summaryrefslogtreecommitdiffstats
path: root/llvm
diff options
context:
space:
mode:
Diffstat (limited to 'llvm')
-rw-r--r--llvm/test/CodeGen/X86/recip-fastmath.ll402
-rw-r--r--llvm/test/CodeGen/X86/recip-fastmath2.ll662
-rw-r--r--llvm/test/CodeGen/X86/sqrt-fastmath.ll350
3 files changed, 1337 insertions, 77 deletions
diff --git a/llvm/test/CodeGen/X86/recip-fastmath.ll b/llvm/test/CodeGen/X86/recip-fastmath.ll
index b1531c9a9d6..14f3255e041 100644
--- a/llvm/test/CodeGen/X86/recip-fastmath.ll
+++ b/llvm/test/CodeGen/X86/recip-fastmath.ll
@@ -833,6 +833,408 @@ define <8 x float> @v8f32_two_step(<8 x float> %x) #2 {
ret <8 x float> %div
}
+define <16 x float> @v16f32_no_estimate(<16 x float> %x) #0 {
+; SSE-LABEL: v16f32_no_estimate:
+; SSE: # %bb.0:
+; SSE-NEXT: movaps {{.*#+}} xmm4 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; SSE-NEXT: movaps %xmm4, %xmm5
+; SSE-NEXT: divps %xmm0, %xmm5
+; SSE-NEXT: movaps %xmm4, %xmm6
+; SSE-NEXT: divps %xmm1, %xmm6
+; SSE-NEXT: movaps %xmm4, %xmm7
+; SSE-NEXT: divps %xmm2, %xmm7
+; SSE-NEXT: divps %xmm3, %xmm4
+; SSE-NEXT: movaps %xmm5, %xmm0
+; SSE-NEXT: movaps %xmm6, %xmm1
+; SSE-NEXT: movaps %xmm7, %xmm2
+; SSE-NEXT: movaps %xmm4, %xmm3
+; SSE-NEXT: retq
+;
+; AVX-RECIP-LABEL: v16f32_no_estimate:
+; AVX-RECIP: # %bb.0:
+; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; AVX-RECIP-NEXT: vdivps %ymm0, %ymm2, %ymm0
+; AVX-RECIP-NEXT: vdivps %ymm1, %ymm2, %ymm1
+; AVX-RECIP-NEXT: retq
+;
+; FMA-RECIP-LABEL: v16f32_no_estimate:
+; FMA-RECIP: # %bb.0:
+; FMA-RECIP-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; FMA-RECIP-NEXT: vdivps %ymm0, %ymm2, %ymm0
+; FMA-RECIP-NEXT: vdivps %ymm1, %ymm2, %ymm1
+; FMA-RECIP-NEXT: retq
+;
+; BTVER2-LABEL: v16f32_no_estimate:
+; BTVER2: # %bb.0:
+; BTVER2-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [5:1.00]
+; BTVER2-NEXT: vdivps %ymm0, %ymm2, %ymm0 # sched: [38:38.00]
+; BTVER2-NEXT: vdivps %ymm1, %ymm2, %ymm1 # sched: [38:38.00]
+; BTVER2-NEXT: retq # sched: [4:1.00]
+;
+; SANDY-LABEL: v16f32_no_estimate:
+; SANDY: # %bb.0:
+; SANDY-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50]
+; SANDY-NEXT: vdivps %ymm0, %ymm2, %ymm0 # sched: [29:28.00]
+; SANDY-NEXT: vdivps %ymm1, %ymm2, %ymm1 # sched: [29:28.00]
+; SANDY-NEXT: retq # sched: [1:1.00]
+;
+; HASWELL-LABEL: v16f32_no_estimate:
+; HASWELL: # %bb.0:
+; HASWELL-NEXT: vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [7:0.50]
+; HASWELL-NEXT: vdivps %ymm0, %ymm2, %ymm0 # sched: [21:14.00]
+; HASWELL-NEXT: vdivps %ymm1, %ymm2, %ymm1 # sched: [21:14.00]
+; HASWELL-NEXT: retq # sched: [7:1.00]
+;
+; HASWELL-NO-FMA-LABEL: v16f32_no_estimate:
+; HASWELL-NO-FMA: # %bb.0:
+; HASWELL-NO-FMA-NEXT: vbroadcastss {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1]
+; HASWELL-NO-FMA-NEXT: vdivps %ymm0, %ymm2, %ymm0
+; HASWELL-NO-FMA-NEXT: vdivps %ymm1, %ymm2, %ymm1
+; HASWELL-NO-FMA-NEXT: retq
+;
+; KNL-LABEL: v16f32_no_estimate:
+; KNL: # %bb.0:
+; KNL-NEXT: vbroadcastss {{.*#+}} zmm1 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] sched: [10:1.00]
+; KNL-NEXT: vdivps %zmm0, %zmm1, %zmm0 # sched: [12:1.00]
+; KNL-NEXT: retq # sched: [7:1.00]
+;
+; SKX-LABEL: v16f32_no_estimate:
+; SKX: # %bb.0:
+; SKX-NEXT: vbroadcastss {{.*#+}} zmm1 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] sched: [8:0.50]
+; SKX-NEXT: vdivps %zmm0, %zmm1, %zmm0 # sched: [18:10.00]
+; SKX-NEXT: retq # sched: [7:1.00]
+ %div = fdiv fast <16 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %x
+ ret <16 x float> %div
+}
+
+define <16 x float> @v16f32_one_step(<16 x float> %x) #1 {
+; SSE-LABEL: v16f32_one_step:
+; SSE: # %bb.0:
+; SSE-NEXT: movaps %xmm3, %xmm4
+; SSE-NEXT: movaps %xmm0, %xmm5
+; SSE-NEXT: rcpps %xmm0, %xmm6
+; SSE-NEXT: mulps %xmm6, %xmm5
+; SSE-NEXT: movaps {{.*#+}} xmm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; SSE-NEXT: movaps %xmm3, %xmm0
+; SSE-NEXT: subps %xmm5, %xmm0
+; SSE-NEXT: mulps %xmm6, %xmm0
+; SSE-NEXT: addps %xmm6, %xmm0
+; SSE-NEXT: rcpps %xmm1, %xmm6
+; SSE-NEXT: mulps %xmm6, %xmm1
+; SSE-NEXT: movaps %xmm3, %xmm5
+; SSE-NEXT: subps %xmm1, %xmm5
+; SSE-NEXT: mulps %xmm6, %xmm5
+; SSE-NEXT: addps %xmm6, %xmm5
+; SSE-NEXT: rcpps %xmm2, %xmm1
+; SSE-NEXT: mulps %xmm1, %xmm2
+; SSE-NEXT: movaps %xmm3, %xmm6
+; SSE-NEXT: subps %xmm2, %xmm6
+; SSE-NEXT: mulps %xmm1, %xmm6
+; SSE-NEXT: addps %xmm1, %xmm6
+; SSE-NEXT: rcpps %xmm4, %xmm1
+; SSE-NEXT: mulps %xmm1, %xmm4
+; SSE-NEXT: subps %xmm4, %xmm3
+; SSE-NEXT: mulps %xmm1, %xmm3
+; SSE-NEXT: addps %xmm1, %xmm3
+; SSE-NEXT: movaps %xmm5, %xmm1
+; SSE-NEXT: movaps %xmm6, %xmm2
+; SSE-NEXT: retq
+;
+; AVX-RECIP-LABEL: v16f32_one_step:
+; AVX-RECIP: # %bb.0:
+; AVX-RECIP-NEXT: vrcpps %ymm0, %ymm2
+; AVX-RECIP-NEXT: vmulps %ymm2, %ymm0, %ymm0
+; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; AVX-RECIP-NEXT: vsubps %ymm0, %ymm3, %ymm0
+; AVX-RECIP-NEXT: vmulps %ymm0, %ymm2, %ymm0
+; AVX-RECIP-NEXT: vaddps %ymm0, %ymm2, %ymm0
+; AVX-RECIP-NEXT: vrcpps %ymm1, %ymm2
+; AVX-RECIP-NEXT: vmulps %ymm2, %ymm1, %ymm1
+; AVX-RECIP-NEXT: vsubps %ymm1, %ymm3, %ymm1
+; AVX-RECIP-NEXT: vmulps %ymm1, %ymm2, %ymm1
+; AVX-RECIP-NEXT: vaddps %ymm1, %ymm2, %ymm1
+; AVX-RECIP-NEXT: retq
+;
+; FMA-RECIP-LABEL: v16f32_one_step:
+; FMA-RECIP: # %bb.0:
+; FMA-RECIP-NEXT: vrcpps %ymm0, %ymm2
+; FMA-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm2 * ymm0) + ymm3
+; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm2) + ymm2
+; FMA-RECIP-NEXT: vrcpps %ymm1, %ymm2
+; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm1 = -(ymm2 * ymm1) + ymm3
+; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm2) + ymm2
+; FMA-RECIP-NEXT: retq
+;
+; BTVER2-LABEL: v16f32_one_step:
+; BTVER2: # %bb.0:
+; BTVER2-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [5:1.00]
+; BTVER2-NEXT: vrcpps %ymm0, %ymm2 # sched: [2:2.00]
+; BTVER2-NEXT: vrcpps %ymm1, %ymm4 # sched: [2:2.00]
+; BTVER2-NEXT: vmulps %ymm2, %ymm0, %ymm0 # sched: [2:2.00]
+; BTVER2-NEXT: vmulps %ymm4, %ymm1, %ymm1 # sched: [2:2.00]
+; BTVER2-NEXT: vsubps %ymm0, %ymm3, %ymm0 # sched: [3:2.00]
+; BTVER2-NEXT: vsubps %ymm1, %ymm3, %ymm1 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps %ymm0, %ymm2, %ymm0 # sched: [2:2.00]
+; BTVER2-NEXT: vmulps %ymm1, %ymm4, %ymm1 # sched: [2:2.00]
+; BTVER2-NEXT: vaddps %ymm0, %ymm2, %ymm0 # sched: [3:2.00]
+; BTVER2-NEXT: vaddps %ymm1, %ymm4, %ymm1 # sched: [3:2.00]
+; BTVER2-NEXT: retq # sched: [4:1.00]
+;
+; SANDY-LABEL: v16f32_one_step:
+; SANDY: # %bb.0:
+; SANDY-NEXT: vrcpps %ymm0, %ymm2 # sched: [7:2.00]
+; SANDY-NEXT: vmulps %ymm2, %ymm0, %ymm0 # sched: [5:1.00]
+; SANDY-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50]
+; SANDY-NEXT: vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm0, %ymm2, %ymm0 # sched: [5:1.00]
+; SANDY-NEXT: vaddps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
+; SANDY-NEXT: vrcpps %ymm1, %ymm2 # sched: [7:2.00]
+; SANDY-NEXT: vmulps %ymm2, %ymm1, %ymm1 # sched: [5:1.00]
+; SANDY-NEXT: vsubps %ymm1, %ymm3, %ymm1 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm1, %ymm2, %ymm1 # sched: [5:1.00]
+; SANDY-NEXT: vaddps %ymm1, %ymm2, %ymm1 # sched: [3:1.00]
+; SANDY-NEXT: retq # sched: [1:1.00]
+;
+; HASWELL-LABEL: v16f32_one_step:
+; HASWELL: # %bb.0:
+; HASWELL-NEXT: vrcpps %ymm0, %ymm2 # sched: [11:2.00]
+; HASWELL-NEXT: vbroadcastss {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] sched: [7:0.50]
+; HASWELL-NEXT: vrcpps %ymm1, %ymm4 # sched: [11:2.00]
+; HASWELL-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm2 * ymm0) + ymm3 sched: [5:0.50]
+; HASWELL-NEXT: vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm2) + ymm2 sched: [5:0.50]
+; HASWELL-NEXT: vfnmadd213ps {{.*#+}} ymm1 = -(ymm4 * ymm1) + ymm3 sched: [5:0.50]
+; HASWELL-NEXT: vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm4) + ymm4 sched: [5:0.50]
+; HASWELL-NEXT: retq # sched: [7:1.00]
+;
+; HASWELL-NO-FMA-LABEL: v16f32_one_step:
+; HASWELL-NO-FMA: # %bb.0:
+; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm2
+; HASWELL-NO-FMA-NEXT: vmulps %ymm2, %ymm0, %ymm0
+; HASWELL-NO-FMA-NEXT: vbroadcastss {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1]
+; HASWELL-NO-FMA-NEXT: vsubps %ymm0, %ymm3, %ymm0
+; HASWELL-NO-FMA-NEXT: vmulps %ymm0, %ymm2, %ymm0
+; HASWELL-NO-FMA-NEXT: vaddps %ymm0, %ymm2, %ymm0
+; HASWELL-NO-FMA-NEXT: vrcpps %ymm1, %ymm2
+; HASWELL-NO-FMA-NEXT: vmulps %ymm2, %ymm1, %ymm1
+; HASWELL-NO-FMA-NEXT: vsubps %ymm1, %ymm3, %ymm1
+; HASWELL-NO-FMA-NEXT: vmulps %ymm1, %ymm2, %ymm1
+; HASWELL-NO-FMA-NEXT: vaddps %ymm1, %ymm2, %ymm1
+; HASWELL-NO-FMA-NEXT: retq
+;
+; KNL-LABEL: v16f32_one_step:
+; KNL: # %bb.0:
