Commit 42fc7852 authored by Craig Topper's avatar Craig Topper
Browse files

[X86] Print k-mask in FMA3 comments.

parent 95192f54

llvm/lib/Target/X86/MCTargetDesc/X86InstComments.cpp

+6 −4
Original line number Diff line number Diff line
@@ -281,7 +281,8 @@ static void printMasking(raw_ostream &OS, const MCInst *MI, 
    OS << " {z}";

}



static bool printFMAComments(const MCInst *MI, raw_ostream &OS) {

static bool printFMAComments(const MCInst *MI, raw_ostream &OS,

                             const MCInstrInfo &MCII) {

  const char *Mul1Name = nullptr, *Mul2Name = nullptr, *AccName = nullptr;

  unsigned NumOperands = MI->getNumOperands();

  bool RegForm = false;
@@ -613,8 +614,9 @@ static bool printFMAComments(const MCInst *MI, raw_ostream &OS) { 
  if (!Mul2Name) Mul2Name = "mem";

  if (!AccName)  AccName = "mem";



  OS << DestName << " = ";

  // TODO: Print masking information?

  OS << DestName;

  printMasking(OS, MI, MCII);

  OS << " = ";



  if (Negate)

    OS << '-';
@@ -641,7 +643,7 @@ bool llvm::EmitAnyX86InstComments(const MCInst *MI, raw_ostream &OS, 
  unsigned NumOperands = MI->getNumOperands();

  bool RegForm = false;



  if (printFMAComments(MI, OS))

  if (printFMAComments(MI, OS, MCII))

    return true;



  switch (MI->getOpcode()) {

llvm/test/CodeGen/X86/avx512-fma-intrinsics-upgrade.ll

+31 −31
Original line number Diff line number Diff line
@@ -18,7 +18,7 @@ define <16 x float> @test_mask_vfnmadd_ps(<16 x float> %a0, <16 x float> %a1, <1 
; CHECK-LABEL: test_mask_vfnmadd_ps:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfnmadd132ps {{.*#+}} zmm0 = -(zmm0 * zmm1) + zmm2

; CHECK-NEXT:    vfnmadd132ps {{.*#+}} zmm0 {%k1} = -(zmm0 * zmm1) + zmm2

; CHECK-NEXT:    retq

  %res = call <16 x float> @llvm.x86.avx512.mask.vfnmadd.ps.512(<16 x float> %a0, <16 x float> %a1, <16 x float> %a2, i16 %mask, i32 4) nounwind

  ret <16 x float> %res
@@ -38,7 +38,7 @@ define <8 x double> @test_mask_vfnmadd_pd(<8 x double> %a0, <8 x double> %a1, <8 
; CHECK-LABEL: test_mask_vfnmadd_pd:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfnmadd132pd {{.*#+}} zmm0 = -(zmm0 * zmm1) + zmm2

; CHECK-NEXT:    vfnmadd132pd {{.*#+}} zmm0 {%k1} = -(zmm0 * zmm1) + zmm2

; CHECK-NEXT:    retq

  %res = call <8 x double> @llvm.x86.avx512.mask.vfnmadd.pd.512(<8 x double> %a0, <8 x double> %a1, <8 x double> %a2, i8 %mask, i32 4) nounwind

  ret <8 x double> %res
@@ -58,7 +58,7 @@ define <16 x float> @test_mask_vfnmsub_ps(<16 x float> %a0, <16 x float> %a1, <1 
; CHECK-LABEL: test_mask_vfnmsub_ps:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfnmsub132ps {{.*#+}} zmm0 = -(zmm0 * zmm1) - zmm2

; CHECK-NEXT:    vfnmsub132ps {{.*#+}} zmm0 {%k1} = -(zmm0 * zmm1) - zmm2

; CHECK-NEXT:    retq

  %res = call <16 x float> @llvm.x86.avx512.mask.vfnmsub.ps.512(<16 x float> %a0, <16 x float> %a1, <16 x float> %a2, i16 %mask, i32 4) nounwind

  ret <16 x float> %res
@@ -78,7 +78,7 @@ define <8 x double> @test_mask_vfnmsub_pd(<8 x double> %a0, <8 x double> %a1, <8 
; CHECK-LABEL: test_mask_vfnmsub_pd:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfnmsub132pd {{.*#+}} zmm0 = -(zmm0 * zmm1) - zmm2

