 = mem[0,2,2,3] sched: [9:1.00] -; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_permq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [3:1.00] -; SKYLAKE-NEXT: vpermq {{.*#+}} ymm1 = mem[0,2,2,3] sched: [10:1.00] -; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_permq: -; SKX: # %bb.0: -; SKX-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [3:1.00] -; SKX-NEXT: vpermq {{.*#+}} ymm1 = mem[0,2,2,3] sched: [10:1.00] -; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_permq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpermq {{.*#+}} ymm1 = mem[0,2,2,3] sched: [9:0.50] -; ZNVER1-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [2:0.25] -; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <4 x i64> %a0, <4 x i64> undef, <4 x i32> <i32 3, i32 2, i32 2, i32 3> - %2 = load <4 x i64>, <4 x i64> *%a1, align 32 - %3 = shufflevector <4 x i64> %2, <4 x i64> undef, <4 x i32> <i32 0, i32 2, i32 2, i32 3> - %4 = add <4 x i64> %1, %3 - ret <4 x i64> %4 -} - -define <4 x i32> @test_pgatherdd(<4 x i32> %a0, i8* %a1, <4 x i32> %a2, <4 x i32> %a3) { -; GENERIC-LABEL: test_pgatherdd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpgatherdd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pgatherdd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpgatherdd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [26:2.67] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pgatherdd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpgatherdd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pgatherdd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpgatherdd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pgatherdd: -; SKX: # %bb.0: -; SKX-NEXT: vpgatherdd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pgatherdd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpgatherdd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i32> @llvm.x86.avx2.gather.d.d(<4 x i32> %a0, i8* %a1, <4 x i32> %a2, <4 x i32> %a3, i8 2) - ret <4 x i32> %1 -} -declare <4 x i32> @llvm.x86.avx2.gather.d.d(<4 x i32>, i8*, <4 x i32>, <4 x i32>, i8) nounwind readonly - -define <8 x i32> @test_pgatherdd_ymm(<8 x i32> %a0, i8* %a1, <8 x i32> %a2, <8 x i32> %a3) { -; GENERIC-LABEL: test_pgatherdd_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpgatherdd %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [5:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pgatherdd_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpgatherdd %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [27:6.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pgatherdd_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpgatherdd %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [5:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pgatherdd_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpgatherdd %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [25:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pgatherdd_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vpgatherdd %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [25:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pgatherdd_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpgatherdd %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.gather.d.d.256(<8 x i32> %a0, i8* %a1, <8 x i32> %a2, <8 x i32> %a3, i8 2) - ret <8 x i32> %1 -} -declare <8 x i32> @llvm.x86.avx2.gather.d.d.256(<8 x i32>, i8*, <8 x i32>, <8 x i32>, i8) nounwind readonly - -define <2 x i64> @test_pgatherdq(<2 x i64> %a0, i8* %a1, <4 x i32> %a2, <2 x i64> %a3) { -; GENERIC-LABEL: test_pgatherdq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpgatherdq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pgatherdq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpgatherdq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [26:2.67] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pgatherdq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpgatherdq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pgatherdq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpgatherdq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pgatherdq: -; SKX: # %bb.0: -; SKX-NEXT: vpgatherdq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pgatherdq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpgatherdq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <2 x i64> @llvm.x86.avx2.gather.d.q(<2 x i64> %a0, i8* %a1, <4 x i32> %a2, <2 x i64> %a3, i8 2) - ret <2 x i64> %1 -} -declare <2 x i64> @llvm.x86.avx2.gather.d.q(<2 x i64>, i8*, <4 x i32>, <2 x i64>, i8) nounwind readonly - -define <4 x i64> @test_pgatherdq_ymm(<4 x i64> %a0, i8* %a1, <4 x i32> %a2, <4 x i64> %a3) { -; GENERIC-LABEL: test_pgatherdq_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpgatherdq %ymm2, (%rdi,%xmm1,2), %ymm0 # sched: [5:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pgatherdq_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpgatherdq %ymm2, (%rdi,%xmm1,2), %ymm0 # sched: [27:4.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pgatherdq_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpgatherdq %ymm2, (%rdi,%xmm1,2), %ymm0 # sched: [5:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pgatherdq_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpgatherdq %ymm2, (%rdi,%xmm1,2), %ymm0 # sched: [25:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pgatherdq_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vpgatherdq %ymm2, (%rdi,%xmm1,2), %ymm0 # sched: [25:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pgatherdq_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpgatherdq %ymm2, (%rdi,%xmm1,2), %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i64> @llvm.x86.avx2.gather.d.q.256(<4 x i64> %a0, i8* %a1, <4 x i32> %a2, <4 x i64> %a3, i8 2) - ret <4 x i64> %1 -} -declare <4 x i64> @llvm.x86.avx2.gather.d.q.256(<4 x i64>, i8*, <4 x i32>, <4 x i64>, i8) nounwind readonly - -define <4 x i32> @test_pgatherqd(<4 x i32> %a0, i8* %a1, <2 x i64> %a2, <4 x i32> %a3) { -; GENERIC-LABEL: test_pgatherqd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpgatherqd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pgatherqd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpgatherqd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [25:5.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pgatherqd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpgatherqd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pgatherqd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpgatherqd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pgatherqd: -; SKX: # %bb.0: -; SKX-NEXT: vpgatherqd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pgatherqd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpgatherqd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i32> @llvm.x86.avx2.gather.q.d(<4 x i32> %a0, i8* %a1, <2 x i64> %a2, <4 x i32> %a3, i8 2) - ret <4 x i32> %1 -} -declare <4 x i32> @llvm.x86.avx2.gather.q.d(<4 x i32>, i8*, <2 x i64>, <4 x i32>, i8) nounwind readonly - -define <4 x i32> @test_pgatherqd_ymm(<4 x i32> %a0, i8* %a1, <4 x i64> %a2, <4 x i32> %a3) { -; GENERIC-LABEL: test_pgatherqd_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [5:0.50] -; GENERIC-NEXT: vzeroupper # sched: [1:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pgatherqd_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [28:5.00] -; HASWELL-NEXT: vzeroupper # sched: [0:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pgatherqd_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [5:0.50] -; BROADWELL-NEXT: vzeroupper # sched: [0:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pgatherqd_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [25:1.00] -; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pgatherqd_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [25:1.00] -; SKX-NEXT: vzeroupper # sched: [0:0.67] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pgatherqd_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [100:0.25] -; ZNVER1-NEXT: vzeroupper # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i32> @llvm.x86.avx2.gather.q.d.256(<4 x i32> %a0, i8* %a1, <4 x i64> %a2, <4 x i32> %a3, i8 2) - ret <4 x i32> %1 -} -declare <4 x i32> @llvm.x86.avx2.gather.q.d.256(<4 x i32>, i8*, <4 x i64>, <4 x i32>, i8) nounwind readonly - -define <2 x i64> @test_pgatherqq(<2 x i64> %a0, i8 *%a1, <2 x i64> %a2, <2 x i64> %a3) { -; GENERIC-LABEL: test_pgatherqq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpgatherqq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pgatherqq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpgatherqq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [23:3.33] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pgatherqq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpgatherqq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pgatherqq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpgatherqq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pgatherqq: -; SKX: # %bb.0: -; SKX-NEXT: vpgatherqq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pgatherqq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpgatherqq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <2 x i64> @llvm.x86.avx2.gather.q.q(<2 x i64> %a0, i8* %a1, <2 x i64> %a2, <2 x i64> %a3, i8 2) - ret <2 x i64> %1 -} -declare <2 x i64> @llvm.x86.avx2.gather.q.q(<2 x i64>, i8*, <2 x i64>, <2 x i64>, i8) nounwind readonly - -define <4 x i64> @test_pgatherqq_ymm(<4 x i64> %a0, i8 *%a1, <4 x i64> %a2, <4 x i64> %a3) { -; GENERIC-LABEL: test_pgatherqq_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpgatherqq %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [5:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pgatherqq_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpgatherqq %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [24:5.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pgatherqq_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpgatherqq %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [5:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pgatherqq_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpgatherqq %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [25:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pgatherqq_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vpgatherqq %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [25:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pgatherqq_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpgatherqq %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i64> @llvm.x86.avx2.gather.q.q.256(<4 x i64> %a0, i8* %a1, <4 x i64> %a2, <4 x i64> %a3, i8 2) - ret <4 x i64> %1 -} -declare <4 x i64> @llvm.x86.avx2.gather.q.q.256(<4 x i64>, i8*, <4 x i64>, <4 x i64>, i8) nounwind readonly - -define <8 x i32> @test_phaddd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_phaddd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vphaddd %ymm1, %ymm0, %ymm0 # sched: [3:1.50] -; GENERIC-NEXT: vphaddd (%rdi), %ymm0, %ymm0 # sched: [10:1.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_phaddd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vphaddd %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; HASWELL-NEXT: vphaddd (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_phaddd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vphaddd %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; BROADWELL-NEXT: vphaddd (%rdi), %ymm0, %ymm0 # sched: [9:2.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_phaddd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vphaddd %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; SKYLAKE-NEXT: vphaddd (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_phaddd: -; SKX: # %bb.0: -; SKX-NEXT: vphaddd %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; SKX-NEXT: vphaddd (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_phaddd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vphaddd %ymm1, %ymm0, %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: vphaddd (%rdi), %ymm0, %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.phadd.d(<8 x i32> %a0, <8 x i32> %a1) - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = call <8 x i32> @llvm.x86.avx2.phadd.d(<8 x i32> %1, <8 x i32> %2) - ret <8 x i32> %3 -} -declare <8 x i32> @llvm.x86.avx2.phadd.d(<8 x i32>, <8 x i32>) nounwind readnone - -define <16 x i16> @test_phaddsw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_phaddsw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vphaddsw %ymm1, %ymm0, %ymm0 # sched: [3:1.50] -; GENERIC-NEXT: vphaddsw (%rdi), %ymm0, %ymm0 # sched: [10:1.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_phaddsw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vphaddsw %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; HASWELL-NEXT: vphaddsw (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_phaddsw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vphaddsw %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; BROADWELL-NEXT: vphaddsw (%rdi), %ymm0, %ymm0 # sched: [9:2.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_phaddsw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vphaddsw %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; SKYLAKE-NEXT: vphaddsw (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_phaddsw: -; SKX: # %bb.0: -; SKX-NEXT: vphaddsw %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; SKX-NEXT: vphaddsw (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_phaddsw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vphaddsw %ymm1, %ymm0, %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: vphaddsw (%rdi), %ymm0, %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.phadd.sw(<16 x i16> %a0, <16 x i16> %a1) - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = call <16 x i16> @llvm.x86.avx2.phadd.sw(<16 x i16> %1, <16 x i16> %2) - ret <16 x i16> %3 -} -declare <16 x i16> @llvm.x86.avx2.phadd.sw(<16 x i16>, <16 x i16>) nounwind readnone - -define <16 x i16> @test_phaddw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_phaddw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vphaddw %ymm1, %ymm0, %ymm0 # sched: [3:1.50] -; GENERIC-NEXT: vphaddw (%rdi), %ymm0, %ymm0 # sched: [10:1.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_phaddw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vphaddw %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; HASWELL-NEXT: vphaddw (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_phaddw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vphaddw %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; BROADWELL-NEXT: vphaddw (%rdi), %ymm0, %ymm0 # sched: [9:2.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_phaddw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vphaddw %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; SKYLAKE-NEXT: vphaddw (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_phaddw: -; SKX: # %bb.0: -; SKX-NEXT: vphaddw %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; SKX-NEXT: vphaddw (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_phaddw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vphaddw %ymm1, %ymm0, %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: vphaddw (%rdi), %ymm0, %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.phadd.w(<16 x i16> %a0, <16 x i16> %a1) - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = call <16 x i16> @llvm.x86.avx2.phadd.w(<16 x i16> %1, <16 x i16> %2) - ret <16 x i16> %3 -} -declare <16 x i16> @llvm.x86.avx2.phadd.w(<16 x i16>, <16 x i16>) nounwind readnone - -define <8 x i32> @test_phsubd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_phsubd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vphsubd %ymm1, %ymm0, %ymm0 # sched: [3:1.50] -; GENERIC-NEXT: vphsubd (%rdi), %ymm0, %ymm0 # sched: [10:1.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_phsubd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vphsubd %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; HASWELL-NEXT: vphsubd (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_phsubd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vphsubd %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; BROADWELL-NEXT: vphsubd (%rdi), %ymm0, %ymm0 # sched: [9:2.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_phsubd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vphsubd %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; SKYLAKE-NEXT: vphsubd (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_phsubd: -; SKX: # %bb.0: -; SKX-NEXT: vphsubd %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; SKX-NEXT: vphsubd (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_phsubd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vphsubd %ymm1, %ymm0, %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: vphsubd (%rdi), %ymm0, %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.phsub.d(<8 x i32> %a0, <8 x i32> %a1) - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = call <8 x i32> @llvm.x86.avx2.phsub.d(<8 x i32> %1, <8 x i32> %2) - ret <8 x i32> %3 -} -declare <8 x i32> @llvm.x86.avx2.phsub.d(<8 x i32>, <8 x i32>) nounwind readnone - -define <16 x i16> @test_phsubsw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_phsubsw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vphsubsw %ymm1, %ymm0, %ymm0 # sched: [3:1.50] -; GENERIC-NEXT: vphsubsw (%rdi), %ymm0, %ymm0 # sched: [10:1.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_phsubsw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vphsubsw %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; HASWELL-NEXT: vphsubsw (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_phsubsw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vphsubsw %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; BROADWELL-NEXT: vphsubsw (%rdi), %ymm0, %ymm0 # sched: [9:2.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_phsubsw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vphsubsw %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; SKYLAKE-NEXT: vphsubsw (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_phsubsw: -; SKX: # %bb.0: -; SKX-NEXT: vphsubsw %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; SKX-NEXT: vphsubsw (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_phsubsw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vphsubsw %ymm1, %ymm0, %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: vphsubsw (%rdi), %ymm0, %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.phsub.sw(<16 x i16> %a0, <16 x i16> %a1) - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = call <16 x i16> @llvm.x86.avx2.phsub.sw(<16 x i16> %1, <16 x i16> %2) - ret <16 x i16> %3 -} -declare <16 x i16> @llvm.x86.avx2.phsub.sw(<16 x i16>, <16 x i16>) nounwind readnone - -define <16 x i16> @test_phsubw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_phsubw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vphsubw %ymm1, %ymm0, %ymm0 # sched: [3:1.50] -; GENERIC-NEXT: vphsubw (%rdi), %ymm0, %ymm0 # sched: [10:1.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_phsubw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vphsubw %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; HASWELL-NEXT: vphsubw (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_phsubw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vphsubw %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; BROADWELL-NEXT: vphsubw (%rdi), %ymm0, %ymm0 # sched: [9:2.