[SelectionDAG] allow FP binops in SimplifyDemandedVectorElts

This is intended to make the backend on par with functionality that was
added to the IR version of SimplifyDemandedVectorElts in:
rL343727
...and the original motivation is that we need to improve demanded-vector-elements
in several ways to avoid problems that would be exposed in D51553.

Differential Revision: https://reviews.llvm.org/D52912

llvm-svn: 344541

diff --git a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
index d3a5078..150d22c 100644 (file)
--- a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -1712,7 +1712,12 @@ bool TargetLowering::SimplifyDemandedVectorElts(
     break;
   }
   case ISD::ADD:
-  case ISD::SUB: {
+  case ISD::SUB:
+  case ISD::FADD:
+  case ISD::FSUB:
+  case ISD::FMUL:
+  case ISD::FDIV:
+  case ISD::FREM: {
     APInt SrcUndef, SrcZero;
     if (SimplifyDemandedVectorElts(Op.getOperand(1), DemandedElts, SrcUndef,
                                    SrcZero, TLO, Depth + 1))

diff --git a/llvm/test/CodeGen/X86/avx512-hadd-hsub.ll b/llvm/test/CodeGen/X86/avx512-hadd-hsub.ll
index 510553b..aed1821 100644 (file)
--- a/llvm/test/CodeGen/X86/avx512-hadd-hsub.ll
+++ b/llvm/test/CodeGen/X86/avx512-hadd-hsub.ll
@@ -178,16 +178,16 @@ define <8 x double> @fhadd_16_4(<8 x double> %x225, <8 x double> %x227) {
 define <4 x double> @fadd_noundef_low(<8 x double> %x225, <8 x double> %x227) {
 ; KNL-LABEL: fadd_noundef_low:
 ; KNL:       # %bb.0:
-; KNL-NEXT:    vunpcklpd {{.*#+}} zmm2 = zmm0[0],zmm1[0],zmm0[2],zmm1[2],zmm0[4],zmm1[4],zmm0[6],zmm1[6]
-; KNL-NEXT:    vunpckhpd {{.*#+}} zmm0 = zmm0[1],zmm1[1],zmm0[3],zmm1[3],zmm0[5],zmm1[5],zmm0[7],zmm1[7]
+; KNL-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; KNL-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; KNL-NEXT:    vaddpd %zmm0, %zmm2, %zmm0
 ; KNL-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: fadd_noundef_low:
 ; SKX:       # %bb.0:
-; SKX-NEXT:    vunpcklpd {{.*#+}} zmm2 = zmm0[0],zmm1[0],zmm0[2],zmm1[2],zmm0[4],zmm1[4],zmm0[6],zmm1[6]
-; SKX-NEXT:    vunpckhpd {{.*#+}} zmm0 = zmm0[1],zmm1[1],zmm0[3],zmm1[3],zmm0[5],zmm1[5],zmm0[7],zmm1[7]
+; SKX-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; SKX-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; SKX-NEXT:    vaddpd %zmm0, %zmm2, %zmm0
 ; SKX-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
 ; SKX-NEXT:    retq