; CHECK-NEXT:    vfnmsub132pd {{.*#+}} zmm0 {%k1} = -(zmm0 * zmm1) - zmm2

; CHECK-NEXT:    retq

  %res = call <8 x double> @llvm.x86.avx512.mask.vfnmsub.pd.512(<8 x double> %a0, <8 x double> %a1, <8 x double> %a2, i8 %mask, i32 4) nounwind

  ret <8 x double> %res
@@ -97,7 +97,7 @@ define <16 x float> @test_mask_fmaddsub_ps(<16 x float> %a, <16 x float> %b, <16 
; CHECK-LABEL: test_mask_fmaddsub_ps:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmaddsub132ps {{.*#+}} zmm0 = (zmm0 * zmm1) +/- zmm2

; CHECK-NEXT:    vfmaddsub132ps {{.*#+}} zmm0 {%k1} = (zmm0 * zmm1) +/- zmm2

; CHECK-NEXT:    retq

  %res = call <16 x float> @llvm.x86.avx512.mask.vfmaddsub.ps.512(<16 x float> %a, <16 x float> %b, <16 x float> %c, i16 %mask, i32 4)

  ret <16 x float> %res
@@ -119,7 +119,7 @@ define <8 x double> @test_mask_vfmaddsub_pd(<8 x double> %a0, <8 x double> %a1, 
; CHECK-LABEL: test_mask_vfmaddsub_pd:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmaddsub132pd {{.*#+}} zmm0 = (zmm0 * zmm1) +/- zmm2

; CHECK-NEXT:    vfmaddsub132pd {{.*#+}} zmm0 {%k1} = (zmm0 * zmm1) +/- zmm2

; CHECK-NEXT:    retq

  %res = call <8 x double> @llvm.x86.avx512.mask.vfmaddsub.pd.512(<8 x double> %a0, <8 x double> %a1, <8 x double> %a2, i8 %mask, i32 4) nounwind

  ret <8 x double> %res
@@ -129,7 +129,7 @@ define <8 x double>@test_int_x86_avx512_mask_vfmaddsub_pd_512(<8 x double> %x0, 
; CHECK-LABEL: test_int_x86_avx512_mask_vfmaddsub_pd_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmaddsub132pd {{.*#+}} zmm0 = (zmm0 * zmm1) +/- zmm2

; CHECK-NEXT:    vfmaddsub132pd {{.*#+}} zmm0 {%k1} = (zmm0 * zmm1) +/- zmm2

; CHECK-NEXT:    retq

  %res = call <8 x double> @llvm.x86.avx512.mask.vfmaddsub.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x double> %x2, i8 %x3, i32 4)

  ret <8 x double> %res
@@ -141,7 +141,7 @@ define <8 x double>@test_int_x86_avx512_mask3_vfmaddsub_pd_512(<8 x double> %x0, 
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmaddsub_pd_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmaddsub231pd {{.*#+}} zmm2 = (zmm0 * zmm1) +/- zmm2

; CHECK-NEXT:    vfmaddsub231pd {{.*#+}} zmm2 {%k1} = (zmm0 * zmm1) +/- zmm2

; CHECK-NEXT:    vmovapd %zmm2, %zmm0

; CHECK-NEXT:    retq

  %res = call <8 x double> @llvm.x86.avx512.mask3.vfmaddsub.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x double> %x2, i8 %x3, i32 4)
@@ -154,7 +154,7 @@ define <8 x double>@test_int_x86_avx512_maskz_vfmaddsub_pd_512(<8 x double> %x0, 
; CHECK-LABEL: test_int_x86_avx512_maskz_vfmaddsub_pd_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmaddsub213pd {{.*#+}} zmm0 = (zmm1 * zmm0) +/- zmm2