+; KNL-NEXT: vbroadcastss {{.*#+}} zmm1 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] sched: [10:1.00]
+; KNL-NEXT: vdivps %zmm0, %zmm1, %zmm0 # sched: [12:1.00]
+; KNL-NEXT: retq # sched: [7:1.00]
+;
+; SKX-LABEL: v16f32_one_step:
+; SKX: # %bb.0:
+; SKX-NEXT: vbroadcastss {{.*#+}} zmm1 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] sched: [8:0.50]
+; SKX-NEXT: vdivps %zmm0, %zmm1, %zmm0 # sched: [18:10.00]
+; SKX-NEXT: retq # sched: [7:1.00]
+ %div = fdiv fast <16 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %x
+ ret <16 x float> %div
+}
+
+define <16 x float> @v16f32_two_step(<16 x float> %x) #2 {
+; SSE-LABEL: v16f32_two_step:
+; SSE: # %bb.0:
+; SSE-NEXT: movaps %xmm3, %xmm4
+; SSE-NEXT: movaps %xmm1, %xmm5
+; SSE-NEXT: movaps %xmm0, %xmm1
+; SSE-NEXT: rcpps %xmm0, %xmm0
+; SSE-NEXT: movaps %xmm1, %xmm6
+; SSE-NEXT: mulps %xmm0, %xmm6
+; SSE-NEXT: movaps {{.*#+}} xmm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; SSE-NEXT: movaps %xmm3, %xmm7
+; SSE-NEXT: subps %xmm6, %xmm7
+; SSE-NEXT: mulps %xmm0, %xmm7
+; SSE-NEXT: addps %xmm0, %xmm7
+; SSE-NEXT: mulps %xmm7, %xmm1
+; SSE-NEXT: movaps %xmm3, %xmm0
+; SSE-NEXT: subps %xmm1, %xmm0
+; SSE-NEXT: mulps %xmm7, %xmm0
+; SSE-NEXT: addps %xmm7, %xmm0
+; SSE-NEXT: rcpps %xmm5, %xmm1
+; SSE-NEXT: movaps %xmm5, %xmm6
+; SSE-NEXT: mulps %xmm1, %xmm6
+; SSE-NEXT: movaps %xmm3, %xmm7
+; SSE-NEXT: subps %xmm6, %xmm7
+; SSE-NEXT: mulps %xmm1, %xmm7
+; SSE-NEXT: addps %xmm1, %xmm7
+; SSE-NEXT: mulps %xmm7, %xmm5
+; SSE-NEXT: movaps %xmm3, %xmm1
+; SSE-NEXT: subps %xmm5, %xmm1
+; SSE-NEXT: mulps %xmm7, %xmm1
+; SSE-NEXT: addps %xmm7, %xmm1
+; SSE-NEXT: rcpps %xmm2, %xmm5
+; SSE-NEXT: movaps %xmm2, %xmm6
+; SSE-NEXT: mulps %xmm5, %xmm6
+; SSE-NEXT: movaps %xmm3, %xmm7
+; SSE-NEXT: subps %xmm6, %xmm7
+; SSE-NEXT: mulps %xmm5, %xmm7
+; SSE-NEXT: addps %xmm5, %xmm7
+; SSE-NEXT: mulps %xmm7, %xmm2
+; SSE-NEXT: movaps %xmm3, %xmm5
+; SSE-NEXT: subps %xmm2, %xmm5
+; SSE-NEXT: mulps %xmm7, %xmm5
+; SSE-NEXT: addps %xmm7, %xmm5
+; SSE-NEXT: rcpps %xmm4, %xmm2
+; SSE-NEXT: movaps %xmm4, %xmm6
+; SSE-NEXT: mulps %xmm2, %xmm6
+; SSE-NEXT: movaps %xmm3, %xmm7
+; SSE-NEXT: subps %xmm6, %xmm7
+; SSE-NEXT: mulps %xmm2, %xmm7
+; SSE-NEXT: addps %xmm2, %xmm7
+; SSE-NEXT: mulps %xmm7, %xmm4
+; SSE-NEXT: subps %xmm4, %xmm3
+; SSE-NEXT: mulps %xmm7, %xmm3
+; SSE-NEXT: addps %xmm7, %xmm3
+; SSE-NEXT: movaps %xmm5, %xmm2
+; SSE-NEXT: retq
+;
+; AVX-RECIP-LABEL: v16f32_two_step:
+; AVX-RECIP: # %bb.0:
+; AVX-RECIP-NEXT: vrcpps %ymm0, %ymm2
+; AVX-RECIP-NEXT: vmulps %ymm2, %ymm0, %ymm3
+; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm4 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; AVX-RECIP-NEXT: vsubps %ymm3, %ymm4, %ymm3
+; AVX-RECIP-NEXT: vmulps %ymm3, %ymm2, %ymm3
+; AVX-RECIP-NEXT: vaddps %ymm3, %ymm2, %ymm2
+; AVX-RECIP-NEXT: vmulps %ymm2, %ymm0, %ymm0
+; AVX-RECIP-NEXT: vsubps %ymm0, %ymm4, %ymm0
+; AVX-RECIP-NEXT: vmulps %ymm0, %ymm2, %ymm0
+; AVX-RECIP-NEXT: vaddps %ymm0, %ymm2, %ymm0
+; AVX-RECIP-NEXT: vrcpps %ymm1, %ymm2
+; AVX-RECIP-NEXT: vmulps %ymm2, %ymm1, %ymm3
+; AVX-RECIP-NEXT: vsubps %ymm3, %ymm4, %ymm3
+; AVX-RECIP-NEXT: vmulps %ymm3, %ymm2, %ymm3
+; AVX-RECIP-NEXT: vaddps %ymm3, %ymm2, %ymm2
+; AVX-RECIP-NEXT: vmulps %ymm2, %ymm1, %ymm1
+; AVX-RECIP-NEXT: vsubps %ymm1, %ymm4, %ymm1
+; AVX-RECIP-NEXT: vmulps %ymm1, %ymm2, %ymm1
+; AVX-RECIP-NEXT: vaddps %ymm1, %ymm2, %ymm1
+; AVX-RECIP-NEXT: retq
+;
+; FMA-RECIP-LABEL: v16f32_two_step:
+; FMA-RECIP: # %bb.0:
+; FMA-RECIP-NEXT: vrcpps %ymm0, %ymm2
+; FMA-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; FMA-RECIP-NEXT: vmovaps %ymm2, %ymm4
+; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm4 = -(ymm0 * ymm4) + ymm3
+; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm4 = (ymm4 * ymm2) + ymm2
+; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm4 * ymm0) + ymm3
+; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm4) + ymm4
+; FMA-RECIP-NEXT: vrcpps %ymm1, %ymm2
+; FMA-RECIP-NEXT: vmovaps %ymm2, %ymm4
+; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm4 = -(ymm1 * ymm4) + ymm3
+; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm4 = (ymm4 * ymm2) + ymm2
+; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm1 = -(ymm4 * ymm1) + ymm3
+; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm4) + ymm4
+; FMA-RECIP-NEXT: retq
+;
+; BTVER2-LABEL: v16f32_two_step:
+; BTVER2: # %bb.0:
+; BTVER2-NEXT: vmovaps {{.*#+}} ymm4 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [5:1.00]
+; BTVER2-NEXT: vrcpps %ymm0, %ymm2 # sched: [2:2.00]
+; BTVER2-NEXT: vmulps %ymm2, %ymm0, %ymm3 # sched: [2:2.00]
+; BTVER2-NEXT: vsubps %ymm3, %ymm4, %ymm3 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps %ymm3, %ymm2, %ymm3 # sched: [2:2.00]
+; BTVER2-NEXT: vaddps %ymm3, %ymm2, %ymm2 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps %ymm2, %ymm0, %ymm0 # sched: [2:2.00]
+; BTVER2-NEXT: vsubps %ymm0, %ymm4, %ymm0 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps %ymm0, %ymm2, %ymm0 # sched: [2:2.00]
+; BTVER2-NEXT: vaddps %ymm0, %ymm2, %ymm0 # sched: [3:2.00]
+; BTVER2-NEXT: vrcpps %ymm1, %ymm2 # sched: [2:2.00]
+; BTVER2-NEXT: vmulps %ymm2, %ymm1, %ymm3 # sched: [2:2.00]
+; BTVER2-NEXT: vsubps %ymm3, %ymm4, %ymm3 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps %ymm3, %ymm2, %ymm3 # sched: [2:2.00]
+; BTVER2-NEXT: vaddps %ymm3, %ymm2, %ymm2 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps %ymm2, %ymm1, %ymm1 # sched: [2:2.00]
+; BTVER2-NEXT: vsubps %ymm1, %ymm4, %ymm1 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps %ymm1, %ymm2, %ymm1 # sched: [2:2.00]
+; BTVER2-NEXT: vaddps %ymm1, %ymm2, %ymm1 # sched: [3:2.00]
+; BTVER2-NEXT: retq # sched: [4:1.00]
+;
+; SANDY-LABEL: v16f32_two_step:
+; SANDY: # %bb.0:
+; SANDY-NEXT: vrcpps %ymm0, %ymm2 # sched: [7:2.00]
+; SANDY-NEXT: vmulps %ymm2, %ymm0, %ymm3 # sched: [5:1.00]
+; SANDY-NEXT: vmovaps {{.*#+}} ymm4 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50]
+; SANDY-NEXT: vsubps %ymm3, %ymm4, %ymm3 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm3, %ymm2, %ymm3 # sched: [5:1.00]
+; SANDY-NEXT: vaddps %ymm3, %ymm2, %ymm2 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm2, %ymm0, %ymm0 # sched: [5:1.00]
+; SANDY-NEXT: vsubps %ymm0, %ymm4, %ymm0 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm0, %ymm2, %ymm0 # sched: [5:1.00]
+; SANDY-NEXT: vaddps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
+; SANDY-NEXT: vrcpps %ymm1, %ymm2 # sched: [7:2.00]
+; SANDY-NEXT: vmulps %ymm2, %ymm1, %ymm3 # sched: [5:1.00]
+; SANDY-NEXT: vsubps %ymm3, %ymm4, %ymm3 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm3, %ymm2, %ymm3 # sched: [5:1.00]
+; SANDY-NEXT: vaddps %ymm3, %ymm2, %ymm2 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm2, %ymm1, %ymm1 # sched: [5:1.00]
+; SANDY-NEXT: vsubps %ymm1, %ymm4, %ymm1 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm1, %ymm2, %ymm1 # sched: [5:1.00]
+; SANDY-NEXT: vaddps %ymm1, %ymm2, %ymm1 # sched: [3:1.00]
+; SANDY-NEXT: retq # sched: [1:1.00]
+;
+; HASWELL-LABEL: v16f32_two_step:
+; HASWELL: # %bb.0:
+; HASWELL-NEXT: vrcpps %ymm0, %ymm2 # sched: [11:2.00]
+; HASWELL-NEXT: vbroadcastss {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] sched: [7:0.50]
+; HASWELL-NEXT: vmovaps %ymm2, %ymm4 # sched: [1:1.00]
+; HASWELL-NEXT: vfnmadd213ps {{.*#+}} ymm4 = -(ymm0 * ymm4) + ymm3 sched: [5:0.50]
+; HASWELL-NEXT: vfmadd132ps {{.*#+}} ymm4 = (ymm4 * ymm2) + ymm2 sched: [5:0.50]
+; HASWELL-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm4 * ymm0) + ymm3 sched: [5:0.50]
+; HASWELL-NEXT: vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm4) + ymm4 sched: [5:0.50]
+; HASWELL-NEXT: vrcpps %ymm1, %ymm2 # sched: [11:2.00]
+; HASWELL-NEXT: vmovaps %ymm2, %ymm4 # sched: [1:1.00]
+; HASWELL-NEXT: vfnmadd213ps {{.*#+}} ymm4 = -(ymm1 * ymm4) + ymm3 sched: [5:0.50]
+; HASWELL-NEXT: vfmadd132ps {{.*#+}} ymm4 = (ymm4 * ymm2) + ymm2 sched: [5:0.50]
+; HASWELL-NEXT: vfnmadd213ps {{.*#+}} ymm1 = -(ymm4 * ymm1) + ymm3 sched: [5:0.50]
+; HASWELL-NEXT: vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm4) + ymm4 sched: [5:0.50]
+; HASWELL-NEXT: retq # sched: [7:1.00]
+;
+; HASWELL-NO-FMA-LABEL: v16f32_two_step:
+; HASWELL-NO-FMA: # %bb.0:
+; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm2
+; HASWELL-NO-FMA-NEXT: vmulps %ymm2, %ymm0, %ymm3
+; HASWELL-NO-FMA-NEXT: vbroadcastss {{.*#+}} ymm4 = [1,1,1,1,1,1,1,1]
+; HASWELL-NO-FMA-NEXT: vsubps %ymm3, %ymm4, %ymm3
+; HASWELL-NO-FMA-NEXT: vmulps %ymm3, %ymm2, %ymm3
+; HASWELL-NO-FMA-NEXT: vaddps %ymm3, %ymm2, %ymm2
+; HASWELL-NO-FMA-NEXT: vmulps %ymm2, %ymm0, %ymm0