@@ -252,17 +252,15 @@ define <8 x i32> @hadd_16_3_sv(<16 x i32> %x225, <16 x i32> %x227) {
 define double @fadd_noundef_eel(<8 x double> %x225, <8 x double> %x227) {
 ; KNL-LABEL: fadd_noundef_eel:
 ; KNL:       # %bb.0:
-; KNL-NEXT:    vunpcklpd {{.*#+}} zmm2 = zmm0[0],zmm1[0],zmm0[2],zmm1[2],zmm0[4],zmm1[4],zmm0[6],zmm1[6]
-; KNL-NEXT:    vunpckhpd {{.*#+}} zmm0 = zmm0[1],zmm1[1],zmm0[3],zmm1[3],zmm0[5],zmm1[5],zmm0[7],zmm1[7]
-; KNL-NEXT:    vaddpd %zmm0, %zmm2, %zmm0
+; KNL-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; KNL-NEXT:    vaddpd %zmm1, %zmm0, %zmm0
 ; KNL-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: fadd_noundef_eel:
 ; SKX:       # %bb.0:
-; SKX-NEXT:    vunpcklpd {{.*#+}} zmm2 = zmm0[0],zmm1[0],zmm0[2],zmm1[2],zmm0[4],zmm1[4],zmm0[6],zmm1[6]
-; SKX-NEXT:    vunpckhpd {{.*#+}} zmm0 = zmm0[1],zmm1[1],zmm0[3],zmm1[3],zmm0[5],zmm1[5],zmm0[7],zmm1[7]
-; SKX-NEXT:    vaddpd %zmm0, %zmm2, %zmm0
+; SKX-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; SKX-NEXT:    vaddpd %zmm1, %zmm0, %zmm0
 ; SKX-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
 ; SKX-NEXT:    vzeroupper
 ; SKX-NEXT:    retq
@@ -278,18 +276,18 @@ define double @fadd_noundef_eel(<8 x double> %x225, <8 x double> %x227) {
 define double @fsub_noundef_ee (<8 x double> %x225, <8 x double> %x227) {
 ; KNL-LABEL: fsub_noundef_ee:
 ; KNL:       # %bb.0:
-; KNL-NEXT:    vunpcklpd {{.*#+}} zmm2 = zmm0[0],zmm1[0],zmm0[2],zmm1[2],zmm0[4],zmm1[4],zmm0[6],zmm1[6]
-; KNL-NEXT:    vunpckhpd {{.*#+}} zmm0 = zmm0[1],zmm1[1],zmm0[3],zmm1[3],zmm0[5],zmm1[5],zmm0[7],zmm1[7]
-; KNL-NEXT:    vsubpd %zmm0, %zmm2, %zmm0
+; KNL-NEXT:    vextractf32x4 $2, %zmm1, %xmm0
+; KNL-NEXT:    vbroadcastsd %xmm0, %zmm0
+; KNL-NEXT:    vsubpd %zmm1, %zmm0, %zmm0
 ; KNL-NEXT:    vextractf32x4 $2, %zmm0, %xmm0
 ; KNL-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
 ; KNL-NEXT:    retq
 ;
 ; SKX-LABEL: fsub_noundef_ee:
 ; SKX:       # %bb.0:
-; SKX-NEXT:    vunpcklpd {{.*#+}} zmm2 = zmm0[0],zmm1[0],zmm0[2],zmm1[2],zmm0[4],zmm1[4],zmm0[6],zmm1[6]
-; SKX-NEXT:    vunpckhpd {{.*#+}} zmm0 = zmm0[1],zmm1[1],zmm0[3],zmm1[3],zmm0[5],zmm1[5],zmm0[7],zmm1[7]
-; SKX-NEXT:    vsubpd %zmm0, %zmm2, %zmm0
+; SKX-NEXT:    vextractf32x4 $2, %zmm1, %xmm0
+; SKX-NEXT:    vbroadcastsd %xmm0, %zmm0
+; SKX-NEXT:    vsubpd %zmm1, %zmm0, %zmm0
 ; SKX-NEXT:    vextractf32x4 $2, %zmm0, %xmm0
 ; SKX-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
 ; SKX-NEXT:    vzeroupper