; CHECK-NEXT:    vfmaddsub213pd {{.*#+}} zmm0 {%k1} {z} = (zmm1 * zmm0) +/- zmm2

; CHECK-NEXT:    retq

  %res = call <8 x double> @llvm.x86.avx512.maskz.vfmaddsub.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x double> %x2, i8 %x3, i32 4)

  ret <8 x double> %res
@@ -164,7 +164,7 @@ define <16 x float>@test_int_x86_avx512_mask_vfmaddsub_ps_512(<16 x float> %x0, 
; CHECK-LABEL: test_int_x86_avx512_mask_vfmaddsub_ps_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmaddsub132ps {{.*#+}} zmm0 = (zmm0 * zmm1) +/- zmm2

; CHECK-NEXT:    vfmaddsub132ps {{.*#+}} zmm0 {%k1} = (zmm0 * zmm1) +/- zmm2

; CHECK-NEXT:    retq

  %res = call <16 x float> @llvm.x86.avx512.mask.vfmaddsub.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x float> %x2, i16 %x3, i32 4)

  ret <16 x float> %res
@@ -176,7 +176,7 @@ define <16 x float>@test_int_x86_avx512_mask3_vfmaddsub_ps_512(<16 x float> %x0, 
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmaddsub_ps_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmaddsub231ps {{.*#+}} zmm2 = (zmm0 * zmm1) +/- zmm2

; CHECK-NEXT:    vfmaddsub231ps {{.*#+}} zmm2 {%k1} = (zmm0 * zmm1) +/- zmm2

; CHECK-NEXT:    vmovaps %zmm2, %zmm0

; CHECK-NEXT:    retq

  %res = call <16 x float> @llvm.x86.avx512.mask3.vfmaddsub.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x float> %x2, i16 %x3, i32 4)
@@ -189,7 +189,7 @@ define <16 x float>@test_int_x86_avx512_maskz_vfmaddsub_ps_512(<16 x float> %x0, 
; CHECK-LABEL: test_int_x86_avx512_maskz_vfmaddsub_ps_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmaddsub213ps {{.*#+}} zmm0 = (zmm1 * zmm0) +/- zmm2

; CHECK-NEXT:    vfmaddsub213ps {{.*#+}} zmm0 {%k1} {z} = (zmm1 * zmm0) +/- zmm2

; CHECK-NEXT:    retq

  %res = call <16 x float> @llvm.x86.avx512.maskz.vfmaddsub.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x float> %x2, i16 %x3, i32 4)

  ret <16 x float> %res
@@ -201,7 +201,7 @@ define <8 x double>@test_int_x86_avx512_mask3_vfmsubadd_pd_512(<8 x double> %x0, 
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsubadd_pd_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmsubadd231pd {{.*#+}} zmm2 = (zmm0 * zmm1) -/+ zmm2

; CHECK-NEXT:    vfmsubadd231pd {{.*#+}} zmm2 {%k1} = (zmm0 * zmm1) -/+ zmm2

; CHECK-NEXT:    vmovapd %zmm2, %zmm0

; CHECK-NEXT:    retq

  %res = call <8 x double> @llvm.x86.avx512.mask3.vfmsubadd.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x double> %x2, i8 %x3, i32 4)
@@ -214,7 +214,7 @@ define <16 x float>@test_int_x86_avx512_mask3_vfmsubadd_ps_512(<16 x float> %x0, 
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsubadd_ps_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmsubadd231ps {{.*#+}} zmm2 = (zmm0 * zmm1) -/+ zmm2

; CHECK-NEXT:    vfmsubadd231ps {{.*#+}} zmm2 {%k1} = (zmm0 * zmm1) -/+ zmm2

; CHECK-NEXT:    vmovaps %zmm2, %zmm0

; CHECK-NEXT:    retq

  %res = call <16 x float> @llvm.x86.avx512.mask3.vfmsubadd.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x float> %x2, i16 %x3, i32 4)
@@ -265,7 +265,7 @@ define <16 x float> @test_mask_round_vfmadd512_ps_rrb_current(<16 x float> %a0, 
; CHECK-LABEL: test_mask_round_vfmadd512_ps_rrb_current:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmadd132ps {{.*#+}} zmm0 = (zmm0 * zmm1) + zmm2