+; HASWELL-NO-FMA-NEXT: vsubps %ymm0, %ymm4, %ymm0
+; HASWELL-NO-FMA-NEXT: vmulps %ymm0, %ymm2, %ymm0
+; HASWELL-NO-FMA-NEXT: vaddps %ymm0, %ymm2, %ymm0
+; HASWELL-NO-FMA-NEXT: vrcpps %ymm1, %ymm2
+; HASWELL-NO-FMA-NEXT: vmulps %ymm2, %ymm1, %ymm3
+; HASWELL-NO-FMA-NEXT: vsubps %ymm3, %ymm4, %ymm3
+; HASWELL-NO-FMA-NEXT: vmulps %ymm3, %ymm2, %ymm3
+; HASWELL-NO-FMA-NEXT: vaddps %ymm3, %ymm2, %ymm2
+; HASWELL-NO-FMA-NEXT: vmulps %ymm2, %ymm1, %ymm1
+; HASWELL-NO-FMA-NEXT: vsubps %ymm1, %ymm4, %ymm1
+; HASWELL-NO-FMA-NEXT: vmulps %ymm1, %ymm2, %ymm1
+; HASWELL-NO-FMA-NEXT: vaddps %ymm1, %ymm2, %ymm1
+; HASWELL-NO-FMA-NEXT: retq
+;
+; KNL-LABEL: v16f32_two_step:
+; KNL: # %bb.0:
+; KNL-NEXT: vbroadcastss {{.*#+}} zmm1 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] sched: [10:1.00]
+; KNL-NEXT: vdivps %zmm0, %zmm1, %zmm0 # sched: [12:1.00]
+; KNL-NEXT: retq # sched: [7:1.00]
+;
+; SKX-LABEL: v16f32_two_step:
+; SKX: # %bb.0:
+; SKX-NEXT: vbroadcastss {{.*#+}} zmm1 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] sched: [8:0.50]
+; SKX-NEXT: vdivps %zmm0, %zmm1, %zmm0 # sched: [18:10.00]
+; SKX-NEXT: retq # sched: [7:1.00]
+ %div = fdiv fast <16 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %x
+ ret <16 x float> %div
+}
+
attributes #0 = { "unsafe-fp-math"="true" "reciprocal-estimates"="!divf,!vec-divf" }
attributes #1 = { "unsafe-fp-math"="true" "reciprocal-estimates"="divf,vec-divf" }
attributes #2 = { "unsafe-fp-math"="true" "reciprocal-estimates"="divf:2,vec-divf:2" }
diff --git a/llvm/test/CodeGen/X86/recip-fastmath2.ll b/llvm/test/CodeGen/X86/recip-fastmath2.ll
index f6eec39a6eb..48672c18ce3 100644
--- a/llvm/test/CodeGen/X86/recip-fastmath2.ll
+++ b/llvm/test/CodeGen/X86/recip-fastmath2.ll
@@ -1190,6 +1190,668 @@ define <8 x float> @v8f32_no_step2(<8 x float> %x) #3 {
ret <8 x float> %div
}
+define <16 x float> @v16f32_one_step2(<16 x float> %x) #1 {
+; SSE-LABEL: v16f32_one_step2:
+; SSE: # %bb.0:
+; SSE-NEXT: movaps %xmm3, %xmm4
+; SSE-NEXT: movaps %xmm2, %xmm5
+; SSE-NEXT: movaps %xmm0, %xmm6
+; SSE-NEXT: rcpps %xmm3, %xmm2
+; SSE-NEXT: mulps %xmm2, %xmm4
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; SSE-NEXT: movaps %xmm0, %xmm3
+; SSE-NEXT: subps %xmm4, %xmm3
+; SSE-NEXT: mulps %xmm2, %xmm3
+; SSE-NEXT: addps %xmm2, %xmm3
+; SSE-NEXT: rcpps %xmm5, %xmm4
+; SSE-NEXT: mulps %xmm4, %xmm5
+; SSE-NEXT: movaps %xmm0, %xmm2
+; SSE-NEXT: subps %xmm5, %xmm2
+; SSE-NEXT: mulps %xmm4, %xmm2
+; SSE-NEXT: addps %xmm4, %xmm2
+; SSE-NEXT: rcpps %xmm1, %xmm5
+; SSE-NEXT: mulps %xmm5, %xmm1
+; SSE-NEXT: movaps %xmm0, %xmm4
+; SSE-NEXT: subps %xmm1, %xmm4
+; SSE-NEXT: mulps %xmm5, %xmm4
+; SSE-NEXT: addps %xmm5, %xmm4
+; SSE-NEXT: rcpps %xmm6, %xmm1
+; SSE-NEXT: mulps %xmm1, %xmm6
+; SSE-NEXT: subps %xmm6, %xmm0
+; SSE-NEXT: mulps %xmm1, %xmm0
+; SSE-NEXT: addps %xmm1, %xmm0
+; SSE-NEXT: mulps {{.*}}(%rip), %xmm0
+; SSE-NEXT: mulps {{.*}}(%rip), %xmm4
+; SSE-NEXT: mulps {{.*}}(%rip), %xmm2
+; SSE-NEXT: mulps {{.*}}(%rip), %xmm3
+; SSE-NEXT: movaps %xmm4, %xmm1
+; SSE-NEXT: retq
+;
+; AVX-RECIP-LABEL: v16f32_one_step2:
+; AVX-RECIP: # %bb.0:
+; AVX-RECIP-NEXT: vrcpps %ymm1, %ymm2
+; AVX-RECIP-NEXT: vmulps %ymm2, %ymm1, %ymm1
+; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; AVX-RECIP-NEXT: vsubps %ymm1, %ymm3, %ymm1
+; AVX-RECIP-NEXT: vmulps %ymm1, %ymm2, %ymm1
+; AVX-RECIP-NEXT: vaddps %ymm1, %ymm2, %ymm1
+; AVX-RECIP-NEXT: vrcpps %ymm0, %ymm2
+; AVX-RECIP-NEXT: vmulps %ymm2, %ymm0, %ymm0
+; AVX-RECIP-NEXT: vsubps %ymm0, %ymm3, %ymm0
+; AVX-RECIP-NEXT: vmulps %ymm0, %ymm2, %ymm0
+; AVX-RECIP-NEXT: vaddps %ymm0, %ymm2, %ymm0
+; AVX-RECIP-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0
+; AVX-RECIP-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1
+; AVX-RECIP-NEXT: retq
+;
+; FMA-RECIP-LABEL: v16f32_one_step2:
+; FMA-RECIP: # %bb.0:
+; FMA-RECIP-NEXT: vrcpps %ymm1, %ymm2
+; FMA-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm1 = -(ymm2 * ymm1) + ymm3
+; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm2) + ymm2
+; FMA-RECIP-NEXT: vrcpps %ymm0, %ymm2
+; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm2 * ymm0) + ymm3
+; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm2) + ymm2
+; FMA-RECIP-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0
+; FMA-RECIP-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1
+; FMA-RECIP-NEXT: retq
+;
+; BTVER2-LABEL: v16f32_one_step2:
+; BTVER2: # %bb.0:
+; BTVER2-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [5:1.00]
+; BTVER2-NEXT: vrcpps %ymm1, %ymm2 # sched: [2:2.00]
+; BTVER2-NEXT: vrcpps %ymm0, %ymm4 # sched: [2:2.00]
+; BTVER2-NEXT: vmulps %ymm2, %ymm1, %ymm1 # sched: [2:2.00]
+; BTVER2-NEXT: vmulps %ymm4, %ymm0, %ymm0 # sched: [2:2.00]
+; BTVER2-NEXT: vsubps %ymm1, %ymm3, %ymm1 # sched: [3:2.00]
+; BTVER2-NEXT: vsubps %ymm0, %ymm3, %ymm0 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps %ymm1, %ymm2, %ymm1 # sched: [2:2.00]
+; BTVER2-NEXT: vmulps %ymm0, %ymm4, %ymm0 # sched: [2:2.00]
+; BTVER2-NEXT: vaddps %ymm1, %ymm2, %ymm1 # sched: [3:2.00]
+; BTVER2-NEXT: vaddps %ymm0, %ymm4, %ymm0 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [7:2.00]
+; BTVER2-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [7:2.00]
+; BTVER2-NEXT: retq # sched: [4:1.00]
+;
+; SANDY-LABEL: v16f32_one_step2:
+; SANDY: # %bb.0:
+; SANDY-NEXT: vrcpps %ymm1, %ymm2 # sched: [7:2.00]
+; SANDY-NEXT: vmulps %ymm2, %ymm1, %ymm1 # sched: [5:1.00]
+; SANDY-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50]
+; SANDY-NEXT: vsubps %ymm1, %ymm3, %ymm1 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm1, %ymm2, %ymm1 # sched: [5:1.00]
+; SANDY-NEXT: vaddps %ymm1, %ymm2, %ymm1 # sched: [3:1.00]
+; SANDY-NEXT: vrcpps %ymm0, %ymm2 # sched: [7:2.00]
+; SANDY-NEXT: vmulps %ymm2, %ymm0, %ymm0 # sched: [5:1.00]
+; SANDY-NEXT: vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm0, %ymm2, %ymm0 # sched: [5:1.00]
+; SANDY-NEXT: vaddps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
+; SANDY-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:1.00]
+; SANDY-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:1.00]
+; SANDY-NEXT: retq # sched: [1:1.00]
+;
+; HASWELL-LABEL: v16f32_one_step2:
+; HASWELL: # %bb.0:
+; HASWELL-NEXT: vrcpps %ymm1, %ymm2 # sched: [11:2.00]
+; HASWELL-NEXT: vbroadcastss {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] sched: [7:0.50]
+; HASWELL-NEXT: vrcpps %ymm0, %ymm4 # sched: [11:2.00]
+; HASWELL-NEXT: vfnmadd213ps {{.*#+}} ymm1 = -(ymm2 * ymm1) + ymm3 sched: [5:0.50]
+; HASWELL-NEXT: vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm2) + ymm2 sched: [5:0.50]
+; HASWELL-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm4 * ymm0) + ymm3 sched: [5:0.50]
+; HASWELL-NEXT: vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm4) + ymm4 sched: [5:0.50]
+; HASWELL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
+; HASWELL-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:0.50]
+; HASWELL-NEXT: retq # sched: [7:1.00]
+;
+; HASWELL-NO-FMA-LABEL: v16f32_one_step2:
+; HASWELL-NO-FMA: # %bb.0:
+; HASWELL-NO-FMA-NEXT: vrcpps %ymm1, %ymm2 # sched: [11:2.00]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm2, %ymm1, %ymm1 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vbroadcastss {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] sched: [7:0.50]
+; HASWELL-NO-FMA-NEXT: vsubps %ymm1, %ymm3, %ymm1 # sched: [3:1.00]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm1, %ymm2, %ymm1 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vaddps %ymm1, %ymm2, %ymm1 # sched: [3:1.00]
+; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm2 # sched: [11:2.00]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm2, %ymm0, %ymm0 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm0, %ymm2, %ymm0 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vaddps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
+; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
+; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:0.50]
+; HASWELL-NO-FMA-NEXT: retq # sched: [7:1.00]
+;
+; KNL-LABEL: v16f32_one_step2:
+; KNL: # %bb.0:
+; KNL-NEXT: vmovaps {{.*#+}} zmm1 = [1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,5.000000e+00,6.000000e+00,7.000000e+00,8.000000e+00,9.000000e+00,1.000000e+01,1.100000e+01,1.200000e+01,1.300000e+01,1.400000e+01,1.500000e+01,1.600000e+01] sched: [5:0.50]
+; KNL-NEXT: vdivps %zmm0, %zmm1, %zmm0 # sched: [12:1.00]
+; KNL-NEXT: retq # sched: [7:1.00]
+;
+; SKX-LABEL: v16f32_one_step2:
+; SKX: # %bb.0:
+; SKX-NEXT: vmovaps {{.*#+}} zmm1 = [1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,5.000000e+00,6.000000e+00,7.000000e+00,8.000000e+00,9.000000e+00,1.000000e+01,1.100000e+01,1.200000e+01,1.300000e+01,1.400000e+01,1.500000e+01,1.600000e+01] sched: [8:0.50]
+; SKX-NEXT: vdivps %zmm0, %zmm1, %zmm0 # sched: [18:10.00]
+; SKX-NEXT: retq # sched: [7:1.00]
+ %div = fdiv fast <16 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0, float 9.0, float 10.0, float 11.0, float 12.0, float 13.0, float 14.0, float 15.0, float 16.0>, %x
+ ret <16 x float> %div
+}
+
+define <16 x float> @v16f32_one_step_2_divs(<16 x float> %x) #1 {
+; SSE-LABEL: v16f32_one_step_2_divs:
+; SSE: # %bb.0:
+; SSE-NEXT: rcpps %xmm0, %xmm6
+; SSE-NEXT: mulps %xmm6, %xmm0
+; SSE-NEXT: movaps {{.*#+}} xmm4 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; SSE-NEXT: movaps %xmm4, %xmm5
+; SSE-NEXT: subps %xmm0, %xmm5
+; SSE-NEXT: mulps %xmm6, %xmm5
+; SSE-NEXT: addps %xmm6, %xmm5
+; SSE-NEXT: rcpps %xmm1, %xmm0
+; SSE-NEXT: mulps %xmm0, %xmm1
+; SSE-NEXT: movaps %xmm4, %xmm6
+; SSE-NEXT: subps %xmm1, %xmm6
+; SSE-NEXT: mulps %xmm0, %xmm6
+; SSE-NEXT: addps %xmm0, %xmm6
+; SSE-NEXT: rcpps %xmm2, %xmm0
+; SSE-NEXT: mulps %xmm0, %xmm2
+; SSE-NEXT: movaps %xmm4, %xmm7
+; SSE-NEXT: subps %xmm2, %xmm7
+; SSE-NEXT: mulps %xmm0, %xmm7
+; SSE-NEXT: addps %xmm0, %xmm7
+; SSE-NEXT: rcpps %xmm3, %xmm0
+; SSE-NEXT: mulps %xmm0, %xmm3
+; SSE-NEXT: subps %xmm3, %xmm4
+; SSE-NEXT: mulps %xmm0, %xmm4
+; SSE-NEXT: addps %xmm0, %xmm4
+; SSE-NEXT: movaps {{.*#+}} xmm3 = [1.300000e+01,1.400000e+01,1.500000e+01,1.600000e+01]
+; SSE-NEXT: mulps %xmm4, %xmm3
+; SSE-NEXT: movaps {{.*#+}} xmm2 = [9.000000e+00,1.000000e+01,1.100000e+01,1.200000e+01]
+; SSE-NEXT: mulps %xmm7, %xmm2
+; SSE-NEXT: movaps {{.*#+}} xmm1 = [5.000000e+00,6.000000e+00,7.000000e+00,8.000000e+00]
+; SSE-NEXT: mulps %xmm6, %xmm1
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00]
+; SSE-NEXT: mulps %xmm5, %xmm0
+; SSE-NEXT: mulps %xmm5, %xmm0
+; SSE-NEXT: mulps %xmm6, %xmm1
+; SSE-NEXT: mulps %xmm7, %xmm2
+; SSE-NEXT: mulps %xmm4, %xmm3
+; SSE-NEXT: retq
+;
+; AVX-RECIP-LABEL: v16f32_one_step_2_divs:
+; AVX-RECIP: # %bb.0:
+; AVX-RECIP-NEXT: vrcpps %ymm0, %ymm2
+; AVX-RECIP-NEXT: vmulps %ymm2, %ymm0, %ymm0
+; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; AVX-RECIP-NEXT: vsubps %ymm0, %ymm3, %ymm0
+; AVX-RECIP-NEXT: vmulps %ymm0, %ymm2, %ymm0
+; AVX-RECIP-NEXT: vaddps %ymm0, %ymm2, %ymm0
+; AVX-RECIP-NEXT: vrcpps %ymm1, %ymm2
+; AVX-RECIP-NEXT: vmulps %ymm2, %ymm1, %ymm1
+; AVX-RECIP-NEXT: vsubps %ymm1, %ymm3, %ymm1
+; AVX-RECIP-NEXT: vmulps %ymm1, %ymm2, %ymm1
+; AVX-RECIP-NEXT: vaddps %ymm1, %ymm2, %ymm1
+; AVX-RECIP-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm2
+; AVX-RECIP-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm3
+; AVX-RECIP-NEXT: vmulps %ymm0, %ymm3, %ymm0
+; AVX-RECIP-NEXT: vmulps %ymm1, %ymm2, %ymm1
+; AVX-RECIP-NEXT: retq
+;
+; FMA-RECIP-LABEL: v16f32_one_step_2_divs:
+; FMA-RECIP: # %bb.0:
+; FMA-RECIP-NEXT: vrcpps %ymm0, %ymm2
+; FMA-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm2 * ymm0) + ymm3
+; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm2) + ymm2
+; FMA-RECIP-NEXT: vrcpps %ymm1, %ymm2
+; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm1 = -(ymm2 * ymm1) + ymm3
+; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm2) + ymm2
+; FMA-RECIP-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm2
+; FMA-RECIP-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm3
+; FMA-RECIP-NEXT: vmulps %ymm0, %ymm3, %ymm0
+; FMA-RECIP-NEXT: vmulps %ymm1, %ymm2, %ymm1
+; FMA-RECIP-NEXT: retq
+;
+; BTVER2-LABEL: v16f32_one_step_2_divs:
+; BTVER2: # %bb.0:
+; BTVER2-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [5:1.00]
+; BTVER2-NEXT: vrcpps %ymm0, %ymm2 # sched: [2:2.00]
+; BTVER2-NEXT: vmulps %ymm2, %ymm0, %ymm0 # sched: [2:2.00]
+; BTVER2-NEXT: vsubps %ymm0, %ymm3, %ymm0 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps %ymm0, %ymm2, %ymm0 # sched: [2:2.00]
+; BTVER2-NEXT: vaddps %ymm0, %ymm2, %ymm0 # sched: [3:2.00]
+; BTVER2-NEXT: vrcpps %ymm1, %ymm2 # sched: [2:2.00]
+; BTVER2-NEXT: vmulps %ymm2, %ymm1, %ymm1 # sched: [2:2.00]
+; BTVER2-NEXT: vsubps %ymm1, %ymm3, %ymm1 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm3 # sched: [7:2.00]
+; BTVER2-NEXT: vmulps %ymm1, %ymm2, %ymm1 # sched: [2:2.00]
+; BTVER2-NEXT: vaddps %ymm1, %ymm2, %ymm1 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm2 # sched: [7:2.00]
+; BTVER2-NEXT: vmulps %ymm0, %ymm3, %ymm0 # sched: [2:2.00]
+; BTVER2-NEXT: vmulps %ymm1, %ymm2, %ymm1 # sched: [2:2.00]
+; BTVER2-NEXT: retq # sched: [4:1.00]
+;
+; SANDY-LABEL: v16f32_one_step_2_divs:
+; SANDY: # %bb.0:
+; SANDY-NEXT: vrcpps %ymm0, %ymm2 # sched: [7:2.00]
+; SANDY-NEXT: vmulps %ymm2, %ymm0, %ymm0 # sched: [5:1.00]
+; SANDY-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50]
+; SANDY-NEXT: vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00]
+; SANDY-NEXT: vrcpps %ymm1, %ymm4 # sched: [7:2.00]
+; SANDY-NEXT: vmulps %ymm0, %ymm2, %ymm0 # sched: [5:1.00]
+; SANDY-NEXT: vaddps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm4, %ymm1, %ymm1 # sched: [5:1.00]
+; SANDY-NEXT: vsubps %ymm1, %ymm3, %ymm1 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm1, %ymm4, %ymm1 # sched: [5:1.00]
+; SANDY-NEXT: vaddps %ymm1, %ymm4, %ymm1 # sched: [3:1.00]
+; SANDY-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm2 # sched: [12:1.00]
+; SANDY-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm3 # sched: [12:1.00]
+; SANDY-NEXT: vmulps %ymm0, %ymm3, %ymm0 # sched: [5:1.00]
+; SANDY-NEXT: vmulps %ymm1, %ymm2, %ymm1 # sched: [5:1.00]
+; SANDY-NEXT: retq # sched: [1:1.00]
+;
+; HASWELL-LABEL: v16f32_one_step_2_divs:
+; HASWELL: # %bb.0:
+; HASWELL-NEXT: vrcpps %ymm0, %ymm2 # sched: [11:2.00]
+; HASWELL-NEXT: vbroadcastss {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] sched: [7:0.50]
+; HASWELL-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm2 * ymm0) + ymm3 sched: [5:0.50]
+; HASWELL-NEXT: vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm2) + ymm2 sched: [5:0.50]
+; HASWELL-NEXT: vrcpps %ymm1, %ymm2 # sched: [11:2.00]
+; HASWELL-NEXT: vfnmadd213ps {{.*#+}} ymm1 = -(ymm2 * ymm1) + ymm3 sched: [5:0.50]
+; HASWELL-NEXT: vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm2) + ymm2 sched: [5:0.50]
+; HASWELL-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm2 # sched: [12:0.50]
+; HASWELL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm3 # sched: [12:0.50]
+; HASWELL-NEXT: vmulps %ymm0, %ymm3, %ymm0 # sched: [5:0.50]
+; HASWELL-NEXT: vmulps %ymm1, %ymm2, %ymm1 # sched: [5:0.50]
+; HASWELL-NEXT: retq # sched: [7:1.00]
+;
+; HASWELL-NO-FMA-LABEL: v16f32_one_step_2_divs:
+; HASWELL-NO-FMA: # %bb.0:
+; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm2 # sched: [11:2.00]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm2, %ymm0, %ymm0 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vbroadcastss {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] sched: [7:0.50]
+; HASWELL-NO-FMA-NEXT: vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00]
+; HASWELL-NO-FMA-NEXT: vrcpps %ymm1, %ymm4 # sched: [11:2.00]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm0, %ymm2, %ymm0 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vaddps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm4, %ymm1, %ymm1 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vsubps %ymm1, %ymm3, %ymm1 # sched: [3:1.00]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm1, %ymm4, %ymm1 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vaddps %ymm1, %ymm4, %ymm1 # sched: [3:1.00]
+; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm2 # sched: [12:0.50]
+; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm3 # sched: [12:0.50]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm0, %ymm3, %ymm0 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm1, %ymm2, %ymm1 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: retq # sched: [7:1.00]
+;
+; KNL-LABEL: v16f32_one_step_2_divs:
+; KNL: # %bb.0:
+; KNL-NEXT: vbroadcastss {{.*#+}} zmm1 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] sched: [10:1.00]
+; KNL-NEXT: vdivps %zmm0, %zmm1, %zmm0 # sched: [12:1.00]
+; KNL-NEXT: vmulps {{.*}}(%rip), %zmm0, %zmm1 # sched: [12:0.50]
+; KNL-NEXT: vmulps %zmm0, %zmm1, %zmm0 # sched: [5:0.50]
+; KNL-NEXT: retq # sched: [7:1.00]
+;
+; SKX-LABEL: v16f32_one_step_2_divs:
+; SKX: # %bb.0:
+; SKX-NEXT: vbroadcastss {{.*#+}} zmm1 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] sched: [8:0.50]
+; SKX-NEXT: vdivps %zmm0, %zmm1, %zmm0 # sched: [18:10.00]
+; SKX-NEXT: vmulps {{.*}}(%rip), %zmm0, %zmm1 # sched: [11:0.50]