diff --git a/llvm/test/CodeGen/X86/avx512-intrinsics-fast-isel.ll b/llvm/test/CodeGen/X86/avx512-intrinsics-fast-isel.ll
index fa37d21..aa89ee7 100644 (file)
--- a/llvm/test/CodeGen/X86/avx512-intrinsics-fast-isel.ll
+++ b/llvm/test/CodeGen/X86/avx512-intrinsics-fast-isel.ll
@@ -7304,7 +7304,7 @@ define float @test_mm512_reduce_add_ps(<16 x float> %__W) {
 ; X86-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; X86-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
 ; X86-NEXT:    vaddps %xmm1, %xmm0, %xmm0
-; X86-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,0,3,2]
+; X86-NEXT:    vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
 ; X86-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; X86-NEXT:    vmovss %xmm0, (%esp)
 ; X86-NEXT:    flds (%esp)
@@ -7321,7 +7321,7 @@ define float @test_mm512_reduce_add_ps(<16 x float> %__W) {
 ; X64-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; X64-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
 ; X64-NEXT:    vaddps %xmm1, %xmm0, %xmm0
-; X64-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,0,3,2]
+; X64-NEXT:    vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
 ; X64-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; X64-NEXT:    vzeroupper
 ; X64-NEXT:    retq
@@ -7354,7 +7354,7 @@ define float @test_mm512_reduce_mul_ps(<16 x float> %__W) {
 ; X86-NEXT:    vmulps %xmm1, %xmm0, %xmm0
 ; X86-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
 ; X86-NEXT:    vmulps %xmm1, %xmm0, %xmm0
-; X86-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,0,3,2]
+; X86-NEXT:    vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
 ; X86-NEXT:    vmulps %xmm1, %xmm0, %xmm0
 ; X86-NEXT:    vmovss %xmm0, (%esp)
 ; X86-NEXT:    flds (%esp)
@@ -7371,7 +7371,7 @@ define float @test_mm512_reduce_mul_ps(<16 x float> %__W) {
 ; X64-NEXT:    vmulps %xmm1, %xmm0, %xmm0
 ; X64-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
 ; X64-NEXT:    vmulps %xmm1, %xmm0, %xmm0
-; X64-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,0,3,2]
+; X64-NEXT:    vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
 ; X64-NEXT:    vmulps %xmm1, %xmm0, %xmm0
 ; X64-NEXT:    vzeroupper
 ; X64-NEXT:    retq
@@ -7516,7 +7516,7 @@ define float @test_mm512_mask_reduce_add_ps(i16 zeroext %__M, <16 x float> %__W)
 ; X86-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; X86-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
 ; X86-NEXT:    vaddps %xmm1, %xmm0, %xmm0
-; X86-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,0,3,2]
+; X86-NEXT:    vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
 ; X86-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; X86-NEXT:    vmovss %xmm0, (%esp)
 ; X86-NEXT:    flds (%esp)
@@ -7535,7 +7535,7 @@ define float @test_mm512_mask_reduce_add_ps(i16 zeroext %__M, <16 x float> %__W)
 ; X64-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; X64-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
 ; X64-NEXT:    vaddps %xmm1, %xmm0, %xmm0
-; X64-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,0,3,2]
+; X64-NEXT:    vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
 ; X64-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; X64-NEXT:    vzeroupper
 ; X64-NEXT:    retq
@@ -7573,7 +7573,7 @@ define float @test_mm512_mask_reduce_mul_ps(i16 zeroext %__M, <16 x float> %__W)
 ; X86-NEXT:    vmulps %xmm1, %xmm0, %xmm0
 ; X86-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
 ; X86-NEXT:    vmulps %xmm1, %xmm0, %xmm0
-; X86-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,0,3,2]
+; X86-NEXT:    vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
 ; X86-NEXT:    vmulps %xmm1, %xmm0, %xmm0
 ; X86-NEXT:    vmovss %xmm0, (%esp)
 ; X86-NEXT:    flds (%esp)
@@ -7593,7 +7593,7 @@ define float @test_mm512_mask_reduce_mul_ps(i16 zeroext %__M, <16 x float> %__W)
 ; X64-NEXT:    vmulps %xmm1, %xmm0, %xmm0
 ; X64-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
 ; X64-NEXT:    vmulps %xmm1, %xmm0, %xmm0
-; X64-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,0,3,2]
+; X64-NEXT:    vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
 ; X64-NEXT:    vmulps %xmm1, %xmm0, %xmm0
 ; X64-NEXT:    vzeroupper
 ; X64-NEXT:    retq

diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
index 2eb9362..01e3668 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
@@ -2703,7 +2703,7 @@ define <4 x float> @PR22377(<4 x float> %a, <4 x float> %b) {
 ; SSE-LABEL: PR22377:
 ; SSE:       # %bb.0: # %entry
 ; SSE-NEXT:    movaps %xmm0, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,3],xmm0[1,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,3],xmm0[2,3]
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2,0,2]
 ; SSE-NEXT:    addps %xmm0, %xmm1
 ; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
@@ -2711,7 +2711,7 @@ define <4 x float> @PR22377(<4 x float> %a, <4 x float> %b) {
 ;
 ; AVX-LABEL: PR22377:
 ; AVX:       # %bb.0: # %entry
-; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,3,1,3]
+; AVX-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,3,2,3]
 ; AVX-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,0,2]
 ; AVX-NEXT:    vaddps %xmm0, %xmm1, %xmm1
 ; AVX-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]