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_phsubw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vphsubw %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; SKYLAKE-NEXT: vphsubw (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_phsubw: -; SKX: # %bb.0: -; SKX-NEXT: vphsubw %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; SKX-NEXT: vphsubw (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_phsubw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vphsubw %ymm1, %ymm0, %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: vphsubw (%rdi), %ymm0, %ymm0 # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.phsub.w(<16 x i16> %a0, <16 x i16> %a1) - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = call <16 x i16> @llvm.x86.avx2.phsub.w(<16 x i16> %1, <16 x i16> %2) - ret <16 x i16> %3 -} -declare <16 x i16> @llvm.x86.avx2.phsub.w(<16 x i16>, <16 x i16>) nounwind readnone - -define <16 x i16> @test_pmaddubsw(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_pmaddubsw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; GENERIC-NEXT: vpmaddubsw (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmaddubsw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: vpmaddubsw (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmaddubsw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; BROADWELL-NEXT: vpmaddubsw (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmaddubsw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKYLAKE-NEXT: vpmaddubsw (%rdi), %ymm0, %ymm0 # sched: [11:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmaddubsw: -; SKX: # %bb.0: -; SKX-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKX-NEXT: vpmaddubsw (%rdi), %ymm0, %ymm0 # sched: [11:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmaddubsw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm0 # sched: [4:1.00] -; ZNVER1-NEXT: vpmaddubsw (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.pmadd.ub.sw(<32 x i8> %a0, <32 x i8> %a1) - %2 = bitcast <16 x i16> %1 to <32 x i8> - %3 = load <32 x i8>, <32 x i8> *%a2, align 32 - %4 = call <16 x i16> @llvm.x86.avx2.pmadd.ub.sw(<32 x i8> %2, <32 x i8> %3) - ret <16 x i16> %4 -} -declare <16 x i16> @llvm.x86.avx2.pmadd.ub.sw(<32 x i8>, <32 x i8>) nounwind readnone - -define <8 x i32> @test_pmaddwd(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_pmaddwd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmaddwd %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; GENERIC-NEXT: vpmaddwd (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmaddwd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmaddwd %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: vpmaddwd (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmaddwd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmaddwd %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; BROADWELL-NEXT: vpmaddwd (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmaddwd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmaddwd %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKYLAKE-NEXT: vpmaddwd (%rdi), %ymm0, %ymm0 # sched: [11:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmaddwd: -; SKX: # %bb.0: -; SKX-NEXT: vpmaddwd %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKX-NEXT: vpmaddwd (%rdi), %ymm0, %ymm0 # sched: [11:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmaddwd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmaddwd %ymm1, %ymm0, %ymm0 # sched: [4:1.00] -; ZNVER1-NEXT: vpmaddwd (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.pmadd.wd(<16 x i16> %a0, <16 x i16> %a1) - %2 = bitcast <8 x i32> %1 to <16 x i16> - %3 = load <16 x i16>, <16 x i16> *%a2, align 32 - %4 = call <8 x i32> @llvm.x86.avx2.pmadd.wd(<16 x i16> %2, <16 x i16> %3) - ret <8 x i32> %4 -} -declare <8 x i32> @llvm.x86.avx2.pmadd.wd(<16 x i16>, <16 x i16>) nounwind readnone - -define <4 x i32> @test_pmaskmovd(i8* %a0, <4 x i32> %a1, <4 x i32> %a2) { -; GENERIC-LABEL: test_pmaskmovd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmaskmovd (%rdi), %xmm0, %xmm2 # sched: [8:1.00] -; GENERIC-NEXT: vpmaskmovd %xmm1, %xmm0, (%rdi) # sched: [5:1.00] -; GENERIC-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmaskmovd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmaskmovd (%rdi), %xmm0, %xmm2 # sched: [8:2.00] -; HASWELL-NEXT: vpmaskmovd %xmm1, %xmm0, (%rdi) # sched: [5:1.00] -; HASWELL-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmaskmovd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmaskmovd (%rdi), %xmm0, %xmm2 # sched: [7:2.00] -; BROADWELL-NEXT: vpmaskmovd %xmm1, %xmm0, (%rdi) # sched: [5:1.00] -; BROADWELL-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmaskmovd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmaskmovd (%rdi), %xmm0, %xmm2 # sched: [7:0.50] -; SKYLAKE-NEXT: vpmaskmovd %xmm1, %xmm0, (%rdi) # sched: [2:1.00] -; SKYLAKE-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmaskmovd: -; SKX: # %bb.0: -; SKX-NEXT: vpmaskmovd (%rdi), %xmm0, %xmm2 # sched: [7:0.50] -; SKX-NEXT: vpmaskmovd %xmm1, %xmm0, (%rdi) # sched: [2:1.00] -; SKX-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmaskmovd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmaskmovd (%rdi), %xmm0, %xmm2 # sched: [100:0.25] -; ZNVER1-NEXT: vpmaskmovd %xmm1, %xmm0, (%rdi) # sched: [100:0.25] -; ZNVER1-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i32> @llvm.x86.avx2.maskload.d(i8* %a0, <4 x i32> %a1) - call void @llvm.x86.avx2.maskstore.d(i8* %a0, <4 x i32> %a1, <4 x i32> %a2) - ret <4 x i32> %1 -} -declare <4 x i32> @llvm.x86.avx2.maskload.d(i8*, <4 x i32>) nounwind readonly -declare void @llvm.x86.avx2.maskstore.d(i8*, <4 x i32>, <4 x i32>) nounwind - -define <8 x i32> @test_pmaskmovd_ymm(i8* %a0, <8 x i32> %a1, <8 x i32> %a2) { -; GENERIC-LABEL: test_pmaskmovd_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmaskmovd (%rdi), %ymm0, %ymm2 # sched: [9:1.00] -; GENERIC-NEXT: vpmaskmovd %ymm1, %ymm0, (%rdi) # sched: [5:1.00] -; GENERIC-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmaskmovd_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmaskmovd (%rdi), %ymm0, %ymm2 # sched: [9:2.00] -; HASWELL-NEXT: vpmaskmovd %ymm1, %ymm0, (%rdi) # sched: [5:1.00] -; HASWELL-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmaskmovd_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmaskmovd (%rdi), %ymm0, %ymm2 # sched: [8:2.00] -; BROADWELL-NEXT: vpmaskmovd %ymm1, %ymm0, (%rdi) # sched: [5:1.00] -; BROADWELL-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.33] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmaskmovd_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmaskmovd (%rdi), %ymm0, %ymm2 # sched: [8:0.50] -; SKYLAKE-NEXT: vpmaskmovd %ymm1, %ymm0, (%rdi) # sched: [2:1.00] -; SKYLAKE-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmaskmovd_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vpmaskmovd (%rdi), %ymm0, %ymm2 # sched: [8:0.50] -; SKX-NEXT: vpmaskmovd %ymm1, %ymm0, (%rdi) # sched: [2:1.00] -; SKX-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmaskmovd_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmaskmovd (%rdi), %ymm0, %ymm2 # sched: [100:0.25] -; ZNVER1-NEXT: vpmaskmovd %ymm1, %ymm0, (%rdi) # sched: [100:0.25] -; ZNVER1-NEXT: vmovdqa %ymm2, %ymm0 # sched: [2:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.maskload.d.256(i8* %a0, <8 x i32> %a1) - call void @llvm.x86.avx2.maskstore.d.256(i8* %a0, <8 x i32> %a1, <8 x i32> %a2) - ret <8 x i32> %1 -} -declare <8 x i32> @llvm.x86.avx2.maskload.d.256(i8*, <8 x i32>) nounwind readonly -declare void @llvm.x86.avx2.maskstore.d.256(i8*, <8 x i32>, <8 x i32>) nounwind - -define <2 x i64> @test_pmaskmovq(i8* %a0, <2 x i64> %a1, <2 x i64> %a2) { -; GENERIC-LABEL: test_pmaskmovq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmaskmovq (%rdi), %xmm0, %xmm2 # sched: [8:1.00] -; GENERIC-NEXT: vpmaskmovq %xmm1, %xmm0, (%rdi) # sched: [5:1.00] -; GENERIC-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmaskmovq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmaskmovq (%rdi), %xmm0, %xmm2 # sched: [8:2.00] -; HASWELL-NEXT: vpmaskmovq %xmm1, %xmm0, (%rdi) # sched: [5:1.00] -; HASWELL-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmaskmovq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmaskmovq (%rdi), %xmm0, %xmm2 # sched: [7:2.00] -; BROADWELL-NEXT: vpmaskmovq %xmm1, %xmm0, (%rdi) # sched: [5:1.00] -; BROADWELL-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmaskmovq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmaskmovq (%rdi), %xmm0, %xmm2 # sched: [7:0.50] -; SKYLAKE-NEXT: vpmaskmovq %xmm1, %xmm0, (%rdi) # sched: [2:1.00] -; SKYLAKE-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmaskmovq: -; SKX: # %bb.0: -; SKX-NEXT: vpmaskmovq (%rdi), %xmm0, %xmm2 # sched: [7:0.50] -; SKX-NEXT: vpmaskmovq %xmm1, %xmm0, (%rdi) # sched: [2:1.00] -; SKX-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmaskmovq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmaskmovq (%rdi), %xmm0, %xmm2 # sched: [8:1.00] -; ZNVER1-NEXT: vpmaskmovq %xmm1, %xmm0, (%rdi) # sched: [100:0.25] -; ZNVER1-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <2 x i64> @llvm.x86.avx2.maskload.q(i8* %a0, <2 x i64> %a1) - call void @llvm.x86.avx2.maskstore.q(i8* %a0, <2 x i64> %a1, <2 x i64> %a2) - ret <2 x i64> %1 -} -declare <2 x i64> @llvm.x86.avx2.maskload.q(i8*, <2 x i64>) nounwind readonly -declare void @llvm.x86.avx2.maskstore.q(i8*, <2 x i64>, <2 x i64>) nounwind - -define <4 x i64> @test_pmaskmovq_ymm(i8* %a0, <4 x i64> %a1, <4 x i64> %a2) { -; GENERIC-LABEL: test_pmaskmovq_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmaskmovq (%rdi), %ymm0, %ymm2 # sched: [9:1.00] -; GENERIC-NEXT: vpmaskmovq %ymm1, %ymm0, (%rdi) # sched: [5:1.00] -; GENERIC-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmaskmovq_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmaskmovq (%rdi), %ymm0, %ymm2 # sched: [9:2.00] -; HASWELL-NEXT: vpmaskmovq %ymm1, %ymm0, (%rdi) # sched: [5:1.00] -; HASWELL-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmaskmovq_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmaskmovq (%rdi), %ymm0, %ymm2 # sched: [8:2.00] -; BROADWELL-NEXT: vpmaskmovq %ymm1, %ymm0, (%rdi) # sched: [5:1.00] -; BROADWELL-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.33] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmaskmovq_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmaskmovq (%rdi), %ymm0, %ymm2 # sched: [8:0.50] -; SKYLAKE-NEXT: vpmaskmovq %ymm1, %ymm0, (%rdi) # sched: [2:1.00] -; SKYLAKE-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmaskmovq_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vpmaskmovq (%rdi), %ymm0, %ymm2 # sched: [8:0.50] -; SKX-NEXT: vpmaskmovq %ymm1, %ymm0, (%rdi) # sched: [2:1.00] -; SKX-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmaskmovq_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmaskmovq (%rdi), %ymm0, %ymm2 # sched: [9:1.50] -; ZNVER1-NEXT: vpmaskmovq %ymm1, %ymm0, (%rdi) # sched: [100:0.25] -; ZNVER1-NEXT: vmovdqa %ymm2, %ymm0 # sched: [2:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i64> @llvm.x86.avx2.maskload.q.256(i8* %a0, <4 x i64> %a1) - call void @llvm.x86.avx2.maskstore.q.256(i8* %a0, <4 x i64> %a1, <4 x i64> %a2) - ret <4 x i64> %1 -} -declare <4 x i64> @llvm.x86.avx2.maskload.q.256(i8*, <4 x i64>) nounwind readonly -declare void @llvm.x86.avx2.maskstore.q.256(i8*, <4 x i64>, <4 x i64>) nounwind - -define <32 x i8> @test_pmaxsb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_pmaxsb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmaxsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpmaxsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmaxsb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmaxsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpmaxsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmaxsb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmaxsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpmaxsb (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmaxsb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmaxsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpmaxsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmaxsb: -; SKX: # %bb.0: -; SKX-NEXT: vpmaxsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpmaxsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmaxsb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmaxsb %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpmaxsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <32 x i8> @llvm.x86.avx2.pmaxs.b(<32 x i8> %a0, <32 x i8> %a1) - %2 = load <32 x i8>, <32 x i8> *%a2, align 32 - %3 = call <32 x i8> @llvm.x86.avx2.pmaxs.b(<32 x i8> %1, <32 x i8> %2) - ret <32 x i8> %3 -} -declare <32 x i8> @llvm.x86.avx2.pmaxs.b(<32 x i8>, <32 x i8>) nounwind readnone - -define <8 x i32> @test_pmaxsd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_pmaxsd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmaxsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpmaxsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmaxsd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmaxsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpmaxsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmaxsd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmaxsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpmaxsd (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmaxsd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmaxsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpmaxsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmaxsd: -; SKX: # %bb.0: -; SKX-NEXT: vpmaxsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpmaxsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmaxsd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmaxsd %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpmaxsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.pmaxs.d(<8 x i32> %a0, <8 x i32> %a1) - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = call <8 x i32> @llvm.x86.avx2.pmaxs.d(<8 x i32> %1, <8 x i32> %2) - ret <8 x i32> %3 -} -declare <8 x i32> @llvm.x86.avx2.pmaxs.d(<8 x i32>, <8 x i32>) nounwind readnone - -define <16 x i16> @test_pmaxsw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_pmaxsw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmaxsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpmaxsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmaxsw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmaxsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpmaxsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmaxsw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmaxsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpmaxsw (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmaxsw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmaxsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpmaxsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmaxsw: -; SKX: # %bb.0: -; SKX-NEXT: vpmaxsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpmaxsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmaxsw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmaxsw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpmaxsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.pmaxs.w(<16 x i16> %a0, <16 x i16> %a1) - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = call <16 x i16> @llvm.x86.avx2.pmaxs.w(<16 x i16> %1, <16 x i16> %2) - ret <16 x i16> %3 -} -declare <16 x i16> @llvm.x86.avx2.pmaxs.w(<16 x i16>, <16 x i16>) nounwind readnone - -define <32 x i8> @test_pmaxub(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_pmaxub: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmaxub %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpmaxub (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmaxub: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmaxub %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpmaxub (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmaxub: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmaxub %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpmaxub (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmaxub: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmaxub %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpmaxub (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmaxub: -; SKX: # %bb.0: -; SKX-NEXT: vpmaxub %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpmaxub (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmaxub: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmaxub %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpmaxub (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <32 x i8> @llvm.x86.avx2.pmaxu.b(<32 x i8> %a0, <32 x i8> %a1) - %2 = load <32 x i8>, <32 x i8> *%a2, align 32 - %3 = call <32 x i8> @llvm.x86.avx2.pmaxu.b(<32 x i8> %1, <32 x i8> %2) - ret <32 x i8> %3 -} -declare <32 x i8> @llvm.x86.avx2.pmaxu.b(<32 x i8>, <32 x i8>) nounwind readnone - -define <8 x i32> @test_pmaxud(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_pmaxud: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmaxud %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpmaxud (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmaxud: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmaxud %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpmaxud (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmaxud: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmaxud %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpmaxud (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmaxud: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmaxud %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpmaxud (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmaxud: -; SKX: # %bb.0: -; SKX-NEXT: vpmaxud %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpmaxud (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmaxud: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmaxud %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpmaxud (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.pmaxu.d(<8 x i32> %a0, <8 x i32> %a1) - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = call <8 x i32> @llvm.x86.avx2.pmaxu.d(<8 x i32> %1, <8 x i32> %2) - ret <8 x i32> %3 -} -declare <8 x i32> @llvm.x86.avx2.pmaxu.d(<8 x i32>, <8 x i32>) nounwind readnone - -define <16 x i16> @test_pmaxuw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_pmaxuw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmaxuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpmaxuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmaxuw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmaxuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpmaxuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmaxuw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmaxuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpmaxuw (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmaxuw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmaxuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpmaxuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmaxuw: -; SKX: # %bb.0: -; SKX-NEXT: vpmaxuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpmaxuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmaxuw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmaxuw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpmaxuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.pmaxu.w(<16 x i16> %a0, <16 x i16> %a1) - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = call <16 x i16> @llvm.x86.avx2.pmaxu.w(<16 x i16> %1, <16 x i16> %2) - ret <16 x i16> %3 -} -declare <16 x i16> @llvm.x86.avx2.pmaxu.w(<16 x i16>, <16 x i16>) nounwind readnone - -define <32 x i8> @test_pminsb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_pminsb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpminsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpminsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pminsb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpminsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpminsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pminsb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpminsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpminsb (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pminsb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpminsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpminsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pminsb: -; SKX: # %bb.0: -; SKX-NEXT: vpminsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpminsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pminsb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpminsb %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpminsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <32 x i8> @llvm.x86.avx2.pmins.b(<32 x i8> %a0, <32 x i8> %a1) - %2 = load <32 x i8>, <32 x i8> *%a2, align 32 - %3 = call <32 x i8> @llvm.x86.avx2.pmins.b(<32 x i8> %1, <32 x i8> %2) - ret <32 x i8> %3 -} -declare <32 x i8> @llvm.x86.avx2.pmins.b(<32 x i8>, <32 x i8>) nounwind readnone - -define <8 x i32> @test_pminsd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_pminsd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpminsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpminsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pminsd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpminsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpminsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pminsd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpminsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpminsd (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pminsd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpminsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpminsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pminsd: -; SKX: # %bb.0: -; SKX-NEXT: vpminsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpminsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pminsd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpminsd %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpminsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.pmins.d(<8 x i32> %a0, <8 x i32> %a1) - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = call <8 x i32> @llvm.x86.avx2.pmins.d(<8 x i32> %1, <8 x i32> %2) - ret <8 x i32> %3 -} -declare <8 x i32> @llvm.x86.avx2.pmins.d(<8 x i32>, <8 x i32>) nounwind readnone - -define <16 x i16> @test_pminsw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_pminsw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpminsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpminsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pminsw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpminsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpminsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pminsw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpminsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpminsw (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pminsw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpminsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpminsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pminsw: -; SKX: # %bb.0: -; SKX-NEXT: vpminsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpminsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pminsw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpminsw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpminsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.pmins.w(<16 x i16> %a0, <16 x i16> %a1) - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = call <16 x i16> @llvm.x86.avx2.pmins.w(<16 x i16> %1, <16 x i16> %2) - ret <16 x i16> %3 -} -declare <16 x i16> @llvm.x86.avx2.pmins.w(<16 x i16>, <16 x i16>) nounwind readnone - -define <32 x i8> @test_pminub(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_pminub: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpminub %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpminub (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pminub: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpminub %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpminub (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pminub: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpminub %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpminub (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pminub: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpminub %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpminub (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pminub: -; SKX: # %bb.0: -; SKX-NEXT: vpminub %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpminub (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pminub: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpminub %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpminub (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <32 x i8> @llvm.x86.avx2.pminu.b(<32 x i8> %a0, <32 x i8> %a1) - %2 = load <32 x i8>, <32 x i8> *%a2, align 32 - %3 = call <32 x i8> @llvm.x86.avx2.pminu.b(<32 x i8> %1, <32 x i8> %2) - ret <32 x i8> %3 -} -declare <32 x i8> @llvm.x86.avx2.pminu.b(<32 x i8>, <32 x i8>) nounwind readnone - -define <8 x i32> @test_pminud(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_pminud: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpminud %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpminud (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pminud: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpminud %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpminud (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pminud: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpminud %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpminud (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pminud: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpminud %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpminud (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pminud: -; SKX: # %bb.0: -; SKX-NEXT: vpminud %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpminud (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pminud: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpminud %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpminud (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.pminu.d(<8 x i32> %a0, <8 x i32> %a1) - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = call <8 x i32> @llvm.x86.avx2.pminu.d(<8 x i32> %1, <8 x i32> %2) - ret <8 x i32> %3 -} -declare <8 x i32> @llvm.x86.avx2.pminu.d(<8 x i32>, <8 x i32>) nounwind readnone - -define <16 x i16> @test_pminuw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_pminuw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpminuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpminuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pminuw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpminuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpminuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pminuw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpminuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpminuw (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pminuw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpminuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpminuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pminuw: -; SKX: # %bb.0: -; SKX-NEXT: vpminuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpminuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pminuw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpminuw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpminuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.pminu.w(<16 x i16> %a0, <16 x i16> %a1) - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = call <16 x i16> @llvm.x86.avx2.pminu.w(<16 x i16> %1, <16 x i16> %2) - ret <16 x i16> %3 -} -declare <16 x i16> @llvm.x86.avx2.pminu.w(<16 x i16>, <16 x i16>) nounwind readnone - -define i32 @test_pmovmskb(<32 x i8> %a0) { -; GENERIC-LABEL: test_pmovmskb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmovmskb %ymm0, %eax # sched: [2:1.00] -; GENERIC-NEXT: vzeroupper # sched: [1:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmovmskb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmovmskb %ymm0, %eax # sched: [3:1.00] -; HASWELL-NEXT: vzeroupper # sched: [0:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmovmskb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmovmskb %ymm0, %eax # sched: [3:1.00] -; BROADWELL-NEXT: vzeroupper # sched: [0:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmovmskb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmovmskb %ymm0, %eax # sched: [2:1.00] -; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmovmskb: -; SKX: # %bb.0: -; SKX-NEXT: vpmovmskb %ymm0, %eax # sched: [2:1.00] -; SKX-NEXT: vzeroupper # sched: [0:0.67] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmovmskb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmovmskb %ymm0, %eax # sched: [2:2.00] -; ZNVER1-NEXT: vzeroupper # sched: [100:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call i32 @llvm.x86.avx2.pmovmskb(<32 x i8> %a0) - ret i32 %1 -} -declare i32 @llvm.x86.avx2.pmovmskb(<32 x i8>) nounwind readnone - -define <8 x i32> @test_pmovsxbd(<16 x i8> %a0, <16 x i8> *%a1) { -; GENERIC-LABEL: test_pmovsxbd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmovsxbd %xmm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpmovsxbd (%rdi), %ymm1 # sched: [8:1.00] -; GENERIC-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmovsxbd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmovsxbd %xmm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vpmovsxbd (%rdi), %ymm1 # sched: [8:1.00] -; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmovsxbd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmovsxbd %xmm0, %ymm0 # sched: [3:1.00] -; BROADWELL-NEXT: vpmovsxbd (%rdi), %ymm1 # sched: [8:1.00] -; BROADWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmovsxbd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmovsxbd %xmm0, %ymm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vpmovsxbd (%rdi), %ymm1 # sched: [8:1.00] -; SKYLAKE-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmovsxbd: -; SKX: # %bb.0: -; SKX-NEXT: vpmovsxbd %xmm0, %ymm0 # sched: [3:1.00] -; SKX-NEXT: vpmovsxbd (%rdi), %ymm1 # sched: [8:1.00] -; SKX-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmovsxbd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmovsxbd (%rdi), %ymm1 # sched: [8:0.50] -; ZNVER1-NEXT: vpmovsxbd %xmm0, %ymm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> - %2 = sext <8 x i8> %1 to <8 x i32> - %3 = load <16 x i8>, <16 x i8> *%a1, align 16 - %4 = shufflevector <16 x i8> %3, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> - %5 = sext <8 x i8> %4 to <8 x i32> - %6 = add <8 x i32> %2, %5 - ret <8 x i32> %6 -} - -define <4 x i64> @test_pmovsxbq(<16 x i8> %a0, <16 x i8> *%a1) { -; GENERIC-LABEL: test_pmovsxbq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmovsxbq %xmm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpmovsxbq (%rdi), %ymm1 # sched: [8:1.00] -; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmovsxbq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmovsxbq %xmm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vpmovsxbq (%rdi), %ymm1 # sched: [8:1.00] -; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmovsxbq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmovsxbq %xmm0, %ymm0 # sched: [3:1.00] -; BROADWELL-NEXT: vpmovsxbq (%rdi), %ymm1 # sched: [8:1.00] -; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmovsxbq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmovsxbq %xmm0, %ymm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vpmovsxbq (%rdi), %ymm1 # sched: [8:1.00] -; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmovsxbq: -; SKX: # %bb.0: -; SKX-NEXT: vpmovsxbq %xmm0, %ymm0 # sched: [3:1.00] -; SKX-NEXT: vpmovsxbq (%rdi), %ymm1 # sched: [8:1.00] -; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmovsxbq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmovsxbq (%rdi), %ymm1 # sched: [8:0.50] -; ZNVER1-NEXT: vpmovsxbq %xmm0, %ymm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3> - %2 = sext <4 x i8> %1 to <4 x i64> - %3 = load <16 x i8>, <16 x i8> *%a1, align 16 - %4 = shufflevector <16 x i8> %3, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3> - %5 = sext <4 x i8> %4 to <4 x i64> - %6 = add <4 x i64> %2, %5 - ret <4 x i64> %6 -} - -define <16 x i16> @test_pmovsxbw(<16 x i8> %a0, <16 x i8> *%a1) { -; GENERIC-LABEL: test_pmovsxbw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmovsxbw %xmm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpmovsxbw (%rdi), %ymm1 # sched: [8:1.00] -; GENERIC-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmovsxbw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmovsxbw %xmm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vpmovsxbw (%rdi), %ymm1 # sched: [9:1.00] -; HASWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmovsxbw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmovsxbw %xmm0, %ymm0 # sched: [3:1.00] -; BROADWELL-NEXT: vpmovsxbw (%rdi), %ymm1 # sched: [8:1.00] -; BROADWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmovsxbw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmovsxbw %xmm0, %ymm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vpmovsxbw (%rdi), %ymm1 # sched: [9:1.00] -; SKYLAKE-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmovsxbw: -; SKX: # %bb.0: -; SKX-NEXT: vpmovsxbw %xmm0, %ymm0 # sched: [3:1.00] -; SKX-NEXT: vpmovsxbw (%rdi), %ymm1 # sched: [9:1.00] -; SKX-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmovsxbw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmovsxbw (%rdi), %ymm1 # sched: [8:0.50] -; ZNVER1-NEXT: vpmovsxbw %xmm0, %ymm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = sext <16 x i8> %a0 to <16 x i16> - %2 = load <16 x i8>, <16 x i8> *%a1, align 16 - %3 = sext <16 x i8> %2 to <16 x i16> - %4 = add <16 x i16> %1, %3 - ret <16 x i16> %4 -} - -define <4 x i64> @test_pmovsxdq(<4 x i32> %a0, <4 x i32> *%a1) { -; GENERIC-LABEL: test_pmovsxdq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmovsxdq %xmm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpmovsxdq (%rdi), %ymm1 # sched: [8:1.00] -; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmovsxdq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmovsxdq %xmm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vpmovsxdq (%rdi), %ymm1 # sched: [9:1.00] -; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmovsxdq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmovsxdq %xmm0, %ymm0 # sched: [3:1.00] -; BROADWELL-NEXT: vpmovsxdq (%rdi), %ymm1 # sched: [8:1.00] -; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmovsxdq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmovsxdq %xmm0, %ymm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vpmovsxdq (%rdi), %ymm1 # sched: [9:1.00] -; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmovsxdq: -; SKX: # %bb.0: -; SKX-NEXT: vpmovsxdq %xmm0, %ymm0 # sched: [3:1.00] -; SKX-NEXT: vpmovsxdq (%rdi), %ymm1 # sched: [9:1.00] -; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmovsxdq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmovsxdq (%rdi), %ymm1 # sched: [8:0.50] -; ZNVER1-NEXT: vpmovsxdq %xmm0, %ymm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = sext <4 x i32> %a0 to <4 x i64> - %2 = load <4 x i32>, <4 x i32> *%a1, align 16 - %3 = sext <4 x i32> %2 to <4 x i64> - %4 = add <4 x i64> %1, %3 - ret <4 x i64> %4 -} - -define <8 x i32> @test_pmovsxwd(<8 x i16> %a0, <8 x i16> *%a1) { -; GENERIC-LABEL: test_pmovsxwd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmovsxwd %xmm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpmovsxwd (%rdi), %ymm1 # sched: [8:1.00] -; GENERIC-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmovsxwd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmovsxwd %xmm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vpmovsxwd (%rdi), %ymm1 # sched: [9:1.00] -; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmovsxwd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmovsxwd %xmm0, %ymm0 # sched: [3:1.00] -; BROADWELL-NEXT: vpmovsxwd (%rdi), %ymm1 # sched: [8:1.00] -; BROADWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmovsxwd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmovsxwd %xmm0, %ymm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vpmovsxwd (%rdi), %ymm1 # sched: [9:1.00] -; SKYLAKE-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmovsxwd: -; SKX: # %bb.0: -; SKX-NEXT: vpmovsxwd %xmm0, %ymm0 # sched: [3:1.00] -; SKX-NEXT: vpmovsxwd (%rdi), %ymm1 # sched: [9:1.00] -; SKX-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmovsxwd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmovsxwd (%rdi), %ymm1 # sched: [8:0.50] -; ZNVER1-NEXT: vpmovsxwd %xmm0, %ymm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = sext <8 x i16> %a0 to <8 x i32> - %2 = load <8 x i16>, <8 x i16> *%a1, align 16 - %3 = sext <8 x i16> %2 to <8 x i32> - %4 = add <8 x i32> %1, %3 - ret <8 x i32> %4 -} - -define <4 x i64> @test_pmovsxwq(<8 x i16> %a0, <8 x i16> *%a1) { -; GENERIC-LABEL: test_pmovsxwq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmovsxwq %xmm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpmovsxwq (%rdi), %ymm1 # sched: [8:1.00] -; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmovsxwq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmovsxwq %xmm0, %ymm0 # sched: [3:1.00] -; HASWELL-NEXT: vpmovsxwq (%rdi), %ymm1 # sched: [8:1.00] -; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmovsxwq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmovsxwq %xmm0, %ymm0 # sched: [3:1.00] -; BROADWELL-NEXT: vpmovsxwq (%rdi), %ymm1 # sched: [8:1.00] -; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmovsxwq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmovsxwq %xmm0, %ymm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vpmovsxwq (%rdi), %ymm1 # sched: [8:1.00] -; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmovsxwq: -; SKX: # %bb.0: -; SKX-NEXT: vpmovsxwq %xmm0, %ymm0 # sched: [3:1.00] -; SKX-NEXT: vpmovsxwq (%rdi), %ymm1 # sched: [8:1.00] -; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmovsxwq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmovsxwq (%rdi), %ymm1 # sched: [8:0.50] -; ZNVER1-NEXT: vpmovsxwq %xmm0, %ymm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <8 x i16> %a0, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3> - %2 = sext <4 x i16> %1 to <4 x i64> - %3 = load <8 x i16>, <8 x i16> *%a1, align 16 - %4 = shufflevector <8 x i16> %3, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3> - %5 = sext <4 x i16> %4 to <4 x i64> - %6 = add <4 x i64> %2, %5 - ret <4 x i64> %6 -} - -define <8 x i32> @test_pmovzxbd(<16 x i8> %a0, <16 x i8> *%a1) { -; GENERIC-LABEL: test_pmovzxbd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmovzxbd {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero sched: [1:1.00] -; GENERIC-NEXT: vpmovzxbd {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero sched: [8:1.00] -; GENERIC-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmovzxbd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmovzxbd {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero sched: [3:1.00] -; HASWELL-NEXT: vpmovzxbd {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero sched: [10:1.00] -; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmovzxbd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmovzxbd {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero sched: [3:1.00] -; BROADWELL-NEXT: vpmovzxbd {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero sched: [9:1.00] -; BROADWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmovzxbd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmovzxbd {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero sched: [3:1.00] -; SKYLAKE-NEXT: vpmovzxbd {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero sched: [10:1.00] -; SKYLAKE-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmovzxbd: -; SKX: # %bb.0: -; SKX-NEXT: vpmovzxbd {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero sched: [3:1.00] -; SKX-NEXT: vpmovzxbd {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero sched: [10:1.00] -; SKX-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmovzxbd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmovzxbd {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero sched: [8:0.50] -; ZNVER1-NEXT: vpmovzxbd {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero sched: [1:0.50] -; ZNVER1-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> - %2 = zext <8 x i8> %1 to <8 x i32> - %3 = load <16 x i8>, <16 x i8> *%a1, align 16 - %4 = shufflevector <16 x i8> %3, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> - %5 = zext <8 x i8> %4 to <8 x i32> - %6 = add <8 x i32> %2, %5 - ret <8 x i32> %6 -} - -define <4 x i64> @test_pmovzxbq(<16 x i8> %a0, <16 x i8> *%a1) { -; GENERIC-LABEL: test_pmovzxbq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmovzxbq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero sched: [1:1.00] -; GENERIC-NEXT: vpmovzxbq {{.