+; SKX-NEXT: vmulps %zmm0, %zmm1, %zmm0 # sched: [4:0.33]
+; SKX-NEXT: retq # sched: [7:1.00]
+ %div = fdiv fast <16 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0, float 9.0, float 10.0, float 11.0, float 12.0, float 13.0, float 14.0, float 15.0, float 16.0>, %x
+ %div2 = fdiv fast <16 x float> %div, %x
+ ret <16 x float> %div2
+}
+
+define <16 x float> @v16f32_two_step2(<16 x float> %x) #2 {
+; SSE-LABEL: v16f32_two_step2:
+; SSE: # %bb.0:
+; SSE-NEXT: movaps %xmm3, %xmm6
+; SSE-NEXT: movaps %xmm2, %xmm5
+; SSE-NEXT: movaps %xmm0, %xmm4
+; SSE-NEXT: rcpps %xmm3, %xmm2
+; SSE-NEXT: mulps %xmm2, %xmm3
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; SSE-NEXT: movaps %xmm0, %xmm7
+; SSE-NEXT: subps %xmm3, %xmm7
+; SSE-NEXT: mulps %xmm2, %xmm7
+; SSE-NEXT: addps %xmm2, %xmm7
+; SSE-NEXT: mulps %xmm7, %xmm6
+; SSE-NEXT: movaps %xmm0, %xmm3
+; SSE-NEXT: subps %xmm6, %xmm3
+; SSE-NEXT: mulps %xmm7, %xmm3
+; SSE-NEXT: addps %xmm7, %xmm3
+; SSE-NEXT: rcpps %xmm5, %xmm2
+; SSE-NEXT: movaps %xmm5, %xmm6
+; SSE-NEXT: mulps %xmm2, %xmm6
+; SSE-NEXT: movaps %xmm0, %xmm7
+; SSE-NEXT: subps %xmm6, %xmm7
+; SSE-NEXT: mulps %xmm2, %xmm7
+; SSE-NEXT: addps %xmm2, %xmm7
+; SSE-NEXT: mulps %xmm7, %xmm5
+; SSE-NEXT: movaps %xmm0, %xmm2
+; SSE-NEXT: subps %xmm5, %xmm2
+; SSE-NEXT: mulps %xmm7, %xmm2
+; SSE-NEXT: addps %xmm7, %xmm2
+; SSE-NEXT: rcpps %xmm1, %xmm5
+; SSE-NEXT: movaps %xmm1, %xmm6
+; SSE-NEXT: mulps %xmm5, %xmm6
+; SSE-NEXT: movaps %xmm0, %xmm7
+; SSE-NEXT: subps %xmm6, %xmm7
+; SSE-NEXT: mulps %xmm5, %xmm7
+; SSE-NEXT: addps %xmm5, %xmm7
+; SSE-NEXT: mulps %xmm7, %xmm1
+; SSE-NEXT: movaps %xmm0, %xmm5
+; SSE-NEXT: subps %xmm1, %xmm5
+; SSE-NEXT: mulps %xmm7, %xmm5
+; SSE-NEXT: addps %xmm7, %xmm5
+; SSE-NEXT: rcpps %xmm4, %xmm1
+; SSE-NEXT: movaps %xmm4, %xmm6
+; SSE-NEXT: mulps %xmm1, %xmm6
+; SSE-NEXT: movaps %xmm0, %xmm7
+; SSE-NEXT: subps %xmm6, %xmm7
+; SSE-NEXT: mulps %xmm1, %xmm7
+; SSE-NEXT: addps %xmm1, %xmm7
+; SSE-NEXT: mulps %xmm7, %xmm4
+; SSE-NEXT: subps %xmm4, %xmm0
+; SSE-NEXT: mulps %xmm7, %xmm0
+; SSE-NEXT: addps %xmm7, %xmm0
+; SSE-NEXT: mulps {{.*}}(%rip), %xmm0
+; SSE-NEXT: mulps {{.*}}(%rip), %xmm5
+; SSE-NEXT: mulps {{.*}}(%rip), %xmm2
+; SSE-NEXT: mulps {{.*}}(%rip), %xmm3
+; SSE-NEXT: movaps %xmm5, %xmm1
+; SSE-NEXT: retq
+;
+; AVX-RECIP-LABEL: v16f32_two_step2:
+; AVX-RECIP: # %bb.0:
+; AVX-RECIP-NEXT: vrcpps %ymm1, %ymm2
+; AVX-RECIP-NEXT: vmulps %ymm2, %ymm1, %ymm3
+; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm4 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; AVX-RECIP-NEXT: vsubps %ymm3, %ymm4, %ymm3
+; AVX-RECIP-NEXT: vmulps %ymm3, %ymm2, %ymm3
+; AVX-RECIP-NEXT: vaddps %ymm3, %ymm2, %ymm2
+; AVX-RECIP-NEXT: vmulps %ymm2, %ymm1, %ymm1
+; AVX-RECIP-NEXT: vsubps %ymm1, %ymm4, %ymm1
+; AVX-RECIP-NEXT: vmulps %ymm1, %ymm2, %ymm1
+; AVX-RECIP-NEXT: vaddps %ymm1, %ymm2, %ymm1
+; AVX-RECIP-NEXT: vrcpps %ymm0, %ymm2
+; AVX-RECIP-NEXT: vmulps %ymm2, %ymm0, %ymm3
+; AVX-RECIP-NEXT: vsubps %ymm3, %ymm4, %ymm3
+; AVX-RECIP-NEXT: vmulps %ymm3, %ymm2, %ymm3
+; AVX-RECIP-NEXT: vaddps %ymm3, %ymm2, %ymm2
+; AVX-RECIP-NEXT: vmulps %ymm2, %ymm0, %ymm0
+; AVX-RECIP-NEXT: vsubps %ymm0, %ymm4, %ymm0
+; AVX-RECIP-NEXT: vmulps %ymm0, %ymm2, %ymm0
+; AVX-RECIP-NEXT: vaddps %ymm0, %ymm2, %ymm0
+; AVX-RECIP-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0
+; AVX-RECIP-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1
+; AVX-RECIP-NEXT: retq
+;
+; FMA-RECIP-LABEL: v16f32_two_step2:
+; FMA-RECIP: # %bb.0:
+; FMA-RECIP-NEXT: vrcpps %ymm1, %ymm2
+; FMA-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; FMA-RECIP-NEXT: vmovaps %ymm2, %ymm4
+; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm4 = -(ymm1 * ymm4) + ymm3
+; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm4 = (ymm4 * ymm2) + ymm2
+; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm1 = -(ymm4 * ymm1) + ymm3
+; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm4) + ymm4
+; FMA-RECIP-NEXT: vrcpps %ymm0, %ymm2
+; FMA-RECIP-NEXT: vmovaps %ymm2, %ymm4
+; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm4 = -(ymm0 * ymm4) + ymm3
+; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm4 = (ymm4 * ymm2) + ymm2
+; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm4 * ymm0) + ymm3
+; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm4) + ymm4
+; FMA-RECIP-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0
+; FMA-RECIP-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1
+; FMA-RECIP-NEXT: retq
+;
+; BTVER2-LABEL: v16f32_two_step2:
+; BTVER2: # %bb.0:
+; BTVER2-NEXT: vmovaps {{.*#+}} ymm4 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [5:1.00]
+; BTVER2-NEXT: vrcpps %ymm1, %ymm2 # sched: [2:2.00]
+; BTVER2-NEXT: vmulps %ymm2, %ymm1, %ymm3 # sched: [2:2.00]
+; BTVER2-NEXT: vsubps %ymm3, %ymm4, %ymm3 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps %ymm3, %ymm2, %ymm3 # sched: [2:2.00]
+; BTVER2-NEXT: vaddps %ymm3, %ymm2, %ymm2 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps %ymm2, %ymm1, %ymm1 # sched: [2:2.00]
+; BTVER2-NEXT: vsubps %ymm1, %ymm4, %ymm1 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps %ymm1, %ymm2, %ymm1 # sched: [2:2.00]
+; BTVER2-NEXT: vaddps %ymm1, %ymm2, %ymm1 # sched: [3:2.00]
+; BTVER2-NEXT: vrcpps %ymm0, %ymm2 # sched: [2:2.00]
+; BTVER2-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [7:2.00]
+; BTVER2-NEXT: vmulps %ymm2, %ymm0, %ymm3 # sched: [2:2.00]
+; BTVER2-NEXT: vsubps %ymm3, %ymm4, %ymm3 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps %ymm3, %ymm2, %ymm3 # sched: [2:2.00]
+; BTVER2-NEXT: vaddps %ymm3, %ymm2, %ymm2 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps %ymm2, %ymm0, %ymm0 # sched: [2:2.00]
+; BTVER2-NEXT: vsubps %ymm0, %ymm4, %ymm0 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps %ymm0, %ymm2, %ymm0 # sched: [2:2.00]
+; BTVER2-NEXT: vaddps %ymm0, %ymm2, %ymm0 # sched: [3:2.00]
+; BTVER2-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [7:2.00]
+; BTVER2-NEXT: retq # sched: [4:1.00]
+;
+; SANDY-LABEL: v16f32_two_step2:
+; SANDY: # %bb.0:
+; SANDY-NEXT: vrcpps %ymm1, %ymm2 # sched: [7:2.00]
+; SANDY-NEXT: vmulps %ymm2, %ymm1, %ymm3 # sched: [5:1.00]
+; SANDY-NEXT: vmovaps {{.*#+}} ymm4 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50]
+; SANDY-NEXT: vsubps %ymm3, %ymm4, %ymm3 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm3, %ymm2, %ymm3 # sched: [5:1.00]
+; SANDY-NEXT: vaddps %ymm3, %ymm2, %ymm2 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm2, %ymm1, %ymm1 # sched: [5:1.00]
+; SANDY-NEXT: vsubps %ymm1, %ymm4, %ymm1 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm1, %ymm2, %ymm1 # sched: [5:1.00]
+; SANDY-NEXT: vaddps %ymm1, %ymm2, %ymm1 # sched: [3:1.00]
+; SANDY-NEXT: vrcpps %ymm0, %ymm2 # sched: [7:2.00]
+; SANDY-NEXT: vmulps %ymm2, %ymm0, %ymm3 # sched: [5:1.00]
+; SANDY-NEXT: vsubps %ymm3, %ymm4, %ymm3 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm3, %ymm2, %ymm3 # sched: [5:1.00]
+; SANDY-NEXT: vaddps %ymm3, %ymm2, %ymm2 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm2, %ymm0, %ymm0 # sched: [5:1.00]
+; SANDY-NEXT: vsubps %ymm0, %ymm4, %ymm0 # sched: [3:1.00]
+; SANDY-NEXT: vmulps %ymm0, %ymm2, %ymm0 # sched: [5:1.00]
+; SANDY-NEXT: vaddps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
+; SANDY-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:1.00]
+; SANDY-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:1.00]
+; SANDY-NEXT: retq # sched: [1:1.00]
+;
+; HASWELL-LABEL: v16f32_two_step2:
+; HASWELL: # %bb.0:
+; HASWELL-NEXT: vrcpps %ymm1, %ymm2 # sched: [11:2.00]
+; HASWELL-NEXT: vbroadcastss {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] sched: [7:0.50]
+; HASWELL-NEXT: vmovaps %ymm2, %ymm4 # sched: [1:1.00]
+; HASWELL-NEXT: vfnmadd213ps {{.*#+}} ymm4 = -(ymm1 * ymm4) + ymm3 sched: [5:0.50]
+; HASWELL-NEXT: vfmadd132ps {{.*#+}} ymm4 = (ymm4 * ymm2) + ymm2 sched: [5:0.50]
+; HASWELL-NEXT: vfnmadd213ps {{.*#+}} ymm1 = -(ymm4 * ymm1) + ymm3 sched: [5:0.50]
+; HASWELL-NEXT: vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm4) + ymm4 sched: [5:0.50]
+; HASWELL-NEXT: vrcpps %ymm0, %ymm2 # sched: [11:2.00]
+; HASWELL-NEXT: vmovaps %ymm2, %ymm4 # sched: [1:1.00]
+; HASWELL-NEXT: vfnmadd213ps {{.*#+}} ymm4 = -(ymm0 * ymm4) + ymm3 sched: [5:0.50]
+; HASWELL-NEXT: vfmadd132ps {{.*#+}} ymm4 = (ymm4 * ymm2) + ymm2 sched: [5:0.50]
+; HASWELL-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm4 * ymm0) + ymm3 sched: [5:0.50]
+; HASWELL-NEXT: vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm4) + ymm4 sched: [5:0.50]
+; HASWELL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
+; HASWELL-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:0.50]