; CHECK-NEXT:    vfmadd132ps {{.*#+}} zmm0 {%k1} = (zmm0 * zmm1) + zmm2

; CHECK-NEXT:    retq

  %res = call <16 x float> @llvm.x86.avx512.mask.vfmadd.ps.512(<16 x float> %a0, <16 x float> %a1, <16 x float> %a2, i16 %mask, i32 4) nounwind

  ret <16 x float> %res
@@ -322,7 +322,7 @@ define <8 x double>@test_int_x86_avx512_mask3_vfmsub_pd_512(<8 x double> %x0, <8 
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsub_pd_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmsub231pd {{.*#+}} zmm2 = (zmm0 * zmm1) - zmm2

; CHECK-NEXT:    vfmsub231pd {{.*#+}} zmm2 {%k1} = (zmm0 * zmm1) - zmm2

; CHECK-NEXT:    vmovapd %zmm2, %zmm0

; CHECK-NEXT:    retq

  %res = call <8 x double> @llvm.x86.avx512.mask3.vfmsub.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x double> %x2, i8 %x3, i32 4)
@@ -335,7 +335,7 @@ define <16 x float>@test_int_x86_avx512_mask3_vfmsub_ps_512(<16 x float> %x0, <1 
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsub_ps_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmsub231ps {{.*#+}} zmm2 = (zmm0 * zmm1) - zmm2

; CHECK-NEXT:    vfmsub231ps {{.*#+}} zmm2 {%k1} = (zmm0 * zmm1) - zmm2

; CHECK-NEXT:    vmovaps %zmm2, %zmm0

; CHECK-NEXT:    retq

  %res = call <16 x float> @llvm.x86.avx512.mask3.vfmsub.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x float> %x2, i16 %x3, i32 4)
@@ -386,7 +386,7 @@ define <8 x double> @test_mask_round_vfmadd512_pd_rrb_current(<8 x double> %a0, 
; CHECK-LABEL: test_mask_round_vfmadd512_pd_rrb_current:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmadd132pd {{.*#+}} zmm0 = (zmm0 * zmm1) + zmm2

; CHECK-NEXT:    vfmadd132pd {{.*#+}} zmm0 {%k1} = (zmm0 * zmm1) + zmm2

; CHECK-NEXT:    retq

  %res = call <8 x double> @llvm.x86.avx512.mask.vfmadd.pd.512(<8 x double> %a0, <8 x double> %a1, <8 x double> %a2, i8 %mask, i32 4) nounwind

  ret <8 x double> %res
@@ -441,7 +441,7 @@ define <8 x double>@test_int_x86_avx512_mask_vfmadd_pd_512(<8 x double> %x0, <8 
; CHECK-LABEL: test_int_x86_avx512_mask_vfmadd_pd_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmadd132pd {{.*#+}} zmm0 = (zmm0 * zmm1) + zmm2

; CHECK-NEXT:    vfmadd132pd {{.*#+}} zmm0 {%k1} = (zmm0 * zmm1) + zmm2

; CHECK-NEXT:    retq

  %res = call <8 x double> @llvm.x86.avx512.mask.vfmadd.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x double> %x2, i8 %x3, i32 4)

  ret <8 x double> %res
@@ -453,7 +453,7 @@ define <8 x double>@test_int_x86_avx512_mask3_vfmadd_pd_512(<8 x double> %x0, <8 
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmadd_pd_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmadd231pd {{.*#+}} zmm2 = (zmm0 * zmm1) + zmm2

; CHECK-NEXT:    vfmadd231pd {{.*#+}} zmm2 {%k1} = (zmm0 * zmm1) + zmm2

; CHECK-NEXT:    vmovapd %zmm2, %zmm0

; CHECK-NEXT:    retq

  %res = call <8 x double> @llvm.x86.avx512.mask3.vfmadd.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x double> %x2, i8 %x3, i32 4)
@@ -466,7 +466,7 @@ define <8 x double>@test_int_x86_avx512_maskz_vfmadd_pd_512(<8 x double> %x0, <8 
; CHECK-LABEL: test_int_x86_avx512_maskz_vfmadd_pd_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmadd213pd {{.*#+}} zmm0 = (zmm1 * zmm0) + zmm2