*#+}} ymm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero,mem[2],zero,zero,zero,zero,zero,zero,zero,mem[3],zero,zero,zero,zero,zero,zero,zero sched: [8:1.00] -; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmovzxbq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmovzxbq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero sched: [3:1.00] -; HASWELL-NEXT: vpmovzxbq {{.*#+}} ymm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero,mem[2],zero,zero,zero,zero,zero,zero,zero,mem[3],zero,zero,zero,zero,zero,zero,zero sched: [10:1.00] -; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmovzxbq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmovzxbq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero sched: [3:1.00] -; BROADWELL-NEXT: vpmovzxbq {{.*#+}} ymm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero,mem[2],zero,zero,zero,zero,zero,zero,zero,mem[3],zero,zero,zero,zero,zero,zero,zero sched: [9:1.00] -; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmovzxbq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmovzxbq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero sched: [3:1.00] -; SKYLAKE-NEXT: vpmovzxbq {{.*#+}} ymm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero,mem[2],zero,zero,zero,zero,zero,zero,zero,mem[3],zero,zero,zero,zero,zero,zero,zero sched: [10:1.00] -; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmovzxbq: -; SKX: # %bb.0: -; SKX-NEXT: vpmovzxbq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero sched: [3:1.00] -; SKX-NEXT: vpmovzxbq {{.*#+}} ymm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero,mem[2],zero,zero,zero,zero,zero,zero,zero,mem[3],zero,zero,zero,zero,zero,zero,zero sched: [10:1.00] -; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmovzxbq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmovzxbq {{.*#+}} ymm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero,mem[2],zero,zero,zero,zero,zero,zero,zero,mem[3],zero,zero,zero,zero,zero,zero,zero sched: [8:0.50] -; ZNVER1-NEXT: vpmovzxbq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero sched: [1:0.50] -; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3> - %2 = zext <4 x i8> %1 to <4 x i64> - %3 = load <16 x i8>, <16 x i8> *%a1, align 16 - %4 = shufflevector <16 x i8> %3, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3> - %5 = zext <4 x i8> %4 to <4 x i64> - %6 = add <4 x i64> %2, %5 - ret <4 x i64> %6 -} - -define <16 x i16> @test_pmovzxbw(<16 x i8> %a0, <16 x i8> *%a1) { -; GENERIC-LABEL: test_pmovzxbw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero sched: [1:1.00] -; GENERIC-NEXT: vpmovzxbw {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero sched: [8:1.00] -; GENERIC-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmovzxbw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero sched: [3:1.00] -; HASWELL-NEXT: vpmovzxbw {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero sched: [10:1.00] -; HASWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmovzxbw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero sched: [3:1.00] -; BROADWELL-NEXT: vpmovzxbw {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero sched: [9:1.00] -; BROADWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmovzxbw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero sched: [3:1.00] -; SKYLAKE-NEXT: vpmovzxbw {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero sched: [10:1.00] -; SKYLAKE-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmovzxbw: -; SKX: # %bb.0: -; SKX-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero sched: [3:1.00] -; SKX-NEXT: vpmovzxbw {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero sched: [10:1.00] -; SKX-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmovzxbw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmovzxbw {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero sched: [8:0.50] -; ZNVER1-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero sched: [1:0.50] -; ZNVER1-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = zext <16 x i8> %a0 to <16 x i16> - %2 = load <16 x i8>, <16 x i8> *%a1, align 16 - %3 = zext <16 x i8> %2 to <16 x i16> - %4 = add <16 x i16> %1, %3 - ret <16 x i16> %4 -} - -define <4 x i64> @test_pmovzxdq(<4 x i32> %a0, <4 x i32> *%a1) { -; GENERIC-LABEL: test_pmovzxdq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [1:1.00] -; GENERIC-NEXT: vpmovzxdq {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [8:1.00] -; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmovzxdq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [3:1.00] -; HASWELL-NEXT: vpmovzxdq {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [10:1.00] -; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmovzxdq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [3:1.00] -; BROADWELL-NEXT: vpmovzxdq {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [9:1.00] -; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmovzxdq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [3:1.00] -; SKYLAKE-NEXT: vpmovzxdq {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [10:1.00] -; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmovzxdq: -; SKX: # %bb.0: -; SKX-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [3:1.00] -; SKX-NEXT: vpmovzxdq {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [10:1.00] -; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmovzxdq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmovzxdq {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [8:0.50] -; ZNVER1-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [1:0.50] -; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = zext <4 x i32> %a0 to <4 x i64> - %2 = load <4 x i32>, <4 x i32> *%a1, align 16 - %3 = zext <4 x i32> %2 to <4 x i64> - %4 = add <4 x i64> %1, %3 - ret <4 x i64> %4 -} - -define <8 x i32> @test_pmovzxwd(<8 x i16> %a0, <8 x i16> *%a1) { -; GENERIC-LABEL: test_pmovzxwd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmovzxwd {{.*#+}} ymm0 = 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] -; GENERIC-NEXT: vpmovzxwd {{.*#+}} ymm1 = 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: [8:1.00] -; GENERIC-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmovzxwd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmovzxwd {{.*#+}} ymm0 = 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: [3:1.00] -; HASWELL-NEXT: vpmovzxwd {{.*#+}} ymm1 = 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: [9:1.00] -; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmovzxwd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmovzxwd {{.*#+}} ymm0 = 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: [3:1.00] -; BROADWELL-NEXT: vpmovzxwd {{.*#+}} ymm1 = 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: [8:1.00] -; BROADWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmovzxwd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmovzxwd {{.*#+}} ymm0 = 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: [3:1.00] -; SKYLAKE-NEXT: vpmovzxwd {{.*#+}} ymm1 = 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: [9:1.00] -; SKYLAKE-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmovzxwd: -; SKX: # %bb.0: -; SKX-NEXT: vpmovzxwd {{.*#+}} ymm0 = 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: [3:1.00] -; SKX-NEXT: vpmovzxwd {{.*#+}} ymm1 = 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: [9:1.00] -; SKX-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmovzxwd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmovzxwd {{.*#+}} ymm1 = 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: [8:0.50] -; ZNVER1-NEXT: vpmovzxwd {{.*#+}} ymm0 = 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] -; ZNVER1-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = zext <8 x i16> %a0 to <8 x i32> - %2 = load <8 x i16>, <8 x i16> *%a1, align 16 - %3 = zext <8 x i16> %2 to <8 x i32> - %4 = add <8 x i32> %1, %3 - ret <8 x i32> %4 -} - -define <4 x i64> @test_pmovzxwq(<8 x i16> %a0, <8 x i16> *%a1) { -; GENERIC-LABEL: test_pmovzxwq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmovzxwq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [1:1.00] -; GENERIC-NEXT: vpmovzxwq {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [8:1.00] -; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmovzxwq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmovzxwq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [3:1.00] -; HASWELL-NEXT: vpmovzxwq {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [10:1.00] -; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmovzxwq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmovzxwq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [3:1.00] -; BROADWELL-NEXT: vpmovzxwq {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [9:1.00] -; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmovzxwq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmovzxwq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [3:1.00] -; SKYLAKE-NEXT: vpmovzxwq {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [10:1.00] -; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmovzxwq: -; SKX: # %bb.0: -; SKX-NEXT: vpmovzxwq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [3:1.00] -; SKX-NEXT: vpmovzxwq {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [10:1.00] -; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmovzxwq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmovzxwq {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [8:0.50] -; ZNVER1-NEXT: vpmovzxwq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [1:0.50] -; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <8 x i16> %a0, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3> - %2 = zext <4 x i16> %1 to <4 x i64> - %3 = load <8 x i16>, <8 x i16> *%a1, align 16 - %4 = shufflevector <8 x i16> %3, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3> - %5 = zext <4 x i16> %4 to <4 x i64> - %6 = add <4 x i64> %2, %5 - ret <4 x i64> %6 -} - -define <4 x i64> @test_pmuldq(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> %a2, <8 x i32> *%a3) { -; GENERIC-LABEL: test_pmuldq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmuldq %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; GENERIC-NEXT: vpmuldq (%rdi), %ymm2, %ymm1 # sched: [12:1.00] -; GENERIC-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmuldq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmuldq %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: vpmuldq (%rdi), %ymm2, %ymm1 # sched: [12:1.00] -; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmuldq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmuldq %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; BROADWELL-NEXT: vpmuldq (%rdi), %ymm2, %ymm1 # sched: [11:1.00] -; BROADWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmuldq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmuldq %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKYLAKE-NEXT: vpmuldq (%rdi), %ymm2, %ymm1 # sched: [11:0.50] -; SKYLAKE-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmuldq: -; SKX: # %bb.0: -; SKX-NEXT: vpmuldq %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKX-NEXT: vpmuldq (%rdi), %ymm2, %ymm1 # sched: [11:0.50] -; SKX-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmuldq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmuldq (%rdi), %ymm2, %ymm2 # sched: [11:1.00] -; ZNVER1-NEXT: vpmuldq %ymm1, %ymm0, %ymm0 # sched: [4:1.00] -; ZNVER1-NEXT: vpor %ymm2, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i64> @llvm.x86.avx2.pmul.dq(<8 x i32> %a0, <8 x i32> %a1) - %2 = load <8 x i32>, <8 x i32> *%a3, align 32 - %3 = call <4 x i64> @llvm.x86.avx2.pmul.dq(<8 x i32> %a2, <8 x i32> %2) - %4 = or <4 x i64> %1, %3 - ret <4 x i64> %4 -} -declare <4 x i64> @llvm.x86.avx2.pmul.dq(<8 x i32>, <8 x i32>) nounwind readnone - -define <16 x i16> @test_pmulhrsw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_pmulhrsw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; GENERIC-NEXT: vpmulhrsw (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmulhrsw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: vpmulhrsw (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmulhrsw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; BROADWELL-NEXT: vpmulhrsw (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmulhrsw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKYLAKE-NEXT: vpmulhrsw (%rdi), %ymm0, %ymm0 # sched: [11:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmulhrsw: -; SKX: # %bb.0: -; SKX-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKX-NEXT: vpmulhrsw (%rdi), %ymm0, %ymm0 # sched: [11:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmulhrsw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm0 # sched: [4:1.00] -; ZNVER1-NEXT: vpmulhrsw (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.pmul.hr.sw(<16 x i16> %a0, <16 x i16> %a1) - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = call <16 x i16> @llvm.x86.avx2.pmul.hr.sw(<16 x i16> %1, <16 x i16> %2) - ret <16 x i16> %3 -} -declare <16 x i16> @llvm.x86.avx2.pmul.hr.sw(<16 x i16>, <16 x i16>) nounwind readnone - -define <16 x i16> @test_pmulhuw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_pmulhuw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmulhuw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; GENERIC-NEXT: vpmulhuw (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmulhuw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmulhuw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: vpmulhuw (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmulhuw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmulhuw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; BROADWELL-NEXT: vpmulhuw (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmulhuw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmulhuw %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKYLAKE-NEXT: vpmulhuw (%rdi), %ymm0, %ymm0 # sched: [11:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmulhuw: -; SKX: # %bb.0: -; SKX-NEXT: vpmulhuw %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKX-NEXT: vpmulhuw (%rdi), %ymm0, %ymm0 # sched: [11:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmulhuw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmulhuw %ymm1, %ymm0, %ymm0 # sched: [4:1.00] -; ZNVER1-NEXT: vpmulhuw (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.pmulhu.w(<16 x i16> %a0, <16 x i16> %a1) - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = call <16 x i16> @llvm.x86.avx2.pmulhu.w(<16 x i16> %1, <16 x i16> %2) - ret <16 x i16> %3 -} -declare <16 x i16> @llvm.x86.avx2.pmulhu.w(<16 x i16>, <16 x i16>) nounwind readnone - -define <16 x i16> @test_pmulhw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_pmulhw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmulhw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; GENERIC-NEXT: vpmulhw (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmulhw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmulhw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: vpmulhw (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmulhw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmulhw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; BROADWELL-NEXT: vpmulhw (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmulhw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmulhw %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKYLAKE-NEXT: vpmulhw (%rdi), %ymm0, %ymm0 # sched: [11:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmulhw: -; SKX: # %bb.0: -; SKX-NEXT: vpmulhw %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKX-NEXT: vpmulhw (%rdi), %ymm0, %ymm0 # sched: [11:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmulhw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmulhw %ymm1, %ymm0, %ymm0 # sched: [4:1.00] -; ZNVER1-NEXT: vpmulhw (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.pmulh.w(<16 x i16> %a0, <16 x i16> %a1) - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = call <16 x i16> @llvm.x86.avx2.pmulh.w(<16 x i16> %1, <16 x i16> %2) - ret <16 x i16> %3 -} -declare <16 x i16> @llvm.x86.avx2.pmulh.w(<16 x i16>, <16 x i16>) nounwind readnone - -define <8 x i32> @test_pmulld(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_pmulld: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmulld %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; GENERIC-NEXT: vpmulld (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmulld: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmulld %ymm1, %ymm0, %ymm0 # sched: [10:2.00] -; HASWELL-NEXT: vpmulld (%rdi), %ymm0, %ymm0 # sched: [17:2.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmulld: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmulld %ymm1, %ymm0, %ymm0 # sched: [10:2.00] -; BROADWELL-NEXT: vpmulld (%rdi), %ymm0, %ymm0 # sched: [16:2.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmulld: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmulld %ymm1, %ymm0, %ymm0 # sched: [10:1.00] -; SKYLAKE-NEXT: vpmulld (%rdi), %ymm0, %ymm0 # sched: [17:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmulld: -; SKX: # %bb.0: -; SKX-NEXT: vpmulld %ymm1, %ymm0, %ymm0 # sched: [10:1.00] -; SKX-NEXT: vpmulld (%rdi), %ymm0, %ymm0 # sched: [17:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmulld: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmulld %ymm1, %ymm0, %ymm0 # sched: [5:2.00] -; ZNVER1-NEXT: vpmulld (%rdi), %ymm0, %ymm0 # sched: [12:2.