+; HASWELL-NEXT: retq # sched: [7:1.00]
+;
+; HASWELL-NO-FMA-LABEL: v16f32_two_step2:
+; HASWELL-NO-FMA: # %bb.0:
+; HASWELL-NO-FMA-NEXT: vrcpps %ymm1, %ymm2 # sched: [11:2.00]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm2, %ymm1, %ymm3 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vbroadcastss {{.*#+}} ymm4 = [1,1,1,1,1,1,1,1] sched: [7:0.50]
+; HASWELL-NO-FMA-NEXT: vsubps %ymm3, %ymm4, %ymm3 # sched: [3:1.00]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm3, %ymm2, %ymm3 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vaddps %ymm3, %ymm2, %ymm2 # sched: [3:1.00]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm2, %ymm1, %ymm1 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vsubps %ymm1, %ymm4, %ymm1 # sched: [3:1.00]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm1, %ymm2, %ymm1 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vaddps %ymm1, %ymm2, %ymm1 # sched: [3:1.00]
+; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm2 # sched: [11:2.00]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm2, %ymm0, %ymm3 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vsubps %ymm3, %ymm4, %ymm3 # sched: [3:1.00]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm3, %ymm2, %ymm3 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vaddps %ymm3, %ymm2, %ymm2 # sched: [3:1.00]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm2, %ymm0, %ymm0 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vsubps %ymm0, %ymm4, %ymm0 # sched: [3:1.00]
+; HASWELL-NO-FMA-NEXT: vmulps %ymm0, %ymm2, %ymm0 # sched: [5:0.50]
+; HASWELL-NO-FMA-NEXT: vaddps %ymm0, %ymm2, %ymm0 # sched: [3:1.00]
+; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
+; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:0.50]
+; HASWELL-NO-FMA-NEXT: retq # sched: [7:1.00]
+;
+; KNL-LABEL: v16f32_two_step2:
+; KNL: # %bb.0:
+; KNL-NEXT: vmovaps {{.*#+}} zmm1 = [1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,5.000000e+00,6.000000e+00,7.000000e+00,8.000000e+00,9.000000e+00,1.000000e+01,1.100000e+01,1.200000e+01,1.300000e+01,1.400000e+01,1.500000e+01,1.600000e+01] sched: [5:0.50]
+; KNL-NEXT: vdivps %zmm0, %zmm1, %zmm0 # sched: [12:1.00]
+; KNL-NEXT: retq # sched: [7:1.00]
+;
+; SKX-LABEL: v16f32_two_step2:
+; SKX: # %bb.0:
+; SKX-NEXT: vmovaps {{.*#+}} zmm1 = [1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,5.000000e+00,6.000000e+00,7.000000e+00,8.000000e+00,9.000000e+00,1.000000e+01,1.100000e+01,1.200000e+01,1.300000e+01,1.400000e+01,1.500000e+01,1.600000e+01] sched: [8:0.50]
+; SKX-NEXT: vdivps %zmm0, %zmm1, %zmm0 # sched: [18:10.00]
+; SKX-NEXT: retq # sched: [7:1.00]
+ %div = fdiv fast <16 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0, float 9.0, float 10.0, float 11.0, float 12.0, float 13.0, float 14.0, float 15.0, float 16.0>, %x
+ ret <16 x float> %div
+}
+
+define <16 x float> @v16f32_no_step(<16 x float> %x) #3 {
+; SSE-LABEL: v16f32_no_step:
+; SSE: # %bb.0:
+; SSE-NEXT: rcpps %xmm0, %xmm0
+; SSE-NEXT: rcpps %xmm1, %xmm1
+; SSE-NEXT: rcpps %xmm2, %xmm2
+; SSE-NEXT: rcpps %xmm3, %xmm3
+; SSE-NEXT: retq
+;
+; AVX-RECIP-LABEL: v16f32_no_step:
+; AVX-RECIP: # %bb.0:
+; AVX-RECIP-NEXT: vrcpps %ymm0, %ymm0
+; AVX-RECIP-NEXT: vrcpps %ymm1, %ymm1
+; AVX-RECIP-NEXT: retq
+;
+; FMA-RECIP-LABEL: v16f32_no_step:
+; FMA-RECIP: # %bb.0:
+; FMA-RECIP-NEXT: vrcpps %ymm0, %ymm0
+; FMA-RECIP-NEXT: vrcpps %ymm1, %ymm1
+; FMA-RECIP-NEXT: retq
+;
+; BTVER2-LABEL: v16f32_no_step:
+; BTVER2: # %bb.0:
+; BTVER2-NEXT: vrcpps %ymm0, %ymm0 # sched: [2:2.00]
+; BTVER2-NEXT: vrcpps %ymm1, %ymm1 # sched: [2:2.00]
+; BTVER2-NEXT: retq # sched: [4:1.00]
+;
+; SANDY-LABEL: v16f32_no_step:
+; SANDY: # %bb.0:
+; SANDY-NEXT: vrcpps %ymm0, %ymm0 # sched: [7:2.00]
+; SANDY-NEXT: vrcpps %ymm1, %ymm1 # sched: [7:2.00]
+; SANDY-NEXT: retq # sched: [1:1.00]
+;
+; HASWELL-LABEL: v16f32_no_step:
+; HASWELL: # %bb.0:
+; HASWELL-NEXT: vrcpps %ymm0, %ymm0 # sched: [11:2.00]
+; HASWELL-NEXT: vrcpps %ymm1, %ymm1 # sched: [11:2.00]
+; HASWELL-NEXT: retq # sched: [7:1.00]
+;
+; HASWELL-NO-FMA-LABEL: v16f32_no_step:
+; HASWELL-NO-FMA: # %bb.0:
+; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm0 # sched: [11:2.00]
+; HASWELL-NO-FMA-NEXT: vrcpps %ymm1, %ymm1 # sched: [11:2.00]
+; HASWELL-NO-FMA-NEXT: retq # sched: [7:1.00]
+;
+; KNL-LABEL: v16f32_no_step:
+; KNL: # %bb.0:
+; KNL-NEXT: vbroadcastss {{.*#+}} zmm1 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] sched: [10:1.00]
+; KNL-NEXT: vdivps %zmm0, %zmm1, %zmm0 # sched: [12:1.00]
+; KNL-NEXT: retq # sched: [7:1.00]
+;
+; SKX-LABEL: v16f32_no_step:
+; SKX: # %bb.0:
+; SKX-NEXT: vbroadcastss {{.*#+}} zmm1 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1] sched: [8:0.50]
+; SKX-NEXT: vdivps %zmm0, %zmm1, %zmm0 # sched: [18:10.00]
+; SKX-NEXT: retq # sched: [7:1.00]
+ %div = fdiv fast <16 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %x
+ ret <16 x float> %div
+}
+
+define <16 x float> @v16f32_no_step2(<16 x float> %x) #3 {
+; SSE-LABEL: v16f32_no_step2:
+; SSE: # %bb.0:
+; SSE-NEXT: rcpps %xmm3, %xmm3
+; SSE-NEXT: rcpps %xmm2, %xmm2
+; SSE-NEXT: rcpps %xmm1, %xmm1
+; SSE-NEXT: rcpps %xmm0, %xmm0
+; SSE-NEXT: mulps {{.*}}(%rip), %xmm0
+; SSE-NEXT: mulps {{.*}}(%rip), %xmm1
+; SSE-NEXT: mulps {{.*}}(%rip), %xmm2
+; SSE-NEXT: mulps {{.*}}(%rip), %xmm3
+; SSE-NEXT: retq
+;
+; AVX-RECIP-LABEL: v16f32_no_step2:
+; AVX-RECIP: # %bb.0:
+; AVX-RECIP-NEXT: vrcpps %ymm1, %ymm1
+; AVX-RECIP-NEXT: vrcpps %ymm0, %ymm0
+; AVX-RECIP-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0
+; AVX-RECIP-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1
+; AVX-RECIP-NEXT: retq
+;
+; FMA-RECIP-LABEL: v16f32_no_step2:
+; FMA-RECIP: # %bb.0:
+; FMA-RECIP-NEXT: vrcpps %ymm1, %ymm1
+; FMA-RECIP-NEXT: vrcpps %ymm0, %ymm0
+; FMA-RECIP-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0
+; FMA-RECIP-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1
+; FMA-RECIP-NEXT: retq
+;
+; BTVER2-LABEL: v16f32_no_step2:
+; BTVER2: # %bb.0:
+; BTVER2-NEXT: vrcpps %ymm1, %ymm1 # sched: [2:2.00]
+; BTVER2-NEXT: vrcpps %ymm0, %ymm0 # sched: [2:2.00]
+; BTVER2-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [7:2.00]
+; BTVER2-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [7:2.00]
+; BTVER2-NEXT: retq # sched: [4:1.00]
+;
+; SANDY-LABEL: v16f32_no_step2:
+; SANDY: # %bb.0:
+; SANDY-NEXT: vrcpps %ymm1, %ymm1 # sched: [7:2.00]
+; SANDY-NEXT: vrcpps %ymm0, %ymm0 # sched: [7:2.00]
+; SANDY-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:1.00]
+; SANDY-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:1.00]
+; SANDY-NEXT: retq # sched: [1:1.00]
+;
+; HASWELL-LABEL: v16f32_no_step2:
+; HASWELL: # %bb.0:
+; HASWELL-NEXT: vrcpps %ymm1, %ymm1 # sched: [11:2.00]
+; HASWELL-NEXT: vrcpps %ymm0, %ymm0 # sched: [11:2.00]
+; HASWELL-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
+; HASWELL-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:0.50]
+; HASWELL-NEXT: retq # sched: [7:1.00]
+;
+; HASWELL-NO-FMA-LABEL: v16f32_no_step2:
+; HASWELL-NO-FMA: # %bb.0:
+; HASWELL-NO-FMA-NEXT: vrcpps %ymm1, %ymm1 # sched: [11:2.00]
+; HASWELL-NO-FMA-NEXT: vrcpps %ymm0, %ymm0 # sched: [11:2.00]
+; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %ymm0, %ymm0 # sched: [12:0.50]
+; HASWELL-NO-FMA-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1 # sched: [12:0.50]
+; HASWELL-NO-FMA-NEXT: retq # sched: [7:1.00]
+;
+; KNL-LABEL: v16f32_no_step2:
+; KNL: # %bb.0:
+; KNL-NEXT: vmovaps {{.*#+}} zmm1 = [1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,5.000000e+00,6.000000e+00,7.000000e+00,8.000000e+00,9.000000e+00,1.000000e+01,1.100000e+01,1.200000e+01,1.300000e+01,1.400000e+01,1.500000e+01,1.600000e+01] sched: [5:0.50]
+; KNL-NEXT: vdivps %zmm0, %zmm1, %zmm0 # sched: [12:1.00]
+; KNL-NEXT: retq # sched: [7:1.00]
+;
+; SKX-LABEL: v16f32_no_step2:
+; SKX: # %bb.0:
+; SKX-NEXT: vmovaps {{.*#+}} zmm1 = [1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00,5.000000e+00,6.000000e+00,7.000000e+00,8.000000e+00,9.000000e+00,1.000000e+01,1.100000e+01,1.200000e+01,1.300000e+01,1.400000e+01,1.500000e+01,1.600000e+01] sched: [8:0.50]
+; SKX-NEXT: vdivps %zmm0, %zmm1, %zmm0 # sched: [18:10.00]
+; SKX-NEXT: retq # sched: [7:1.00]
+ %div = fdiv fast <16 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0, float 9.0, float 10.0, float 11.0, float 12.0, float 13.0, float 14.0, float 15.0, float 16.0>, %x
+ ret <16 x float> %div
+}
+
attributes #0 = { "unsafe-fp-math"="true" "reciprocal-estimates"="!divf,!vec-divf" }
attributes #1 = { "unsafe-fp-math"="true" "reciprocal-estimates"="divf,vec-divf" }