; CHECK-NEXT:    vfmadd213pd {{.*#+}} zmm0 {%k1} {z} = (zmm1 * zmm0) + zmm2

; CHECK-NEXT:    retq

  %res = call <8 x double> @llvm.x86.avx512.maskz.vfmadd.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x double> %x2, i8 %x3, i32 4)

  ret <8 x double> %res
@@ -476,7 +476,7 @@ define <16 x float>@test_int_x86_avx512_mask_vfmadd_ps_512(<16 x float> %x0, <16 
; CHECK-LABEL: test_int_x86_avx512_mask_vfmadd_ps_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmadd132ps {{.*#+}} zmm0 = (zmm0 * zmm1) + zmm2

; CHECK-NEXT:    vfmadd132ps {{.*#+}} zmm0 {%k1} = (zmm0 * zmm1) + zmm2

; CHECK-NEXT:    retq

  %res = call <16 x float> @llvm.x86.avx512.mask.vfmadd.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x float> %x2, i16 %x3, i32 4)

  ret <16 x float> %res
@@ -488,7 +488,7 @@ define <16 x float>@test_int_x86_avx512_mask3_vfmadd_ps_512(<16 x float> %x0, <1 
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmadd_ps_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmadd231ps {{.*#+}} zmm2 = (zmm0 * zmm1) + zmm2

; CHECK-NEXT:    vfmadd231ps {{.*#+}} zmm2 {%k1} = (zmm0 * zmm1) + zmm2

; CHECK-NEXT:    vmovaps %zmm2, %zmm0

; CHECK-NEXT:    retq

  %res = call <16 x float> @llvm.x86.avx512.mask3.vfmadd.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x float> %x2, i16 %x3, i32 4)
@@ -501,7 +501,7 @@ define <16 x float>@test_int_x86_avx512_maskz_vfmadd_ps_512(<16 x float> %x0, <1 
; CHECK-LABEL: test_int_x86_avx512_maskz_vfmadd_ps_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfmadd213ps {{.*#+}} zmm0 = (zmm1 * zmm0) + zmm2

; CHECK-NEXT:    vfmadd213ps {{.*#+}} zmm0 {%k1} {z} = (zmm1 * zmm0) + zmm2

; CHECK-NEXT:    retq

  %res = call <16 x float> @llvm.x86.avx512.maskz.vfmadd.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x float> %x2, i16 %x3, i32 4)

  ret <16 x float> %res
@@ -552,7 +552,7 @@ define <8 x double> @test_mask_round_vfnmsub512_pd_rrb_current(<8 x double> %a0, 
; CHECK-LABEL: test_mask_round_vfnmsub512_pd_rrb_current:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfnmsub132pd {{.*#+}} zmm0 = -(zmm0 * zmm1) - zmm2

; CHECK-NEXT:    vfnmsub132pd {{.*#+}} zmm0 {%k1} = -(zmm0 * zmm1) - zmm2

; CHECK-NEXT:    retq

  %res = call <8 x double> @llvm.x86.avx512.mask.vfnmsub.pd.512(<8 x double> %a0, <8 x double> %a1, <8 x double> %a2, i8 %mask, i32 4) nounwind

  ret <8 x double> %res
@@ -607,7 +607,7 @@ define <8 x double>@test_int_x86_avx512_mask_vfnmsub_pd_512(<8 x double> %x0, <8 
; CHECK-LABEL: test_int_x86_avx512_mask_vfnmsub_pd_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfnmsub132pd {{.*#+}} zmm0 = -(zmm0 * zmm1) - zmm2

; CHECK-NEXT:    vfnmsub132pd {{.*#+}} zmm0 {%k1} = -(zmm0 * zmm1) - zmm2

; CHECK-NEXT:    retq

  %res = call <8 x double> @llvm.x86.avx512.mask.vfnmsub.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x double> %x2, i8 %x3, i32 4)

  ret <8 x double> %res
@@ -619,7 +619,7 @@ define <8 x double>@test_int_x86_avx512_mask3_vfnmsub_pd_512(<8 x double> %x0, < 
; CHECK-LABEL: test_int_x86_avx512_mask3_vfnmsub_pd_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfnmsub231pd {{.*#+}} zmm2 = -(zmm0 * zmm1) - zmm2