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = mul <8 x i32> %a0, %a1 - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = mul <8 x i32> %1, %2 - ret <8 x i32> %3 -} - -define <16 x i16> @test_pmullw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_pmullw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmullw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; GENERIC-NEXT: vpmullw (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmullw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmullw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: vpmullw (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmullw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmullw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; BROADWELL-NEXT: vpmullw (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmullw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmullw %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKYLAKE-NEXT: vpmullw (%rdi), %ymm0, %ymm0 # sched: [11:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmullw: -; SKX: # %bb.0: -; SKX-NEXT: vpmullw %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKX-NEXT: vpmullw (%rdi), %ymm0, %ymm0 # sched: [11:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmullw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmullw %ymm1, %ymm0, %ymm0 # sched: [4:1.00] -; ZNVER1-NEXT: vpmullw (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = mul <16 x i16> %a0, %a1 - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = mul <16 x i16> %1, %2 - ret <16 x i16> %3 -} - -define <4 x i64> @test_pmuludq(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_pmuludq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpmuludq %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; GENERIC-NEXT: vpmuludq (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pmuludq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpmuludq %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: vpmuludq (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pmuludq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpmuludq %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; BROADWELL-NEXT: vpmuludq (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pmuludq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpmuludq %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKYLAKE-NEXT: vpmuludq (%rdi), %ymm0, %ymm0 # sched: [11:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pmuludq: -; SKX: # %bb.0: -; SKX-NEXT: vpmuludq %ymm1, %ymm0, %ymm0 # sched: [4:0.50] -; SKX-NEXT: vpmuludq (%rdi), %ymm0, %ymm0 # sched: [11:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pmuludq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpmuludq %ymm1, %ymm0, %ymm0 # sched: [4:1.00] -; ZNVER1-NEXT: vpmuludq (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i64> @llvm.x86.avx2.pmulu.dq(<8 x i32> %a0, <8 x i32> %a1) - %2 = bitcast <4 x i64> %1 to <8 x i32> - %3 = load <8 x i32>, <8 x i32> *%a2, align 32 - %4 = call <4 x i64> @llvm.x86.avx2.pmulu.dq(<8 x i32> %2, <8 x i32> %3) - ret <4 x i64> %4 -} -declare <4 x i64> @llvm.x86.avx2.pmulu.dq(<8 x i32>, <8 x i32>) nounwind readnone - -define <4 x i64> @test_por(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { -; GENERIC-LABEL: test_por: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; GENERIC-NEXT: vpor (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_por: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: vpor (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_por: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; BROADWELL-NEXT: vpor (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_por: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: vpor (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_por: -; SKX: # %bb.0: -; SKX-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: vpor (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_por: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpor (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = or <4 x i64> %a0, %a1 - %2 = load <4 x i64>, <4 x i64> *%a2, align 32 - %3 = or <4 x i64> %1, %2 - %4 = add <4 x i64> %3, %a1 - ret <4 x i64> %4 -} - -define <4 x i64> @test_psadbw(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_psadbw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsadbw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; GENERIC-NEXT: vpsadbw (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psadbw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsadbw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; HASWELL-NEXT: vpsadbw (%rdi), %ymm0, %ymm0 # sched: [12:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psadbw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsadbw %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; BROADWELL-NEXT: vpsadbw (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psadbw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsadbw %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SKYLAKE-NEXT: vpsadbw (%rdi), %ymm0, %ymm0 # sched: [10:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psadbw: -; SKX: # %bb.0: -; SKX-NEXT: vpsadbw %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; SKX-NEXT: vpsadbw (%rdi), %ymm0, %ymm0 # sched: [10:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psadbw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsadbw %ymm1, %ymm0, %ymm0 # sched: [3:1.00] -; ZNVER1-NEXT: vpsadbw (%rdi), %ymm0, %ymm0 # sched: [10:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i64> @llvm.x86.avx2.psad.bw(<32 x i8> %a0, <32 x i8> %a1) - %2 = bitcast <4 x i64> %1 to <32 x i8> - %3 = load <32 x i8>, <32 x i8> *%a2, align 32 - %4 = call <4 x i64> @llvm.x86.avx2.psad.bw(<32 x i8> %2, <32 x i8> %3) - ret <4 x i64> %4 -} -declare <4 x i64> @llvm.x86.avx2.psad.bw(<32 x i8>, <32 x i8>) nounwind readnone - -define <32 x i8> @test_pshufb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_pshufb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpshufb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpshufb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pshufb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpshufb %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: vpshufb (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pshufb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpshufb %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; BROADWELL-NEXT: vpshufb (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pshufb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpshufb %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; SKYLAKE-NEXT: vpshufb (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pshufb: -; SKX: # %bb.0: -; SKX-NEXT: vpshufb %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; SKX-NEXT: vpshufb (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pshufb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpshufb %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpshufb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %a0, <32 x i8> %a1) - %2 = load <32 x i8>, <32 x i8> *%a2, align 32 - %3 = call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %1, <32 x i8> %2) - ret <32 x i8> %3 -} -declare <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8>, <32 x i8>) nounwind readnone - -define <8 x i32> @test_pshufd(<8 x i32> %a0, <8 x i32> *%a1) { -; GENERIC-LABEL: test_pshufd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00] -; GENERIC-NEXT: vpshufd {{.*#+}} ymm1 = mem[1,0,3,2,5,4,7,6] sched: [8:1.00] -; GENERIC-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pshufd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00] -; HASWELL-NEXT: vpshufd {{.*#+}} ymm1 = mem[1,0,3,2,5,4,7,6] sched: [8:1.00] -; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pshufd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00] -; BROADWELL-NEXT: vpshufd {{.*#+}} ymm1 = mem[1,0,3,2,5,4,7,6] sched: [7:1.00] -; BROADWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pshufd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00] -; SKYLAKE-NEXT: vpshufd {{.*#+}} ymm1 = mem[1,0,3,2,5,4,7,6] sched: [8:1.00] -; SKYLAKE-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pshufd: -; SKX: # %bb.0: -; SKX-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00] -; SKX-NEXT: vpshufd {{.*#+}} ymm1 = mem[1,0,3,2,5,4,7,6] sched: [8:1.00] -; SKX-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pshufd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpshufd {{.*#+}} ymm1 = mem[1,0,3,2,5,4,7,6] sched: [8:0.50] -; ZNVER1-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:0.25] -; ZNVER1-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <8 x i32> %a0, <8 x i32> undef, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4> - %2 = load <8 x i32>, <8 x i32> *%a1, align 32 - %3 = shufflevector <8 x i32> %2, <8 x i32> undef, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6> - %4 = add <8 x i32> %1, %3 - ret <8 x i32> %4 -} - -define <16 x i16> @test_pshufhw(<16 x i16> %a0, <16 x i16> *%a1) { -; GENERIC-LABEL: test_pshufhw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12] sched: [1:1.00] -; GENERIC-NEXT: vpshufhw {{.*#+}} ymm1 = mem[0,1,2,3,5,4,7,6,8,9,10,11,13,12,15,14] sched: [8:1.00] -; GENERIC-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pshufhw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12] sched: [1:1.00] -; HASWELL-NEXT: vpshufhw {{.*#+}} ymm1 = mem[0,1,2,3,5,4,7,6,8,9,10,11,13,12,15,14] sched: [8:1.00] -; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pshufhw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12] sched: [1:1.00] -; BROADWELL-NEXT: vpshufhw {{.*#+}} ymm1 = mem[0,1,2,3,5,4,7,6,8,9,10,11,13,12,15,14] sched: [7:1.00] -; BROADWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pshufhw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12] sched: [1:1.00] -; SKYLAKE-NEXT: vpshufhw {{.*#+}} ymm1 = mem[0,1,2,3,5,4,7,6,8,9,10,11,13,12,15,14] sched: [8:1.00] -; SKYLAKE-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pshufhw: -; SKX: # %bb.0: -; SKX-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12] sched: [1:1.00] -; SKX-NEXT: vpshufhw {{.*#+}} ymm1 = mem[0,1,2,3,5,4,7,6,8,9,10,11,13,12,15,14] sched: [8:1.00] -; SKX-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pshufhw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpshufhw {{.*#+}} ymm1 = mem[0,1,2,3,5,4,7,6,8,9,10,11,13,12,15,14] sched: [8:0.50] -; ZNVER1-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12] sched: [1:0.25] -; ZNVER1-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <16 x i16> %a0, <16 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 7, i32 6, i32 5, i32 4, i32 8, i32 9, i32 10, i32 11, i32 15, i32 14, i32 13, i32 12> - %2 = load <16 x i16>, <16 x i16> *%a1, align 32 - %3 = shufflevector <16 x i16> %2, <16 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 7, i32 6, i32 8, i32 9, i32 10, i32 11, i32 13, i32 12, i32 15, i32 14> - %4 = or <16 x i16> %1, %3 - ret <16 x i16> %4 -} - -define <16 x i16> @test_pshuflw(<16 x i16> %a0, <16 x i16> *%a1) { -; GENERIC-LABEL: test_pshuflw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpshuflw {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15] sched: [1:1.00] -; GENERIC-NEXT: vpshuflw {{.*#+}} ymm1 = mem[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15] sched: [8:1.00] -; GENERIC-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pshuflw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpshuflw {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15] sched: [1:1.00] -; HASWELL-NEXT: vpshuflw {{.*#+}} ymm1 = mem[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15] sched: [8:1.00] -; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pshuflw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpshuflw {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15] sched: [1:1.00] -; BROADWELL-NEXT: vpshuflw {{.*#+}} ymm1 = mem[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15] sched: [7:1.00] -; BROADWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pshuflw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpshuflw {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15] sched: [1:1.00] -; SKYLAKE-NEXT: vpshuflw {{.*#+}} ymm1 = mem[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15] sched: [8:1.00] -; SKYLAKE-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pshuflw: -; SKX: # %bb.0: -; SKX-NEXT: vpshuflw {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15] sched: [1:1.00] -; SKX-NEXT: vpshuflw {{.*#+}} ymm1 = mem[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15] sched: [8:1.00] -; SKX-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pshuflw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpshuflw {{.*#+}} ymm1 = mem[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15] sched: [8:0.50] -; ZNVER1-NEXT: vpshuflw {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15] sched: [1:0.25] -; ZNVER1-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <16 x i16> %a0, <16 x i16> undef, <16 x i32> <i32 3, i32 2, i32 1, i32 0, i32 4, i32 5, i32 6, i32 7, i32 11, i32 10, i32 9, i32 8, i32 12, i32 13, i32 14, i32 15> - %2 = load <16 x i16>, <16 x i16> *%a1, align 32 - %3 = shufflevector <16 x i16> %2, <16 x i16> undef, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 4, i32 5, i32 6, i32 7, i32 9, i32 8, i32 11, i32 10, i32 12, i32 13, i32 14, i32 15> - %4 = or <16 x i16> %1, %3 - ret <16 x i16> %4 -} - -define <32 x i8> @test_psignb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_psignb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsignb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpsignb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psignb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsignb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsignb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psignb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsignb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpsignb (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psignb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsignb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsignb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psignb: -; SKX: # %bb.0: -; SKX-NEXT: vpsignb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpsignb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psignb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsignb %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpsignb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <32 x i8> @llvm.x86.avx2.psign.b(<32 x i8> %a0, <32 x i8> %a1) - %2 = load <32 x i8>, <32 x i8> *%a2, align 32 - %3 = call <32 x i8> @llvm.x86.avx2.psign.b(<32 x i8> %1, <32 x i8> %2) - ret <32 x i8> %3 -} -declare <32 x i8> @llvm.x86.avx2.psign.b(<32 x i8>, <32 x i8>) nounwind readnone - -define <8 x i32> @test_psignd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_psignd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsignd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpsignd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psignd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsignd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsignd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psignd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsignd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpsignd (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psignd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsignd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsignd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psignd: -; SKX: # %bb.0: -; SKX-NEXT: vpsignd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpsignd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psignd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsignd %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpsignd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.psign.d(<8 x i32> %a0, <8 x i32> %a1) - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = call <8 x i32> @llvm.x86.avx2.psign.d(<8 x i32> %1, <8 x i32> %2) - ret <8 x i32> %3 -} -declare <8 x i32> @llvm.x86.avx2.psign.d(<8 x i32>, <8 x i32>) nounwind readnone - -define <16 x i16> @test_psignw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_psignw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsignw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpsignw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psignw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsignw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsignw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psignw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsignw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpsignw (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psignw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsignw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsignw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psignw: -; SKX: # %bb.0: -; SKX-NEXT: vpsignw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpsignw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psignw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsignw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpsignw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.psign.w(<16 x i16> %a0, <16 x i16> %a1) - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = call <16 x i16> @llvm.x86.avx2.psign.w(<16 x i16> %1, <16 x i16> %2) - ret <16 x i16> %3 -} -declare <16 x i16> @llvm.x86.avx2.psign.w(<16 x i16>, <16 x i16>) nounwind readnone - -define <8 x i32> @test_pslld(<8 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { -; GENERIC-LABEL: test_pslld: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpslld %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; GENERIC-NEXT: vpslld (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; GENERIC-NEXT: vpslld $2, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pslld: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpslld %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; HASWELL-NEXT: vpslld (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; HASWELL-NEXT: vpslld $2, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pslld: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpslld %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; BROADWELL-NEXT: vpslld (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; BROADWELL-NEXT: vpslld $2, %ymm0, %ymm0 # sched: [1:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pslld: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpslld %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; SKYLAKE-NEXT: vpslld (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: vpslld $2, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pslld: -; SKX: # %bb.0: -; SKX-NEXT: vpslld %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; SKX-NEXT: vpslld (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: vpslld $2, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pslld: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpslld %xmm1, %ymm0, %ymm0 # sched: [2:1.00] -; ZNVER1-NEXT: vpslld (%rdi), %ymm0, %ymm0 # sched: [9:1.00] -; ZNVER1-NEXT: vpslld $2, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.psll.d(<8 x i32> %a0, <4 x i32> %a1) - %2 = load <4 x i32>, <4 x i32> *%a2, align 16 - %3 = call <8 x i32> @llvm.x86.avx2.psll.d(<8 x i32> %1, <4 x i32> %2) - %4 = shl <8 x i32> %3, <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2> - ret <8 x i32> %4 -} -declare <8 x i32> @llvm.x86.avx2.psll.d(<8 x i32>, <4 x i32>) nounwind readnone - -define <32 x i8> @test_pslldq(<32 x i8> %a0) { -; GENERIC-LABEL: test_pslldq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpslldq {{.*#+}} ymm0 = zero,zero,zero,ymm0[0,1,2,3,4,5,6,7,8,9,10,11,12],zero,zero,zero,ymm0[16,17,18,19,20,21,22,23,24,25,26,27,28] sched: [1:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pslldq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpslldq {{.*#+}} ymm0 = zero,zero,zero,ymm0[0,1,2,3,4,5,6,7,8,9,10,11,12],zero,zero,zero,ymm0[16,17,18,19,20,21,22,23,24,25,26,27,28] sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pslldq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpslldq {{.*#+}} ymm0 = zero,zero,zero,ymm0[0,1,2,3,4,5,6,7,8,9,10,11,12],zero,zero,zero,ymm0[16,17,18,19,20,21,22,23,24,25,26,27,28] sched: [1:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pslldq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpslldq {{.