attributes #2 = { "unsafe-fp-math"="true" "reciprocal-estimates"="divf:2,vec-divf:2" }
diff --git a/llvm/test/CodeGen/X86/sqrt-fastmath.ll b/llvm/test/CodeGen/X86/sqrt-fastmath.ll
index 07495e159e1..91ce663ccef 100644
--- a/llvm/test/CodeGen/X86/sqrt-fastmath.ll
+++ b/llvm/test/CodeGen/X86/sqrt-fastmath.ll
@@ -1,6 +1,7 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefix=CHECK --check-prefix=SSE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=CHECK --check-prefix=AVX
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX1
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX512
declare double @__sqrt_finite(double)
declare float @__sqrtf_finite(float)
@@ -8,6 +9,7 @@ declare x86_fp80 @__sqrtl_finite(x86_fp80)
declare float @llvm.sqrt.f32(float)
declare <4 x float> @llvm.sqrt.v4f32(<4 x float>)
declare <8 x float> @llvm.sqrt.v8f32(<8 x float>)
+declare <16 x float> @llvm.sqrt.v16f32(<16 x float>)
define double @finite_f64_no_estimate(double %d) #0 {
@@ -70,18 +72,31 @@ define float @finite_f32_estimate(float %f) #1 {
; SSE-NEXT: andnps %xmm2, %xmm0
; SSE-NEXT: retq
;
-; AVX-LABEL: finite_f32_estimate:
-; AVX: # %bb.0:
-; AVX-NEXT: vrsqrtss %xmm0, %xmm0, %xmm1
-; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm2
-; AVX-NEXT: vmulss %xmm1, %xmm2, %xmm1
-; AVX-NEXT: vaddss {{.*}}(%rip), %xmm1, %xmm1
-; AVX-NEXT: vmulss {{.*}}(%rip), %xmm2, %xmm2
-; AVX-NEXT: vmulss %xmm1, %xmm2, %xmm1
-; AVX-NEXT: vxorps %xmm2, %xmm2, %xmm2
-; AVX-NEXT: vcmpeqss %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vandnps %xmm1, %xmm0, %xmm0
-; AVX-NEXT: retq
+; AVX1-LABEL: finite_f32_estimate:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vrsqrtss %xmm0, %xmm0, %xmm1
+; AVX1-NEXT: vmulss %xmm1, %xmm0, %xmm2
+; AVX1-NEXT: vmulss %xmm1, %xmm2, %xmm1
+; AVX1-NEXT: vaddss {{.*}}(%rip), %xmm1, %xmm1
+; AVX1-NEXT: vmulss {{.*}}(%rip), %xmm2, %xmm2
+; AVX1-NEXT: vmulss %xmm1, %xmm2, %xmm1
+; AVX1-NEXT: vxorps %xmm2, %xmm2, %xmm2
+; AVX1-NEXT: vcmpeqss %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vandnps %xmm1, %xmm0, %xmm0
+; AVX1-NEXT: retq
+;
+; AVX512-LABEL: finite_f32_estimate:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vrsqrtss %xmm0, %xmm0, %xmm1
+; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm2
+; AVX512-NEXT: vfmadd213ss {{.*#+}} xmm1 = (xmm2 * xmm1) + mem
+; AVX512-NEXT: vmulss {{.*}}(%rip), %xmm2, %xmm2
+; AVX512-NEXT: vmulss %xmm1, %xmm2, %xmm1
+; AVX512-NEXT: vxorps %xmm2, %xmm2, %xmm2
+; AVX512-NEXT: vcmpeqss %xmm2, %xmm0, %k1
+; AVX512-NEXT: vmovss %xmm2, %xmm0, %xmm1 {%k1}
+; AVX512-NEXT: vmovaps %xmm1, %xmm0
+; AVX512-NEXT: retq
%call = tail call float @__sqrtf_finite(float %f) #2
ret float %call
}
@@ -126,18 +141,33 @@ define float @sqrtf_check_denorms(float %x) #3 {
; SSE-NEXT: andnps %xmm2, %xmm0
; SSE-NEXT: retq
;
-; AVX-LABEL: sqrtf_check_denorms:
-; AVX: # %bb.0:
-; AVX-NEXT: vrsqrtss %xmm0, %xmm0, %xmm1
-; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm2
-; AVX-NEXT: vmulss %xmm1, %xmm2, %xmm1
-; AVX-NEXT: vaddss {{.*}}(%rip), %xmm1, %xmm1
-; AVX-NEXT: vmulss {{.*}}(%rip), %xmm2, %xmm2
-; AVX-NEXT: vmulss %xmm1, %xmm2, %xmm1
-; AVX-NEXT: vandps {{.*}}(%rip), %xmm0, %xmm0
-; AVX-NEXT: vcmpltss {{.*}}(%rip), %xmm0, %xmm0
-; AVX-NEXT: vandnps %xmm1, %xmm0, %xmm0
-; AVX-NEXT: retq
+; AVX1-LABEL: sqrtf_check_denorms:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vrsqrtss %xmm0, %xmm0, %xmm1
+; AVX1-NEXT: vmulss %xmm1, %xmm0, %xmm2
+; AVX1-NEXT: vmulss %xmm1, %xmm2, %xmm1
+; AVX1-NEXT: vaddss {{.*}}(%rip), %xmm1, %xmm1
+; AVX1-NEXT: vmulss {{.*}}(%rip), %xmm2, %xmm2
+; AVX1-NEXT: vmulss %xmm1, %xmm2, %xmm1
+; AVX1-NEXT: vandps {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT: vcmpltss {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT: vandnps %xmm1, %xmm0, %xmm0
+; AVX1-NEXT: retq
+;
+; AVX512-LABEL: sqrtf_check_denorms:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vrsqrtss %xmm0, %xmm0, %xmm1
+; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm2
+; AVX512-NEXT: vfmadd213ss {{.*#+}} xmm1 = (xmm2 * xmm1) + mem
+; AVX512-NEXT: vmulss {{.*}}(%rip), %xmm2, %xmm2
+; AVX512-NEXT: vmulss %xmm1, %xmm2, %xmm1
+; AVX512-NEXT: vbroadcastss {{.*#+}} xmm2 = [NaN,NaN,NaN,NaN]
+; AVX512-NEXT: vandps %xmm2, %xmm0, %xmm0
+; AVX512-NEXT: vcmpltss {{.*}}(%rip), %xmm0, %k1
+; AVX512-NEXT: vxorps %xmm0, %xmm0, %xmm0
+; AVX512-NEXT: vmovss %xmm0, %xmm0, %xmm1 {%k1}
+; AVX512-NEXT: vmovaps %xmm1, %xmm0
+; AVX512-NEXT: retq
%call = tail call float @__sqrtf_finite(float %x) #2
ret float %call
}
@@ -160,19 +190,35 @@ define <4 x float> @sqrt_v4f32_check_denorms(<4 x float> %x) #3 {
; SSE-NEXT: movaps %xmm1, %xmm0
; SSE-NEXT: retq
;
-; AVX-LABEL: sqrt_v4f32_check_denorms:
-; AVX: # %bb.0:
-; AVX-NEXT: vrsqrtps %xmm0, %xmm1
-; AVX-NEXT: vmulps %xmm1, %xmm0, %xmm2
-; AVX-NEXT: vmulps {{.*}}(%rip), %xmm2, %xmm3
-; AVX-NEXT: vmulps %xmm1, %xmm2, %xmm1
-; AVX-NEXT: vaddps {{.*}}(%rip), %xmm1, %xmm1
-; AVX-NEXT: vmulps %xmm1, %xmm3, %xmm1
-; AVX-NEXT: vandps {{.*}}(%rip), %xmm0, %xmm0
-; AVX-NEXT: vmovaps {{.*#+}} xmm2 = [1.175494e-38,1.175494e-38,1.175494e-38,1.175494e-38]
-; AVX-NEXT: vcmpleps %xmm0, %xmm2, %xmm0
-; AVX-NEXT: vandps %xmm1, %xmm0, %xmm0
-; AVX-NEXT: retq
+; AVX1-LABEL: sqrt_v4f32_check_denorms:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vrsqrtps %xmm0, %xmm1
+; AVX1-NEXT: vmulps %xmm1, %xmm0, %xmm2
+; AVX1-NEXT: vmulps {{.*}}(%rip), %xmm2, %xmm3
+; AVX1-NEXT: vmulps %xmm1, %xmm2, %xmm1
+; AVX1-NEXT: vaddps {{.*}}(%rip), %xmm1, %xmm1
+; AVX1-NEXT: vmulps %xmm1, %xmm3, %xmm1
+; AVX1-NEXT: vandps {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT: vmovaps {{.*#+}} xmm2 = [1.175494e-38,1.175494e-38,1.175494e-38,1.175494e-38]
+; AVX1-NEXT: vcmpleps %xmm0, %xmm2, %xmm0
+; AVX1-NEXT: vandps %xmm1, %xmm0, %xmm0
+; AVX1-NEXT: retq
+;
+; AVX512-LABEL: sqrt_v4f32_check_denorms:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vrsqrtps %xmm0, %xmm1
+; AVX512-NEXT: vmulps %xmm1, %xmm0, %xmm2
+; AVX512-NEXT: vbroadcastss {{.*#+}} xmm3 = [-3,-3,-3,-3]
+; AVX512-NEXT: vfmadd231ps {{.*#+}} xmm3 = (xmm2 * xmm1) + xmm3
+; AVX512-NEXT: vbroadcastss {{.*#+}} xmm1 = [-0.5,-0.5,-0.5,-0.5]
+; AVX512-NEXT: vmulps %xmm3, %xmm1, %xmm1
+; AVX512-NEXT: vmulps %xmm1, %xmm2, %xmm1
+; AVX512-NEXT: vbroadcastss {{.*#+}} xmm2 = [NaN,NaN,NaN,NaN]
+; AVX512-NEXT: vandps %xmm2, %xmm0, %xmm0
+; AVX512-NEXT: vbroadcastss {{.*#+}} xmm2 = [1.17549435E-38,1.17549435E-38,1.17549435E-38,1.17549435E-38]
+; AVX512-NEXT: vcmpleps %xmm0, %xmm2, %xmm0
+; AVX512-NEXT: vandps %xmm1, %xmm0, %xmm0
+; AVX512-NEXT: retq
%call = tail call <4 x float> @llvm.sqrt.v4f32(<4 x float> %x) #2
ret <4 x float> %call
}
@@ -209,15 +255,24 @@ define float @f32_estimate(float %x) #1 {
; SSE-NEXT: movaps %xmm1, %xmm0
; SSE-NEXT: retq
;
-; AVX-LABEL: f32_estimate:
-; AVX: # %bb.0:
-; AVX-NEXT: vrsqrtss %xmm0, %xmm0, %xmm1
-; AVX-NEXT: vmulss %xmm1, %xmm1, %xmm2
-; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vaddss {{.*}}(%rip), %xmm0, %xmm0
-; AVX-NEXT: vmulss {{.*}}(%rip), %xmm1, %xmm1
-; AVX-NEXT: vmulss %xmm0, %xmm1, %xmm0
-; AVX-NEXT: retq
+; AVX1-LABEL: f32_estimate:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vrsqrtss %xmm0, %xmm0, %xmm1
+; AVX1-NEXT: vmulss %xmm1, %xmm1, %xmm2
+; AVX1-NEXT: vmulss %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vaddss {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT: vmulss {{.*}}(%rip), %xmm1, %xmm1
+; AVX1-NEXT: vmulss %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: retq
+;
+; AVX512-LABEL: f32_estimate:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vrsqrtss %xmm0, %xmm0, %xmm1
+; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm0
+; AVX512-NEXT: vfmadd213ss {{.*#+}} xmm0 = (xmm1 * xmm0) + mem
+; AVX512-NEXT: vmulss {{.*}}(%rip), %xmm1, %xmm1
+; AVX512-NEXT: vmulss %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: retq
%sqrt = tail call float @llvm.sqrt.f32(float %x)
%div = fdiv fast float 1.0, %sqrt
ret float %div
@@ -231,12 +286,19 @@ define <4 x float> @v4f32_no_estimate(<4 x float> %x) #0 {
; SSE-NEXT: divps %xmm1, %xmm0
; SSE-NEXT: retq
;
-; AVX-LABEL: v4f32_no_estimate:
-; AVX: # %bb.0:
-; AVX-NEXT: vsqrtps %xmm0, %xmm0
-; AVX-NEXT: vmovaps {{.*#+}} xmm1 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
-; AVX-NEXT: vdivps %xmm0, %xmm1, %xmm0
-; AVX-NEXT: retq
+; AVX1-LABEL: v4f32_no_estimate:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vsqrtps %xmm0, %xmm0
+; AVX1-NEXT: vmovaps {{.*#+}} xmm1 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; AVX1-NEXT: vdivps %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: retq
+;