; CHECK-NEXT:    vfnmsub231pd {{.*#+}} zmm2 {%k1} = -(zmm0 * zmm1) - zmm2

; CHECK-NEXT:    vmovapd %zmm2, %zmm0

; CHECK-NEXT:    retq

  %res = call <8 x double> @llvm.x86.avx512.mask3.vfnmsub.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x double> %x2, i8 %x3, i32 4)
@@ -630,7 +630,7 @@ define <16 x float>@test_int_x86_avx512_mask_vfnmsub_ps_512(<16 x float> %x0, <1 
; CHECK-LABEL: test_int_x86_avx512_mask_vfnmsub_ps_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfnmsub132ps {{.*#+}} zmm0 = -(zmm0 * zmm1) - zmm2

; CHECK-NEXT:    vfnmsub132ps {{.*#+}} zmm0 {%k1} = -(zmm0 * zmm1) - zmm2

; CHECK-NEXT:    retq

  %res = call <16 x float> @llvm.x86.avx512.mask.vfnmsub.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x float> %x2, i16 %x3, i32 4)

  ret <16 x float> %res
@@ -642,7 +642,7 @@ define <16 x float>@test_int_x86_avx512_mask3_vfnmsub_ps_512(<16 x float> %x0, < 
; CHECK-LABEL: test_int_x86_avx512_mask3_vfnmsub_ps_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfnmsub231ps {{.*#+}} zmm2 = -(zmm0 * zmm1) - zmm2

; CHECK-NEXT:    vfnmsub231ps {{.*#+}} zmm2 {%k1} = -(zmm0 * zmm1) - zmm2

; CHECK-NEXT:    vmovaps %zmm2, %zmm0

; CHECK-NEXT:    retq

  %res = call <16 x float> @llvm.x86.avx512.mask3.vfnmsub.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x float> %x2, i16 %x3, i32 4)
@@ -653,7 +653,7 @@ define <8 x double>@test_int_x86_avx512_mask_vfnmadd_pd_512(<8 x double> %x0, <8 
; CHECK-LABEL: test_int_x86_avx512_mask_vfnmadd_pd_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfnmadd132pd {{.*#+}} zmm0 = -(zmm0 * zmm1) + zmm2

; CHECK-NEXT:    vfnmadd132pd {{.*#+}} zmm0 {%k1} = -(zmm0 * zmm1) + zmm2

; CHECK-NEXT:    retq

  %res = call <8 x double> @llvm.x86.avx512.mask.vfnmadd.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x double> %x2, i8 %x3, i32 4)

  ret <8 x double> %res
@@ -663,7 +663,7 @@ define <16 x float>@test_int_x86_avx512_mask_vfnmadd_ps_512(<16 x float> %x0, <1 
; CHECK-LABEL: test_int_x86_avx512_mask_vfnmadd_ps_512:

; CHECK:       ## %bb.0:

; CHECK-NEXT:    kmovw %edi, %k1

; CHECK-NEXT:    vfnmadd132ps {{.*#+}} zmm0 = -(zmm0 * zmm1) + zmm2

; CHECK-NEXT:    vfnmadd132ps {{.*#+}} zmm0 {%k1} = -(zmm0 * zmm1) + zmm2

; CHECK-NEXT:    retq

  %res = call <16 x float> @llvm.x86.avx512.mask.vfnmadd.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x float> %x2, i16 %x3, i32 4)

  ret <16 x float> %res

llvm/test/CodeGen/X86/avx512-fma-intrinsics.ll

+62 −62

File changed.

Preview size limit exceeded, changes collapsed.

llvm/test/CodeGen/X86/avx512-fma.ll

+6 −6
Original line number Diff line number Diff line
@@ -125,14 +125,14 @@ define <16 x float> @test_x86_fmadd132_ps(<16 x float> %a0, <16 x float> %a1, <1 
; KNL-NEXT:    vpmovsxbd %xmm2, %zmm2

; KNL-NEXT:    vpslld $31, %zmm2, %zmm2

; KNL-NEXT:    vptestmd %zmm2, %zmm2, %k1

; KNL-NEXT:    vfmadd132ps {{.*#+}} zmm0 = (zmm0 * mem) + zmm1

; KNL-NEXT:    vfmadd132ps {{.*#+}} zmm0 {%k1} = (zmm0 * mem) + zmm1

; KNL-NEXT:    retq

;