*#+}} ymm0 = zero,zero,zero,ymm0[0,1,2,3,4,5,6,7,8,9,10,11,12],zero,zero,zero,ymm0[16,17,18,19,20,21,22,23,24,25,26,27,28] sched: [1:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pslldq: -; SKX: # %bb.0: -; SKX-NEXT: vpslldq {{.*#+}} ymm0 = zero,zero,zero,ymm0[0,1,2,3,4,5,6,7,8,9,10,11,12],zero,zero,zero,ymm0[16,17,18,19,20,21,22,23,24,25,26,27,28] sched: [1:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pslldq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpslldq {{.*#+}} ymm0 = zero,zero,zero,ymm0[0,1,2,3,4,5,6,7,8,9,10,11,12],zero,zero,zero,ymm0[16,17,18,19,20,21,22,23,24,25,26,27,28] sched: [2:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <32 x i8> zeroinitializer, <32 x i8> %a0, <32 x i32> <i32 13, i32 14, i32 15, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 29, i32 30, i32 31, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60> - ret <32 x i8> %1 -} - -define <4 x i64> @test_psllq(<4 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) { -; GENERIC-LABEL: test_psllq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsllq %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; GENERIC-NEXT: vpsllq (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; GENERIC-NEXT: vpsllq $2, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psllq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsllq %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; HASWELL-NEXT: vpsllq (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; HASWELL-NEXT: vpsllq $2, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psllq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsllq %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; BROADWELL-NEXT: vpsllq (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; BROADWELL-NEXT: vpsllq $2, %ymm0, %ymm0 # sched: [1:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psllq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsllq %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; SKYLAKE-NEXT: vpsllq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: vpsllq $2, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psllq: -; SKX: # %bb.0: -; SKX-NEXT: vpsllq %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; SKX-NEXT: vpsllq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: vpsllq $2, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psllq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsllq %xmm1, %ymm0, %ymm0 # sched: [2:1.00] -; ZNVER1-NEXT: vpsllq (%rdi), %ymm0, %ymm0 # sched: [9:1.00] -; ZNVER1-NEXT: vpsllq $2, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i64> @llvm.x86.avx2.psll.q(<4 x i64> %a0, <2 x i64> %a1) - %2 = load <2 x i64>, <2 x i64> *%a2, align 16 - %3 = call <4 x i64> @llvm.x86.avx2.psll.q(<4 x i64> %1, <2 x i64> %2) - %4 = shl <4 x i64> %3, <i64 2, i64 2, i64 2, i64 2> - ret <4 x i64> %4 -} -declare <4 x i64> @llvm.x86.avx2.psll.q(<4 x i64>, <2 x i64>) nounwind readnone - -define <4 x i32> @test_psllvd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { -; GENERIC-LABEL: test_psllvd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsllvd %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; GENERIC-NEXT: vpsllvd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psllvd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsllvd %xmm1, %xmm0, %xmm0 # sched: [3:2.00] -; HASWELL-NEXT: vpsllvd (%rdi), %xmm0, %xmm0 # sched: [9:2.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psllvd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsllvd %xmm1, %xmm0, %xmm0 # sched: [3:2.00] -; BROADWELL-NEXT: vpsllvd (%rdi), %xmm0, %xmm0 # sched: [8:2.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psllvd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsllvd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsllvd (%rdi), %xmm0, %xmm0 # sched: [7:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psllvd: -; SKX: # %bb.0: -; SKX-NEXT: vpsllvd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SKX-NEXT: vpsllvd (%rdi), %xmm0, %xmm0 # sched: [7:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psllvd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsllvd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpsllvd (%rdi), %xmm0, %xmm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i32> @llvm.x86.avx2.psllv.d(<4 x i32> %a0, <4 x i32> %a1) - %2 = load <4 x i32>, <4 x i32> *%a2, align 16 - %3 = call <4 x i32> @llvm.x86.avx2.psllv.d(<4 x i32> %1, <4 x i32> %2) - ret <4 x i32> %3 -} -declare <4 x i32> @llvm.x86.avx2.psllv.d(<4 x i32>, <4 x i32>) nounwind readnone - -define <8 x i32> @test_psllvd_ymm(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_psllvd_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsllvd %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpsllvd (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psllvd_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsllvd %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; HASWELL-NEXT: vpsllvd (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psllvd_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsllvd %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; BROADWELL-NEXT: vpsllvd (%rdi), %ymm0, %ymm0 # sched: [9:2.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psllvd_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsllvd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsllvd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psllvd_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vpsllvd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpsllvd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psllvd_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsllvd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpsllvd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.psllv.d.256(<8 x i32> %a0, <8 x i32> %a1) - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = call <8 x i32> @llvm.x86.avx2.psllv.d.256(<8 x i32> %1, <8 x i32> %2) - ret <8 x i32> %3 -} -declare <8 x i32> @llvm.x86.avx2.psllv.d.256(<8 x i32>, <8 x i32>) nounwind readnone - -define <2 x i64> @test_psllvq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) { -; GENERIC-LABEL: test_psllvq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsllvq %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; GENERIC-NEXT: vpsllvq (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psllvq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsllvq %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpsllvq (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psllvq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsllvq %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; BROADWELL-NEXT: vpsllvq (%rdi), %xmm0, %xmm0 # sched: [6:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psllvq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsllvq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsllvq (%rdi), %xmm0, %xmm0 # sched: [7:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psllvq: -; SKX: # %bb.0: -; SKX-NEXT: vpsllvq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SKX-NEXT: vpsllvq (%rdi), %xmm0, %xmm0 # sched: [7:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psllvq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsllvq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpsllvq (%rdi), %xmm0, %xmm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <2 x i64> @llvm.x86.avx2.psllv.q(<2 x i64> %a0, <2 x i64> %a1) - %2 = load <2 x i64>, <2 x i64> *%a2, align 16 - %3 = call <2 x i64> @llvm.x86.avx2.psllv.q(<2 x i64> %1, <2 x i64> %2) - ret <2 x i64> %3 -} -declare <2 x i64> @llvm.x86.avx2.psllv.q(<2 x i64>, <2 x i64>) nounwind readnone - -define <4 x i64> @test_psllvq_ymm(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { -; GENERIC-LABEL: test_psllvq_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsllvq %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpsllvq (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psllvq_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsllvq %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: vpsllvq (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psllvq_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsllvq %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; BROADWELL-NEXT: vpsllvq (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psllvq_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsllvq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsllvq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psllvq_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vpsllvq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpsllvq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psllvq_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsllvq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpsllvq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i64> @llvm.x86.avx2.psllv.q.256(<4 x i64> %a0, <4 x i64> %a1) - %2 = load <4 x i64>, <4 x i64> *%a2, align 32 - %3 = call <4 x i64> @llvm.x86.avx2.psllv.q.256(<4 x i64> %1, <4 x i64> %2) - ret <4 x i64> %3 -} -declare <4 x i64> @llvm.x86.avx2.psllv.q.256(<4 x i64>, <4 x i64>) nounwind readnone - -define <16 x i16> @test_psllw(<16 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { -; GENERIC-LABEL: test_psllw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsllw %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; GENERIC-NEXT: vpsllw (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; GENERIC-NEXT: vpsllw $2, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psllw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsllw %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; HASWELL-NEXT: vpsllw (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; HASWELL-NEXT: vpsllw $2, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psllw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsllw %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; BROADWELL-NEXT: vpsllw (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; BROADWELL-NEXT: vpsllw $2, %ymm0, %ymm0 # sched: [1:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psllw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsllw %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; SKYLAKE-NEXT: vpsllw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: vpsllw $2, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psllw: -; SKX: # %bb.0: -; SKX-NEXT: vpsllw %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; SKX-NEXT: vpsllw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: vpsllw $2, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psllw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsllw %xmm1, %ymm0, %ymm0 # sched: [2:1.00] -; ZNVER1-NEXT: vpsllw (%rdi), %ymm0, %ymm0 # sched: [9:1.00] -; ZNVER1-NEXT: vpsllw $2, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.psll.w(<16 x i16> %a0, <8 x i16> %a1) - %2 = load <8 x i16>, <8 x i16> *%a2, align 16 - %3 = call <16 x i16> @llvm.x86.avx2.psll.w(<16 x i16> %1, <8 x i16> %2) - %4 = shl <16 x i16> %3, <i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2> - ret <16 x i16> %4 -} -declare <16 x i16> @llvm.x86.avx2.psll.w(<16 x i16>, <8 x i16>) nounwind readnone - -define <8 x i32> @test_psrad(<8 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { -; GENERIC-LABEL: test_psrad: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsrad %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; GENERIC-NEXT: vpsrad (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; GENERIC-NEXT: vpsrad $2, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psrad: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsrad %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; HASWELL-NEXT: vpsrad (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; HASWELL-NEXT: vpsrad $2, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psrad: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsrad %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; BROADWELL-NEXT: vpsrad (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; BROADWELL-NEXT: vpsrad $2, %ymm0, %ymm0 # sched: [1:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psrad: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsrad %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; SKYLAKE-NEXT: vpsrad (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: vpsrad $2, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psrad: -; SKX: # %bb.0: -; SKX-NEXT: vpsrad %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; SKX-NEXT: vpsrad (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: vpsrad $2, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psrad: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsrad %xmm1, %ymm0, %ymm0 # sched: [2:1.00] -; ZNVER1-NEXT: vpsrad (%rdi), %ymm0, %ymm0 # sched: [9:1.00] -; ZNVER1-NEXT: vpsrad $2, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.psra.d(<8 x i32> %a0, <4 x i32> %a1) - %2 = load <4 x i32>, <4 x i32> *%a2, align 16 - %3 = call <8 x i32> @llvm.x86.avx2.psra.d(<8 x i32> %1, <4 x i32> %2) - %4 = ashr <8 x i32> %3, <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2> - ret <8 x i32> %4 -} -declare <8 x i32> @llvm.x86.avx2.psra.d(<8 x i32>, <4 x i32>) nounwind readnone - -define <4 x i32> @test_psravd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { -; GENERIC-LABEL: test_psravd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsravd %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; GENERIC-NEXT: vpsravd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psravd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsravd %xmm1, %xmm0, %xmm0 # sched: [3:2.00] -; HASWELL-NEXT: vpsravd (%rdi), %xmm0, %xmm0 # sched: [9:2.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psravd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsravd %xmm1, %xmm0, %xmm0 # sched: [3:2.00] -; BROADWELL-NEXT: vpsravd (%rdi), %xmm0, %xmm0 # sched: [8:2.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psravd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsravd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsravd (%rdi), %xmm0, %xmm0 # sched: [7:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psravd: -; SKX: # %bb.0: -; SKX-NEXT: vpsravd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SKX-NEXT: vpsravd (%rdi), %xmm0, %xmm0 # sched: [7:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psravd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsravd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpsravd (%rdi), %xmm0, %xmm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i32> @llvm.x86.avx2.psrav.d(<4 x i32> %a0, <4 x i32> %a1) - %2 = load <4 x i32>, <4 x i32> *%a2, align 16 - %3 = call <4 x i32> @llvm.x86.avx2.psrav.d(<4 x i32> %1, <4 x i32> %2) - ret <4 x i32> %3 -} -declare <4 x i32> @llvm.x86.avx2.psrav.d(<4 x i32>, <4 x i32>) nounwind readnone - -define <8 x i32> @test_psravd_ymm(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_psravd_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsravd %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpsravd (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psravd_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsravd %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; HASWELL-NEXT: vpsravd (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psravd_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsravd %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; BROADWELL-NEXT: vpsravd (%rdi), %ymm0, %ymm0 # sched: [9:2.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psravd_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsravd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsravd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psravd_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vpsravd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpsravd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psravd_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsravd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpsravd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.psrav.d.256(<8 x i32> %a0, <8 x i32> %a1) - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = call <8 x i32> @llvm.x86.avx2.psrav.d.256(<8 x i32> %1, <8 x i32> %2) - ret <8 x i32> %3 -} -declare <8 x i32> @llvm.x86.avx2.psrav.d.256(<8 x i32>, <8 x i32>) nounwind readnone - -define <16 x i16> @test_psraw(<16 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { -; GENERIC-LABEL: test_psraw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsraw %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; GENERIC-NEXT: vpsraw (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; GENERIC-NEXT: vpsraw $2, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psraw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsraw %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; HASWELL-NEXT: vpsraw (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; HASWELL-NEXT: vpsraw $2, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psraw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsraw %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; BROADWELL-NEXT: vpsraw (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; BROADWELL-NEXT: vpsraw $2, %ymm0, %ymm0 # sched: [1:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psraw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsraw %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; SKYLAKE-NEXT: vpsraw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: vpsraw $2, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psraw: -; SKX: # %bb.0: -; SKX-NEXT: vpsraw %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; SKX-NEXT: vpsraw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: vpsraw $2, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psraw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsraw %xmm1, %ymm0, %ymm0 # sched: [2:1.00] -; ZNVER1-NEXT: vpsraw (%rdi), %ymm0, %ymm0 # sched: [9:1.00] -; ZNVER1-NEXT: vpsraw $2, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.psra.w(<16 x i16> %a0, <8 x i16> %a1) - %2 = load <8 x i16>, <8 x i16> *%a2, align 16 - %3 = call <16 x i16> @llvm.x86.avx2.psra.w(<16 x i16> %1, <8 x i16> %2) - %4 = ashr <16 x i16> %3, <i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2> - ret <16 x i16> %4 -} -declare <16 x i16> @llvm.x86.avx2.psra.w(<16 x i16>, <8 x i16>) nounwind readnone - -define <8 x i32> @test_psrld(<8 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { -; GENERIC-LABEL: test_psrld: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsrld %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; GENERIC-NEXT: vpsrld (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; GENERIC-NEXT: vpsrld $2, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psrld: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsrld %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; HASWELL-NEXT: vpsrld (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; HASWELL-NEXT: vpsrld $2, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psrld: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsrld %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; BROADWELL-NEXT: vpsrld (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; BROADWELL-NEXT: vpsrld $2, %ymm0, %ymm0 # sched: [1:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psrld: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsrld %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; SKYLAKE-NEXT: vpsrld (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: vpsrld $2, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psrld: -; SKX: # %bb.0: -; SKX-NEXT: vpsrld %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; SKX-NEXT: vpsrld (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: vpsrld $2, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psrld: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsrld %xmm1, %ymm0, %ymm0 # sched: [2:1.00] -; ZNVER1-NEXT: vpsrld (%rdi), %ymm0, %ymm0 # sched: [9:1.00] -; ZNVER1-NEXT: vpsrld $2, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.psrl.d(<8 x i32> %a0, <4 x i32> %a1) - %2 = load <4 x i32>, <4 x i32> *%a2, align 16 - %3 = call <8 x i32> @llvm.x86.avx2.psrl.d(<8 x i32> %1, <4 x i32> %2) - %4 = lshr <8 x i32> %3, <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2> - ret <8 x i32> %4 -} -declare <8 x i32> @llvm.x86.avx2.psrl.d(<8 x i32>, <4 x i32>) nounwind readnone - -define <32 x i8> @test_psrldq(<32 x i8> %a0) { -; GENERIC-LABEL: test_psrldq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,ymm0[19,20,21,22,23,24,25,26,27,28,29,30,31],zero,zero,zero sched: [1:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psrldq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,ymm0[19,20,21,22,23,24,25,26,27,28,29,30,31],zero,zero,zero sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psrldq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,ymm0[19,20,21,22,23,24,25,26,27,28,29,30,31],zero,zero,zero sched: [1:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psrldq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,ymm0[19,20,21,22,23,24,25,26,27,28,29,30,31],zero,zero,zero sched: [1:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psrldq: -; SKX: # %bb.0: -; SKX-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,ymm0[19,20,21,22,23,24,25,26,27,28,29,30,31],zero,zero,zero sched: [1:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psrldq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,ymm0[19,20,21,22,23,24,25,26,27,28,29,30,31],zero,zero,zero sched: [2:1.00] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <32 x i8> %a0, <32 x i8> zeroinitializer, <32 x i32> <i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 32, i32 33, i32 34, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 48, i32 49, i32 50> - ret <32 x i8> %1 -} - -define <4 x i64> @test_psrlq(<4 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) { -; GENERIC-LABEL: test_psrlq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsrlq %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; GENERIC-NEXT: vpsrlq (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; GENERIC-NEXT: vpsrlq $2, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psrlq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsrlq %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; HASWELL-NEXT: vpsrlq (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; HASWELL-NEXT: vpsrlq $2, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psrlq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsrlq %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; BROADWELL-NEXT: vpsrlq (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; BROADWELL-NEXT: vpsrlq $2, %ymm0, %ymm0 # sched: [1:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psrlq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsrlq %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; SKYLAKE-NEXT: vpsrlq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: vpsrlq $2, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psrlq: -; SKX: # %bb.0: -; SKX-NEXT: vpsrlq %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; SKX-NEXT: vpsrlq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: vpsrlq $2, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psrlq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsrlq %xmm1, %ymm0, %ymm0 # sched: [2:1.00] -; ZNVER1-NEXT: vpsrlq (%rdi), %ymm0, %ymm0 # sched: [9:1.00] -; ZNVER1-NEXT: vpsrlq $2, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i64> @llvm.x86.avx2.psrl.q(<4 x i64> %a0, <2 x i64> %a1) - %2 = load <2 x i64>, <2 x i64> *%a2, align 16 - %3 = call <4 x i64> @llvm.x86.avx2.psrl.q(<4 x i64> %1, <2 x i64> %2) - %4 = lshr <4 x i64> %3, <i64 2, i64 2, i64 2, i64 2> - ret <4 x i64> %4 -} -declare <4 x i64> @llvm.x86.avx2.psrl.q(<4 x i64>, <2 x i64>) nounwind readnone - -define <4 x i32> @test_psrlvd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) { -; GENERIC-LABEL: test_psrlvd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsrlvd %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; GENERIC-NEXT: vpsrlvd (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psrlvd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsrlvd %xmm1, %xmm0, %xmm0 # sched: [3:2.00] -; HASWELL-NEXT: vpsrlvd (%rdi), %xmm0, %xmm0 # sched: [9:2.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psrlvd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsrlvd %xmm1, %xmm0, %xmm0 # sched: [3:2.00] -; BROADWELL-NEXT: vpsrlvd (%rdi), %xmm0, %xmm0 # sched: [8:2.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psrlvd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsrlvd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsrlvd (%rdi), %xmm0, %xmm0 # sched: [7:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psrlvd: -; SKX: # %bb.0: -; SKX-NEXT: vpsrlvd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SKX-NEXT: vpsrlvd (%rdi), %xmm0, %xmm0 # sched: [7:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psrlvd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsrlvd %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpsrlvd (%rdi), %xmm0, %xmm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i32> @llvm.x86.avx2.psrlv.d(<4 x i32> %a0, <4 x i32> %a1) - %2 = load <4 x i32>, <4 x i32> *%a2, align 16 - %3 = call <4 x i32> @llvm.x86.avx2.psrlv.d(<4 x i32> %1, <4 x i32> %2) - ret <4 x i32> %3 -} -declare <4 x i32> @llvm.x86.avx2.psrlv.d(<4 x i32>, <4 x i32>) nounwind readnone - -define <8 x i32> @test_psrlvd_ymm(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_psrlvd_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpsrlvd (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psrlvd_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; HASWELL-NEXT: vpsrlvd (%rdi), %ymm0, %ymm0 # sched: [10:2.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psrlvd_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0 # sched: [3:2.00] -; BROADWELL-NEXT: vpsrlvd (%rdi), %ymm0, %ymm0 # sched: [9:2.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psrlvd_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsrlvd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psrlvd_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpsrlvd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psrlvd_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpsrlvd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <8 x i32> @llvm.x86.avx2.psrlv.d.256(<8 x i32> %a0, <8 x i32> %a1) - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = call <8 x i32> @llvm.x86.avx2.psrlv.d.256(<8 x i32> %1, <8 x i32> %2) - ret <8 x i32> %3 -} -declare <8 x i32> @llvm.x86.avx2.psrlv.d.256(<8 x i32>, <8 x i32>) nounwind readnone - -define <2 x i64> @test_psrlvq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) { -; GENERIC-LABEL: test_psrlvq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsrlvq %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; GENERIC-NEXT: vpsrlvq (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psrlvq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsrlvq %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; HASWELL-NEXT: vpsrlvq (%rdi), %xmm0, %xmm0 # sched: [7:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psrlvq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsrlvq %xmm1, %xmm0, %xmm0 # sched: [1:1.00] -; BROADWELL-NEXT: vpsrlvq (%rdi), %xmm0, %xmm0 # sched: [6:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psrlvq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsrlvq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsrlvq (%rdi), %xmm0, %xmm0 # sched: [7:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psrlvq: -; SKX: # %bb.0: -; SKX-NEXT: vpsrlvq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; SKX-NEXT: vpsrlvq (%rdi), %xmm0, %xmm0 # sched: [7:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psrlvq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsrlvq %xmm1, %xmm0, %xmm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpsrlvq (%rdi), %xmm0, %xmm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <2 x i64> @llvm.x86.avx2.psrlv.q(<2 x i64> %a0, <2 x i64> %a1) - %2 = load <2 x i64>, <2 x i64> *%a2, align 16 - %3 = call <2 x i64> @llvm.x86.avx2.psrlv.q(<2 x i64> %1, <2 x i64> %2) - ret <2 x i64> %3 -} -declare <2 x i64> @llvm.x86.avx2.psrlv.q(<2 x i64>, <2 x i64>) nounwind readnone - -define <4 x i64> @test_psrlvq_ymm(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { -; GENERIC-LABEL: test_psrlvq_ymm: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsrlvq %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: vpsrlvq (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psrlvq_ymm: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsrlvq %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: vpsrlvq (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psrlvq_ymm: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsrlvq %ymm1, %ymm0, %ymm0 # sched: [1:1.00] -; BROADWELL-NEXT: vpsrlvq (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psrlvq_ymm: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsrlvq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsrlvq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psrlvq_ymm: -; SKX: # %bb.0: -; SKX-NEXT: vpsrlvq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpsrlvq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psrlvq_ymm: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsrlvq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; ZNVER1-NEXT: vpsrlvq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <4 x i64> @llvm.x86.avx2.psrlv.q.256(<4 x i64> %a0, <4 x i64> %a1) - %2 = load <4 x i64>, <4 x i64> *%a2, align 32 - %3 = call <4 x i64> @llvm.x86.avx2.psrlv.q.256(<4 x i64> %1, <4 x i64> %2) - ret <4 x i64> %3 -} -declare <4 x i64> @llvm.x86.avx2.psrlv.q.256(<4 x i64>, <4 x i64>) nounwind readnone - -define <16 x i16> @test_psrlw(<16 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) { -; GENERIC-LABEL: test_psrlw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsrlw %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; GENERIC-NEXT: vpsrlw (%rdi), %ymm0, %ymm0 # sched: [11:1.00] -; GENERIC-NEXT: vpsrlw $2, %ymm0, %ymm0 # sched: [1:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psrlw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsrlw %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; HASWELL-NEXT: vpsrlw (%rdi), %ymm0, %ymm0 # sched: [8:1.00] -; HASWELL-NEXT: vpsrlw $2, %ymm0, %ymm0 # sched: [1:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psrlw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsrlw %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; BROADWELL-NEXT: vpsrlw (%rdi), %ymm0, %ymm0 # sched: [7:1.00] -; BROADWELL-NEXT: vpsrlw $2, %ymm0, %ymm0 # sched: [1:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psrlw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsrlw %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; SKYLAKE-NEXT: vpsrlw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: vpsrlw $2, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psrlw: -; SKX: # %bb.0: -; SKX-NEXT: vpsrlw %xmm1, %ymm0, %ymm0 # sched: [4:1.00] -; SKX-NEXT: vpsrlw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: vpsrlw $2, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psrlw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsrlw %xmm1, %ymm0, %ymm0 # sched: [2:1.00] -; ZNVER1-NEXT: vpsrlw (%rdi), %ymm0, %ymm0 # sched: [9:1.00] -; ZNVER1-NEXT: vpsrlw $2, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.x86.avx2.psrl.w(<16 x i16> %a0, <8 x i16> %a1) - %2 = load <8 x i16>, <8 x i16> *%a2, align 16 - %3 = call <16 x i16> @llvm.x86.avx2.psrl.w(<16 x i16> %1, <8 x i16> %2) - %4 = lshr <16 x i16> %3, <i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2> - ret <16 x i16> %4 -} -declare <16 x i16> @llvm.x86.avx2.psrl.w(<16 x i16>, <8 x i16>) nounwind readnone - -define <32 x i8> @test_psubb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_psubb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsubb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpsubb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psubb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsubb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psubb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsubb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpsubb (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psubb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsubb %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: vpsubb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psubb: -; SKX: # %bb.0: -; SKX-NEXT: vpsubb %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: vpsubb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psubb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsubb %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpsubb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = sub <32 x i8> %a0, %a1 - %2 = load <32 x i8>, <32 x i8> *%a2, align 32 - %3 = sub <32 x i8> %1, %2 - ret <32 x i8> %3 -} - -define <8 x i32> @test_psubd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_psubd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpsubd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psubd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psubd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpsubd (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psubd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: vpsubd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psubd: -; SKX: # %bb.0: -; SKX-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: vpsubd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psubd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpsubd (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = sub <8 x i32> %a0, %a1 - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = sub <8 x i32> %1, %2 - ret <8 x i32> %3 -} - -define <4 x i64> @test_psubq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { -; GENERIC-LABEL: test_psubq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsubq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpsubq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psubq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsubq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psubq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsubq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpsubq (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psubq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsubq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: vpsubq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psubq: -; SKX: # %bb.0: -; SKX-NEXT: vpsubq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: vpsubq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psubq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsubq %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpsubq (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = sub <4 x i64> %a0, %a1 - %2 = load <4 x i64>, <4 x i64> *%a2, align 32 - %3 = sub <4 x i64> %1, %2 - ret <4 x i64> %3 -} - -define <32 x i8> @test_psubsb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_psubsb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psubsb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psubsb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psubsb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psubsb: -; SKX: # %bb.0: -; SKX-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psubsb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <32 x i8> @llvm.ssub.sat.v32i8(<32 x i8> %a0, <32 x i8> %a1) - %2 = load <32 x i8>, <32 x i8> *%a2, align 32 - %3 = call <32 x i8> @llvm.ssub.sat.v32i8(<32 x i8> %1, <32 x i8> %2) - ret <32 x i8> %3 -} -declare <32 x i8> @llvm.ssub.sat.v32i8(<32 x i8>, <32 x i8>) nounwind readnone - -define <16 x i16> @test_psubsw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_psubsw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psubsw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psubsw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psubsw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psubsw: -; SKX: # %bb.0: -; SKX-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psubsw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.ssub.sat.v16i16(<16 x i16> %a0, <16 x i16> %a1) - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = call <16 x i16> @llvm.ssub.sat.v16i16(<16 x i16> %1, <16 x i16> %2) - ret <16 x i16> %3 -} -declare <16 x i16> @llvm.ssub.sat.v16i16(<16 x i16>, <16 x i16>) nounwind readnone - -define <32 x i8> @test_psubusb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_psubusb: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpsubusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psubusb: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psubusb: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpsubusb (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psubusb: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsubusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psubusb: -; SKX: # %bb.0: -; SKX-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpsubusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psubusb: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpsubusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <32 x i8> @llvm.usub.sat.v32i8(<32 x i8> %a0, <32 x i8> %a1) - %2 = load <32 x i8>, <32 x i8> *%a2, align 32 - %3 = call <32 x i8> @llvm.usub.sat.v32i8(<32 x i8> %1, <32 x i8> %2) - ret <32 x i8> %3 -} -declare <32 x i8> @llvm.usub.sat.v32i8(<32 x i8>, <32 x i8>) nounwind readnone - -define <16 x i16> @test_psubusw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_psubusw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpsubusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psubusw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psubusw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpsubusw (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psubusw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsubusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psubusw: -; SKX: # %bb.0: -; SKX-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; SKX-NEXT: vpsubusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psubusw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpsubusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = call <16 x i16> @llvm.usub.sat.v16i16(<16 x i16> %a0, <16 x i16> %a1) - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = call <16 x i16> @llvm.usub.sat.v16i16(<16 x i16> %1, <16 x i16> %2) - ret <16 x i16> %3 -} -declare <16 x i16> @llvm.usub.sat.v16i16(<16 x i16>, <16 x i16>) nounwind readnone - -define <16 x i16> @test_psubw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_psubw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpsubw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: vpsubw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_psubw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpsubw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: vpsubw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_psubw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpsubw %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: vpsubw (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_psubw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpsubw %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: vpsubw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_psubw: -; SKX: # %bb.0: -; SKX-NEXT: vpsubw %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: vpsubw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_psubw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpsubw %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpsubw (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = sub <16 x i16> %a0, %a1 - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = sub <16 x i16> %1, %2 - ret <16 x i16> %3 -} - -define <32 x i8> @test_punpckhbw(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_punpckhbw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpunpckhbw {{.