+; AVX512-LABEL: v4f32_no_estimate:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vsqrtps %xmm0, %xmm0
+; AVX512-NEXT: vbroadcastss {{.*#+}} xmm1 = [1,1,1,1]
+; AVX512-NEXT: vdivps %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: retq
%sqrt = tail call <4 x float> @llvm.sqrt.v4f32(<4 x float> %x)
%div = fdiv fast <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>, %sqrt
ret <4 x float> %div
@@ -255,15 +317,26 @@ define <4 x float> @v4f32_estimate(<4 x float> %x) #1 {
; SSE-NEXT: movaps %xmm1, %xmm0
; SSE-NEXT: retq
;
-; AVX-LABEL: v4f32_estimate:
-; AVX: # %bb.0:
-; AVX-NEXT: vrsqrtps %xmm0, %xmm1
-; AVX-NEXT: vmulps %xmm1, %xmm1, %xmm2
-; AVX-NEXT: vmulps %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vaddps {{.*}}(%rip), %xmm0, %xmm0
-; AVX-NEXT: vmulps {{.*}}(%rip), %xmm1, %xmm1
-; AVX-NEXT: vmulps %xmm0, %xmm1, %xmm0
-; AVX-NEXT: retq
+; AVX1-LABEL: v4f32_estimate:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vrsqrtps %xmm0, %xmm1
+; AVX1-NEXT: vmulps %xmm1, %xmm1, %xmm2
+; AVX1-NEXT: vmulps %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vaddps {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT: vmulps {{.*}}(%rip), %xmm1, %xmm1
+; AVX1-NEXT: vmulps %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: retq
+;
+; AVX512-LABEL: v4f32_estimate:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vrsqrtps %xmm0, %xmm1
+; AVX512-NEXT: vmulps %xmm1, %xmm0, %xmm0
+; AVX512-NEXT: vbroadcastss {{.*#+}} xmm2 = [-3,-3,-3,-3]
+; AVX512-NEXT: vfmadd213ps {{.*#+}} xmm0 = (xmm1 * xmm0) + xmm2
+; AVX512-NEXT: vbroadcastss {{.*#+}} xmm2 = [-0.5,-0.5,-0.5,-0.5]
+; AVX512-NEXT: vmulps %xmm0, %xmm2, %xmm0
+; AVX512-NEXT: vmulps %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: retq
%sqrt = tail call <4 x float> @llvm.sqrt.v4f32(<4 x float> %x)
%div = fdiv fast <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>, %sqrt
ret <4 x float> %div
@@ -280,12 +353,19 @@ define <8 x float> @v8f32_no_estimate(<8 x float> %x) #0 {
; SSE-NEXT: divps %xmm2, %xmm1
; SSE-NEXT: retq
;
-; AVX-LABEL: v8f32_no_estimate:
-; AVX: # %bb.0:
-; AVX-NEXT: vsqrtps %ymm0, %ymm0
-; AVX-NEXT: vmovaps {{.*#+}} ymm1 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
-; AVX-NEXT: vdivps %ymm0, %ymm1, %ymm0
-; AVX-NEXT: retq
+; AVX1-LABEL: v8f32_no_estimate:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vsqrtps %ymm0, %ymm0
+; AVX1-NEXT: vmovaps {{.*#+}} ymm1 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; AVX1-NEXT: vdivps %ymm0, %ymm1, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX512-LABEL: v8f32_no_estimate:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vsqrtps %ymm0, %ymm0
+; AVX512-NEXT: vbroadcastss {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
+; AVX512-NEXT: vdivps %ymm0, %ymm1, %ymm0
+; AVX512-NEXT: retq
%sqrt = tail call <8 x float> @llvm.sqrt.v8f32(<8 x float> %x)
%div = fdiv fast <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %sqrt
ret <8 x float> %div
@@ -314,20 +394,136 @@ define <8 x float> @v8f32_estimate(<8 x float> %x) #1 {
; SSE-NEXT: movaps %xmm3, %xmm1
; SSE-NEXT: retq
;
-; AVX-LABEL: v8f32_estimate:
-; AVX: # %bb.0:
-; AVX-NEXT: vrsqrtps %ymm0, %ymm1
-; AVX-NEXT: vmulps %ymm1, %ymm1, %ymm2
-; AVX-NEXT: vmulps %ymm2, %ymm0, %ymm0
-; AVX-NEXT: vaddps {{.*}}(%rip), %ymm0, %ymm0
-; AVX-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1
-; AVX-NEXT: vmulps %ymm0, %ymm1, %ymm0
-; AVX-NEXT: retq
+; AVX1-LABEL: v8f32_estimate:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vrsqrtps %ymm0, %ymm1
+; AVX1-NEXT: vmulps %ymm1, %ymm1, %ymm2
+; AVX1-NEXT: vmulps %ymm2, %ymm0, %ymm0
+; AVX1-NEXT: vaddps {{.*}}(%rip), %ymm0, %ymm0
+; AVX1-NEXT: vmulps {{.*}}(%rip), %ymm1, %ymm1
+; AVX1-NEXT: vmulps %ymm0, %ymm1, %ymm0
+; AVX1-NEXT: retq
+;
+; AVX512-LABEL: v8f32_estimate:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vrsqrtps %ymm0, %ymm1
+; AVX512-NEXT: vmulps %ymm1, %ymm0, %ymm0
+; AVX512-NEXT: vbroadcastss {{.*#+}} ymm2 = [-3,-3,-3,-3,-3,-3,-3,-3]
+; AVX512-NEXT: vfmadd213ps {{.*#+}} ymm0 = (ymm1 * ymm0) + ymm2
+; AVX512-NEXT: vbroadcastss {{.*#+}} ymm2 = [-0.5,-0.5,-0.5,-0.5,-0.5,-0.5,-0.5,-0.5]
+; AVX512-NEXT: vmulps %ymm0, %ymm2, %ymm0
+; AVX512-NEXT: vmulps %ymm0, %ymm1, %ymm0
+; AVX512-NEXT: retq
%sqrt = tail call <8 x float> @llvm.sqrt.v8f32(<8 x float> %x)
%div = fdiv fast <8 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %sqrt
ret <8 x float> %div
}
+define <16 x float> @v16f32_no_estimate(<16 x float> %x) #0 {
+; SSE-LABEL: v16f32_no_estimate:
+; SSE: # %bb.0:
+; SSE-NEXT: sqrtps %xmm3, %xmm4
+; SSE-NEXT: sqrtps %xmm2, %xmm5
+; SSE-NEXT: sqrtps %xmm1, %xmm2
+; SSE-NEXT: sqrtps %xmm0, %xmm1
+; SSE-NEXT: movaps {{.*#+}} xmm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; SSE-NEXT: movaps %xmm3, %xmm0
+; SSE-NEXT: divps %xmm1, %xmm0
+; SSE-NEXT: movaps %xmm3, %xmm1
+; SSE-NEXT: divps %xmm2, %xmm1
+; SSE-NEXT: movaps %xmm3, %xmm2
+; SSE-NEXT: divps %xmm5, %xmm2
+; SSE-NEXT: divps %xmm4, %xmm3
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: v16f32_no_estimate:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vsqrtps %ymm1, %ymm1
+; AVX1-NEXT: vsqrtps %ymm0, %ymm0
+; AVX1-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00]
+; AVX1-NEXT: vdivps %ymm0, %ymm2, %ymm0
+; AVX1-NEXT: vdivps %ymm1, %ymm2, %ymm1
+; AVX1-NEXT: retq
+;
+; AVX512-LABEL: v16f32_no_estimate:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vsqrtps %zmm0, %zmm0
+; AVX512-NEXT: vbroadcastss {{.*#+}} zmm1 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
+; AVX512-NEXT: vdivps %zmm0, %zmm1, %zmm0
+; AVX512-NEXT: retq
+ %sqrt = tail call <16 x float> @llvm.sqrt.v16f32(<16 x float> %x)
+ %div = fdiv fast <16 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %sqrt
+ ret <16 x float> %div
+}
+
+define <16 x float> @v16f32_estimate(<16 x float> %x) #1 {
+; SSE-LABEL: v16f32_estimate:
+; SSE: # %bb.0:
+; SSE-NEXT: movaps %xmm1, %xmm4
+; SSE-NEXT: movaps %xmm0, %xmm1
+; SSE-NEXT: rsqrtps %xmm0, %xmm5
+; SSE-NEXT: movaps {{.*#+}} xmm6 = [-5.000000e-01,-5.000000e-01,-5.000000e-01,-5.000000e-01]
+; SSE-NEXT: movaps %xmm5, %xmm0
+; SSE-NEXT: mulps %xmm5, %xmm0
+; SSE-NEXT: mulps %xmm1, %xmm0
+; SSE-NEXT: movaps {{.*#+}} xmm7 = [-3.000000e+00,-3.000000e+00,-3.000000e+00,-3.000000e+00]
+; SSE-NEXT: addps %xmm7, %xmm0
+; SSE-NEXT: mulps %xmm6, %xmm0
+; SSE-NEXT: mulps %xmm5, %xmm0
+; SSE-NEXT: rsqrtps %xmm4, %xmm5
+; SSE-NEXT: movaps %xmm5, %xmm1
+; SSE-NEXT: mulps %xmm5, %xmm1
+; SSE-NEXT: mulps %xmm4, %xmm1
+; SSE-NEXT: addps %xmm7, %xmm1
+; SSE-NEXT: mulps %xmm6, %xmm1
+; SSE-NEXT: mulps %xmm5, %xmm1
+; SSE-NEXT: rsqrtps %xmm2, %xmm5
+; SSE-NEXT: movaps %xmm5, %xmm4
+; SSE-NEXT: mulps %xmm5, %xmm4
+; SSE-NEXT: mulps %xmm2, %xmm4
+; SSE-NEXT: addps %xmm7, %xmm4
+; SSE-NEXT: mulps %xmm6, %xmm4
+; SSE-NEXT: mulps %xmm5, %xmm4
+; SSE-NEXT: rsqrtps %xmm3, %xmm2
+; SSE-NEXT: movaps %xmm2, %xmm5
+; SSE-NEXT: mulps %xmm2, %xmm5
+; SSE-NEXT: mulps %xmm3, %xmm5
+; SSE-NEXT: addps %xmm7, %xmm5
+; SSE-NEXT: mulps %xmm6, %xmm5
+; SSE-NEXT: mulps %xmm2, %xmm5
+; SSE-NEXT: movaps %xmm4, %xmm2
+; SSE-NEXT: movaps %xmm5, %xmm3
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: v16f32_estimate:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vrsqrtps %ymm0, %ymm2
+; AVX1-NEXT: vmovaps {{.*#+}} ymm3 = [-5.000000e-01,-5.000000e-01,-5.000000e-01,-5.000000e-01,-5.000000e-01,-5.000000e-01,-5.000000e-01,-5.000000e-01]
+; AVX1-NEXT: vmulps %ymm2, %ymm2, %ymm4
+; AVX1-NEXT: vmulps %ymm4, %ymm0, %ymm0
+; AVX1-NEXT: vmovaps {{.*#+}} ymm4 = [-3.000000e+00,-3.000000e+00,-3.000000e+00,-3.000000e+00,-3.000000e+00,-3.000000e+00,-3.000000e+00,-3.000000e+00]
+; AVX1-NEXT: vaddps %ymm4, %ymm0, %ymm0
+; AVX1-NEXT: vmulps %ymm0, %ymm3, %ymm0
+; AVX1-NEXT: vmulps %ymm0, %ymm2, %ymm0
+; AVX1-NEXT: vrsqrtps %ymm1, %ymm2
+; AVX1-NEXT: vmulps %ymm2, %ymm2, %ymm5
+; AVX1-NEXT: vmulps %ymm5, %ymm1, %ymm1
+; AVX1-NEXT: vaddps %ymm4, %ymm1, %ymm1
+; AVX1-NEXT: vmulps %ymm1, %ymm3, %ymm1
+; AVX1-NEXT: vmulps %ymm1, %ymm2, %ymm1
+; AVX1-NEXT: retq
+;
+; AVX512-LABEL: v16f32_estimate:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vsqrtps %zmm0, %zmm0
+; AVX512-NEXT: vbroadcastss {{.*#+}} zmm1 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
+; AVX512-NEXT: vdivps %zmm0, %zmm1, %zmm0
+; AVX512-NEXT: retq
+ %sqrt = tail call <16 x float> @llvm.sqrt.v16f32(<16 x float> %x)
+ %div = fdiv fast <16 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, %sqrt
+ ret <16 x float> %div
+}
+
attributes #0 = { "unsafe-fp-math"="true" "reciprocal-estimates"="!sqrtf,!vec-sqrtf,!divf,!vec-divf" }
attributes #1 = { "unsafe-fp-math"="true" "reciprocal-estimates"="sqrt,vec-sqrt" }
OpenPOWER on IntegriCloud