; SKX-LABEL: test_x86_fmadd132_ps:

; SKX:       ## %bb.0:

; SKX-NEXT:    vpsllw $7, %xmm2, %xmm2

; SKX-NEXT:    vpmovb2m %xmm2, %k1

; SKX-NEXT:    vfmadd132ps {{.*#+}} zmm0 = (zmm0 * mem) + zmm1

; SKX-NEXT:    vfmadd132ps {{.*#+}} zmm0 {%k1} = (zmm0 * mem) + zmm1

; SKX-NEXT:    retq

  %a2   = load <16 x float>,<16 x float> *%a2_ptrt,align 1

  %x = fmul <16 x float> %a0, %a2
@@ -148,7 +148,7 @@ define <16 x float> @test_x86_fmadd231_ps(<16 x float> %a0, <16 x float> %a1, <1 
; KNL-NEXT:    vpmovsxbd %xmm2, %zmm2

; KNL-NEXT:    vpslld $31, %zmm2, %zmm2

; KNL-NEXT:    vptestmd %zmm2, %zmm2, %k1

; KNL-NEXT:    vfmadd231ps {{.*#+}} zmm1 = (zmm0 * mem) + zmm1

; KNL-NEXT:    vfmadd231ps {{.*#+}} zmm1 {%k1} = (zmm0 * mem) + zmm1

; KNL-NEXT:    vmovaps %zmm1, %zmm0

; KNL-NEXT:    retq

;
@@ -156,7 +156,7 @@ define <16 x float> @test_x86_fmadd231_ps(<16 x float> %a0, <16 x float> %a1, <1 
; SKX:       ## %bb.0:

; SKX-NEXT:    vpsllw $7, %xmm2, %xmm2

; SKX-NEXT:    vpmovb2m %xmm2, %k1

; SKX-NEXT:    vfmadd231ps {{.*#+}} zmm1 = (zmm0 * mem) + zmm1

; SKX-NEXT:    vfmadd231ps {{.*#+}} zmm1 {%k1} = (zmm0 * mem) + zmm1

; SKX-NEXT:    vmovaps %zmm1, %zmm0

; SKX-NEXT:    retq

  %a2   = load <16 x float>,<16 x float> *%a2_ptrt,align 1
@@ -173,7 +173,7 @@ define <16 x float> @test_x86_fmadd213_ps(<16 x float> %a0, <16 x float> %a1, <1 
; KNL-NEXT:    vpmovsxbd %xmm2, %zmm2

; KNL-NEXT:    vpslld $31, %zmm2, %zmm2

; KNL-NEXT:    vptestmd %zmm2, %zmm2, %k1

; KNL-NEXT:    vfmadd213ps {{.*#+}} zmm1 = (zmm0 * zmm1) + mem

; KNL-NEXT:    vfmadd213ps {{.*#+}} zmm1 {%k1} = (zmm0 * zmm1) + mem

; KNL-NEXT:    vmovaps %zmm1, %zmm0

; KNL-NEXT:    retq

;
@@ -181,7 +181,7 @@ define <16 x float> @test_x86_fmadd213_ps(<16 x float> %a0, <16 x float> %a1, <1 
; SKX:       ## %bb.0:

; SKX-NEXT:    vpsllw $7, %xmm2, %xmm2

; SKX-NEXT:    vpmovb2m %xmm2, %k1

; SKX-NEXT:    vfmadd213ps {{.*#+}} zmm1 = (zmm0 * zmm1) + mem

; SKX-NEXT:    vfmadd213ps {{.*#+}} zmm1 {%k1} = (zmm0 * zmm1) + mem

; SKX-NEXT:    vmovaps %zmm1, %zmm0

; SKX-NEXT:    retq

  %a2   = load <16 x float>,<16 x float> *%a2_ptrt,align 1

llvm/test/CodeGen/X86/avx512-intrinsics-canonical.ll

+168 −168

File changed.

Preview size limit exceeded, changes collapsed.