*#+}} ymm0 = 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] sched: [1:1.00] -; GENERIC-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15],ymm0[24],mem[24],ymm0[25],mem[25],ymm0[26],mem[26],ymm0[27],mem[27],ymm0[28],mem[28],ymm0[29],mem[29],ymm0[30],mem[30],ymm0[31],mem[31] sched: [8:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_punpckhbw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpunpckhbw {{.*#+}} ymm0 = 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] sched: [1:1.00] -; HASWELL-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15],ymm0[24],mem[24],ymm0[25],mem[25],ymm0[26],mem[26],ymm0[27],mem[27],ymm0[28],mem[28],ymm0[29],mem[29],ymm0[30],mem[30],ymm0[31],mem[31] sched: [8:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_punpckhbw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpunpckhbw {{.*#+}} ymm0 = 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] sched: [1:1.00] -; BROADWELL-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15],ymm0[24],mem[24],ymm0[25],mem[25],ymm0[26],mem[26],ymm0[27],mem[27],ymm0[28],mem[28],ymm0[29],mem[29],ymm0[30],mem[30],ymm0[31],mem[31] sched: [7:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_punpckhbw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpunpckhbw {{.*#+}} ymm0 = 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] sched: [1:1.00] -; SKYLAKE-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15],ymm0[24],mem[24],ymm0[25],mem[25],ymm0[26],mem[26],ymm0[27],mem[27],ymm0[28],mem[28],ymm0[29],mem[29],ymm0[30],mem[30],ymm0[31],mem[31] sched: [8:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_punpckhbw: -; SKX: # %bb.0: -; SKX-NEXT: vpunpckhbw {{.*#+}} ymm0 = 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] sched: [1:1.00] -; SKX-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15],ymm0[24],mem[24],ymm0[25],mem[25],ymm0[26],mem[26],ymm0[27],mem[27],ymm0[28],mem[28],ymm0[29],mem[29],ymm0[30],mem[30],ymm0[31],mem[31] sched: [8:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_punpckhbw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpunpckhbw {{.*#+}} ymm0 = 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] sched: [1:0.25] -; ZNVER1-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15],ymm0[24],mem[24],ymm0[25],mem[25],ymm0[26],mem[26],ymm0[27],mem[27],ymm0[28],mem[28],ymm0[29],mem[29],ymm0[30],mem[30],ymm0[31],mem[31] sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <32 x i8> %a0, <32 x i8> %a1, <32 x i32> <i32 8, i32 40, i32 9, i32 41, i32 10, i32 42, i32 11, i32 43, i32 12, i32 44, i32 13, i32 45, i32 14, i32 46, i32 15, i32 47, i32 24, i32 56, i32 25, i32 57, i32 26, i32 58, i32 27, i32 59, i32 28, i32 60, i32 29, i32 61, i32 30, i32 62, i32 31, i32 63> - %2 = load <32 x i8>, <32 x i8> *%a2, align 32 - %3 = shufflevector <32 x i8> %1, <32 x i8> %2, <32 x i32> <i32 8, i32 40, i32 9, i32 41, i32 10, i32 42, i32 11, i32 43, i32 12, i32 44, i32 13, i32 45, i32 14, i32 46, i32 15, i32 47, i32 24, i32 56, i32 25, i32 57, i32 26, i32 58, i32 27, i32 59, i32 28, i32 60, i32 29, i32 61, i32 30, i32 62, i32 31, i32 63> - ret <32 x i8> %3 -} - -define <8 x i32> @test_punpckhdq(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_punpckhdq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00] -; GENERIC-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:1.00] -; GENERIC-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50] -; GENERIC-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_punpckhdq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00] -; HASWELL-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:1.00] -; HASWELL-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50] -; HASWELL-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_punpckhdq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00] -; BROADWELL-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [7:1.00] -; BROADWELL-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50] -; BROADWELL-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_punpckhdq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00] -; SKYLAKE-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:1.00] -; SKYLAKE-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_punpckhdq: -; SKX: # %bb.0: -; SKX-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00] -; SKX-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:1.00] -; SKX-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50] -; SKX-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_punpckhdq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:0.25] -; ZNVER1-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:0.50] -; ZNVER1-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.25] -; ZNVER1-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <8 x i32> %a0, <8 x i32> %a1, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 6, i32 14, i32 7, i32 15> - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = shufflevector <8 x i32> %1, <8 x i32> %2, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 6, i32 14, i32 7, i32 15> - %4 = add <8 x i32> %3, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> - ret <8 x i32> %4 -} - -define <4 x i64> @test_punpckhqdq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { -; GENERIC-LABEL: test_punpckhqdq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpunpckhqdq {{.*#+}} ymm1 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00] -; GENERIC-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],mem[1],ymm0[3],mem[3] sched: [8:1.00] -; GENERIC-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_punpckhqdq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpunpckhqdq {{.*#+}} ymm1 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00] -; HASWELL-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],mem[1],ymm0[3],mem[3] sched: [8:1.00] -; HASWELL-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_punpckhqdq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpunpckhqdq {{.*#+}} ymm1 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00] -; BROADWELL-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],mem[1],ymm0[3],mem[3] sched: [7:1.00] -; BROADWELL-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_punpckhqdq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpunpckhqdq {{.*#+}} ymm1 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00] -; SKYLAKE-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],mem[1],ymm0[3],mem[3] sched: [8:1.00] -; SKYLAKE-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_punpckhqdq: -; SKX: # %bb.0: -; SKX-NEXT: vpunpckhqdq {{.*#+}} ymm1 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00] -; SKX-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],mem[1],ymm0[3],mem[3] sched: [8:1.00] -; SKX-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_punpckhqdq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpunpckhqdq {{.*#+}} ymm1 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:0.25] -; ZNVER1-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],mem[1],ymm0[3],mem[3] sched: [8:0.50] -; ZNVER1-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <4 x i64> %a0, <4 x i64> %a1, <4 x i32> <i32 1, i32 5, i32 3, i32 7> - %2 = load <4 x i64>, <4 x i64> *%a2, align 32 - %3 = shufflevector <4 x i64> %a0, <4 x i64> %2, <4 x i32> <i32 1, i32 5, i32 3, i32 7> - %4 = add <4 x i64> %1, %3 - ret <4 x i64> %4 -} - -define <16 x i16> @test_punpckhwd(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_punpckhwd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15] sched: [1:1.00] -; GENERIC-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15] sched: [8:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_punpckhwd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15] sched: [1:1.00] -; HASWELL-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15] sched: [8:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_punpckhwd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15] sched: [1:1.00] -; BROADWELL-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15] sched: [7:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_punpckhwd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15] sched: [1:1.00] -; SKYLAKE-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15] sched: [8:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_punpckhwd: -; SKX: # %bb.0: -; SKX-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15] sched: [1:1.00] -; SKX-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15] sched: [8:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_punpckhwd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15] sched: [1:0.25] -; ZNVER1-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15] sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <16 x i16> %a0, <16 x i16> %a1, <16 x i32> <i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31> - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = shufflevector <16 x i16> %1, <16 x i16> %2, <16 x i32> <i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31> - ret <16 x i16> %3 -} - -define <32 x i8> @test_punpcklbw(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) { -; GENERIC-LABEL: test_punpcklbw: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23] sched: [1:1.00] -; GENERIC-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[16],mem[16],ymm0[17],mem[17],ymm0[18],mem[18],ymm0[19],mem[19],ymm0[20],mem[20],ymm0[21],mem[21],ymm0[22],mem[22],ymm0[23],mem[23] sched: [8:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_punpcklbw: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23] sched: [1:1.00] -; HASWELL-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[16],mem[16],ymm0[17],mem[17],ymm0[18],mem[18],ymm0[19],mem[19],ymm0[20],mem[20],ymm0[21],mem[21],ymm0[22],mem[22],ymm0[23],mem[23] sched: [8:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_punpcklbw: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23] sched: [1:1.00] -; BROADWELL-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[16],mem[16],ymm0[17],mem[17],ymm0[18],mem[18],ymm0[19],mem[19],ymm0[20],mem[20],ymm0[21],mem[21],ymm0[22],mem[22],ymm0[23],mem[23] sched: [7:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_punpcklbw: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23] sched: [1:1.00] -; SKYLAKE-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[16],mem[16],ymm0[17],mem[17],ymm0[18],mem[18],ymm0[19],mem[19],ymm0[20],mem[20],ymm0[21],mem[21],ymm0[22],mem[22],ymm0[23],mem[23] sched: [8:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_punpcklbw: -; SKX: # %bb.0: -; SKX-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23] sched: [1:1.00] -; SKX-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[16],mem[16],ymm0[17],mem[17],ymm0[18],mem[18],ymm0[19],mem[19],ymm0[20],mem[20],ymm0[21],mem[21],ymm0[22],mem[22],ymm0[23],mem[23] sched: [8:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_punpcklbw: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23] sched: [1:0.25] -; ZNVER1-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[16],mem[16],ymm0[17],mem[17],ymm0[18],mem[18],ymm0[19],mem[19],ymm0[20],mem[20],ymm0[21],mem[21],ymm0[22],mem[22],ymm0[23],mem[23] sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <32 x i8> %a0, <32 x i8> %a1, <32 x i32> <i32 0, i32 32, i32 1, i32 33, i32 2, i32 34, i32 3, i32 35, i32 4, i32 36, i32 5, i32 37, i32 6, i32 38, i32 7, i32 39, i32 16, i32 48, i32 17, i32 49, i32 18, i32 50, i32 19, i32 51, i32 20, i32 52, i32 21, i32 53, i32 22, i32 54, i32 23, i32 55> - %2 = load <32 x i8>, <32 x i8> *%a2, align 32 - %3 = shufflevector <32 x i8> %1, <32 x i8> %2, <32 x i32> <i32 0, i32 32, i32 1, i32 33, i32 2, i32 34, i32 3, i32 35, i32 4, i32 36, i32 5, i32 37, i32 6, i32 38, i32 7, i32 39, i32 16, i32 48, i32 17, i32 49, i32 18, i32 50, i32 19, i32 51, i32 20, i32 52, i32 21, i32 53, i32 22, i32 54, i32 23, i32 55> - ret <32 x i8> %3 -} - -define <8 x i32> @test_punpckldq(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) { -; GENERIC-LABEL: test_punpckldq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00] -; GENERIC-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:1.00] -; GENERIC-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50] -; GENERIC-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_punpckldq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00] -; HASWELL-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:1.00] -; HASWELL-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50] -; HASWELL-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_punpckldq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00] -; BROADWELL-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [7:1.00] -; BROADWELL-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50] -; BROADWELL-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_punpckldq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00] -; SKYLAKE-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:1.00] -; SKYLAKE-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50] -; SKYLAKE-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_punpckldq: -; SKX: # %bb.0: -; SKX-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00] -; SKX-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:1.00] -; SKX-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50] -; SKX-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_punpckldq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:0.25] -; ZNVER1-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:0.50] -; ZNVER1-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.25] -; ZNVER1-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <8 x i32> %a0, <8 x i32> %a1, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 4, i32 12, i32 5, i32 13> - %2 = load <8 x i32>, <8 x i32> *%a2, align 32 - %3 = shufflevector <8 x i32> %1, <8 x i32> %2, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 4, i32 12, i32 5, i32 13> - %4 = add <8 x i32> %3, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> - ret <8 x i32> %4 -} - -define <4 x i64> @test_punpcklqdq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { -; GENERIC-LABEL: test_punpcklqdq: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00] -; GENERIC-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2] sched: [8:1.00] -; GENERIC-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_punpcklqdq: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00] -; HASWELL-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2] sched: [8:1.00] -; HASWELL-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_punpcklqdq: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00] -; BROADWELL-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2] sched: [7:1.00] -; BROADWELL-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_punpcklqdq: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00] -; SKYLAKE-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2] sched: [8:1.00] -; SKYLAKE-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_punpcklqdq: -; SKX: # %bb.0: -; SKX-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00] -; SKX-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2] sched: [8:1.00] -; SKX-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_punpcklqdq: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:0.25] -; ZNVER1-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2] sched: [8:0.50] -; ZNVER1-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <4 x i64> %a0, <4 x i64> %a1, <4 x i32> <i32 0, i32 4, i32 2, i32 6> - %2 = load <4 x i64>, <4 x i64> *%a2, align 32 - %3 = shufflevector <4 x i64> %a0, <4 x i64> %2, <4 x i32> <i32 0, i32 4, i32 2, i32 6> - %4 = add <4 x i64> %1, %3 - ret <4 x i64> %4 -} - -define <16 x i16> @test_punpcklwd(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) { -; GENERIC-LABEL: test_punpcklwd: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11] sched: [1:1.00] -; GENERIC-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11] sched: [8:1.00] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_punpcklwd: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11] sched: [1:1.00] -; HASWELL-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11] sched: [8:1.00] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_punpcklwd: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11] sched: [1:1.00] -; BROADWELL-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11] sched: [7:1.00] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_punpcklwd: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11] sched: [1:1.00] -; SKYLAKE-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11] sched: [8:1.00] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_punpcklwd: -; SKX: # %bb.0: -; SKX-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11] sched: [1:1.00] -; SKX-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11] sched: [8:1.00] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_punpcklwd: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11] sched: [1:0.25] -; ZNVER1-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11] sched: [8:0.50] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = shufflevector <16 x i16> %a0, <16 x i16> %a1, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27> - %2 = load <16 x i16>, <16 x i16> *%a2, align 32 - %3 = shufflevector <16 x i16> %1, <16 x i16> %2, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27> - ret <16 x i16> %3 -} - -define <4 x i64> @test_pxor(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) { -; GENERIC-LABEL: test_pxor: -; GENERIC: # %bb.0: -; GENERIC-NEXT: vpxor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; GENERIC-NEXT: vpxor (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; GENERIC-NEXT: retq # sched: [1:1.00] -; -; HASWELL-LABEL: test_pxor: -; HASWELL: # %bb.0: -; HASWELL-NEXT: vpxor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; HASWELL-NEXT: vpxor (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; HASWELL-NEXT: retq # sched: [7:1.00] -; -; BROADWELL-LABEL: test_pxor: -; BROADWELL: # %bb.0: -; BROADWELL-NEXT: vpxor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; BROADWELL-NEXT: vpxor (%rdi), %ymm0, %ymm0 # sched: [7:0.50] -; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50] -; BROADWELL-NEXT: retq # sched: [7:1.00] -; -; SKYLAKE-LABEL: test_pxor: -; SKYLAKE: # %bb.0: -; SKYLAKE-NEXT: vpxor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: vpxor (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKYLAKE-NEXT: retq # sched: [7:1.00] -; -; SKX-LABEL: test_pxor: -; SKX: # %bb.0: -; SKX-NEXT: vpxor %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: vpxor (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33] -; SKX-NEXT: retq # sched: [7:1.00] -; -; ZNVER1-LABEL: test_pxor: -; ZNVER1: # %bb.0: -; ZNVER1-NEXT: vpxor %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: vpxor (%rdi), %ymm0, %ymm0 # sched: [8:0.50] -; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25] -; ZNVER1-NEXT: retq # sched: [1:0.50] - %1 = xor <4 x i64> %a0, %a1 - %2 = load <4 x i64>, <4 x i64> *%a2, align 32 - %3 = xor <4 x i64> %1, %2 - %4 = add <4 x i64> %3, %a1 - ret <4 x i64> %4 -} - -!0 = !{i32 1} |
