From d87d4991ce23d1e6e300b054fbea3bedb6cf8f4b Mon Sep 17 00:00:00 2001 From: Simon Pilgrim Date: Thu, 26 Nov 2015 20:53:28 +0000 Subject: [PATCH] [X86][FMA] Begun adding AVX512 FMA tests As discussed on D14909 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@254180 91177308-0d34-0410-b5e6-96231b3b80d8 --- test/CodeGen/X86/fma_patterns.ll | 991 ++++++++++++++++---------- test/CodeGen/X86/fma_patterns_wide.ll | 157 ++-- 2 files changed, 706 insertions(+), 442 deletions(-) diff --git a/test/CodeGen/X86/fma_patterns.ll b/test/CodeGen/X86/fma_patterns.ll index 9c8b12d63fa..f3695408d0e 100644 --- a/test/CodeGen/X86/fma_patterns.ll +++ b/test/CodeGen/X86/fma_patterns.ll @@ -1,66 +1,88 @@ -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+fma -fp-contract=fast | FileCheck %s --check-prefix=ALL --check-prefix=CHECK_FMA -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+fma4,+fma -fp-contract=fast | FileCheck %s --check-prefix=ALL --check-prefix=CHECK_FMA -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+fma4 -fp-contract=fast | FileCheck %s --check-prefix=ALL --check-prefix=CHECK_FMA4 +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+fma -fp-contract=fast | FileCheck %s --check-prefix=ALL --check-prefix=FMA +; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+fma4,+fma -fp-contract=fast | FileCheck %s --check-prefix=ALL --check-prefix=FMA +; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+fma4 -fp-contract=fast | FileCheck %s --check-prefix=ALL --check-prefix=FMA4 +; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512vl -fp-contract=fast | FileCheck %s --check-prefix=ALL --check-prefix=AVX512 ; ; Patterns (+ fneg variants): add(mul(x,y),z), sub(mul(x,y),z) ; define <4 x float> @test_x86_fmadd_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) { -; CHECK_FMA-LABEL: test_x86_fmadd_ps: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmadd213ps %xmm2, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fmadd_ps: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmaddps %xmm2, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fmadd_ps: +; FMA: # BB#0: +; FMA-NEXT: vfmadd213ps %xmm2, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fmadd_ps: +; FMA4: # BB#0: +; FMA4-NEXT: vfmaddps %xmm2, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fmadd_ps: +; AVX512: # BB#0: +; AVX512-NEXT: vfmadd213ps %xmm2, %xmm1, %xmm0 +; AVX512-NEXT: retq %x = fmul <4 x float> %a0, %a1 %res = fadd <4 x float> %x, %a2 ret <4 x float> %res } define <4 x float> @test_x86_fmsub_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) { -; CHECK_FMA-LABEL: test_x86_fmsub_ps: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmsub213ps %xmm2, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fmsub_ps: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmsubps %xmm2, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fmsub_ps: +; FMA: # BB#0: +; FMA-NEXT: vfmsub213ps %xmm2, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fmsub_ps: +; FMA4: # BB#0: +; FMA4-NEXT: vfmsubps %xmm2, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fmsub_ps: +; AVX512: # BB#0: +; AVX512-NEXT: vfmsub213ps %xmm2, %xmm1, %xmm0 +; AVX512-NEXT: retq %x = fmul <4 x float> %a0, %a1 %res = fsub <4 x float> %x, %a2 ret <4 x float> %res } define <4 x float> @test_x86_fnmadd_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) { -; CHECK_FMA-LABEL: test_x86_fnmadd_ps: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmadd213ps %xmm2, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fnmadd_ps: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmaddps %xmm2, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fnmadd_ps: +; FMA: # BB#0: +; FMA-NEXT: vfnmadd213ps %xmm2, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fnmadd_ps: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmaddps %xmm2, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fnmadd_ps: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmadd213ps %xmm2, %xmm1, %xmm0 +; AVX512-NEXT: retq %x = fmul <4 x float> %a0, %a1 %res = fsub <4 x float> %a2, %x ret <4 x float> %res } define <4 x float> @test_x86_fnmsub_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) { -; CHECK_FMA-LABEL: test_x86_fnmsub_ps: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmsub213ps %xmm2, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fnmsub_ps: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmsubps %xmm2, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fnmsub_ps: +; FMA: # BB#0: +; FMA-NEXT: vfnmsub213ps %xmm2, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fnmsub_ps: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmsubps %xmm2, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fnmsub_ps: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmsub213ps %xmm2, %xmm1, %xmm0 +; AVX512-NEXT: retq %x = fmul <4 x float> %a0, %a1 %y = fsub <4 x float> , %x %res = fsub <4 x float> %y, %a2 @@ -68,60 +90,80 @@ define <4 x float> @test_x86_fnmsub_ps(<4 x float> %a0, <4 x float> %a1, <4 x fl } define <8 x float> @test_x86_fmadd_ps_y(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) { -; CHECK_FMA-LABEL: test_x86_fmadd_ps_y: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmadd213ps %ymm2, %ymm1, %ymm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fmadd_ps_y: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmaddps %ymm2, %ymm1, %ymm0, %ymm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fmadd_ps_y: +; FMA: # BB#0: +; FMA-NEXT: vfmadd213ps %ymm2, %ymm1, %ymm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fmadd_ps_y: +; FMA4: # BB#0: +; FMA4-NEXT: vfmaddps %ymm2, %ymm1, %ymm0, %ymm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fmadd_ps_y: +; AVX512: # BB#0: +; AVX512-NEXT: vfmadd213ps %ymm2, %ymm1, %ymm0 +; AVX512-NEXT: retq %x = fmul <8 x float> %a0, %a1 %res = fadd <8 x float> %x, %a2 ret <8 x float> %res } define <8 x float> @test_x86_fmsub_ps_y(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) { -; CHECK_FMA-LABEL: test_x86_fmsub_ps_y: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmsub213ps %ymm2, %ymm1, %ymm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fmsub_ps_y: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmsubps %ymm2, %ymm1, %ymm0, %ymm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fmsub_ps_y: +; FMA: # BB#0: +; FMA-NEXT: vfmsub213ps %ymm2, %ymm1, %ymm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fmsub_ps_y: +; FMA4: # BB#0: +; FMA4-NEXT: vfmsubps %ymm2, %ymm1, %ymm0, %ymm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fmsub_ps_y: +; AVX512: # BB#0: +; AVX512-NEXT: vfmsub213ps %ymm2, %ymm1, %ymm0 +; AVX512-NEXT: retq %x = fmul <8 x float> %a0, %a1 %res = fsub <8 x float> %x, %a2 ret <8 x float> %res } define <8 x float> @test_x86_fnmadd_ps_y(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) { -; CHECK_FMA-LABEL: test_x86_fnmadd_ps_y: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmadd213ps %ymm2, %ymm1, %ymm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fnmadd_ps_y: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmaddps %ymm2, %ymm1, %ymm0, %ymm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fnmadd_ps_y: +; FMA: # BB#0: +; FMA-NEXT: vfnmadd213ps %ymm2, %ymm1, %ymm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fnmadd_ps_y: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmaddps %ymm2, %ymm1, %ymm0, %ymm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fnmadd_ps_y: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmadd213ps %ymm2, %ymm1, %ymm0 +; AVX512-NEXT: retq %x = fmul <8 x float> %a0, %a1 %res = fsub <8 x float> %a2, %x ret <8 x float> %res } define <8 x float> @test_x86_fnmsub_ps_y(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) { -; CHECK_FMA-LABEL: test_x86_fnmsub_ps_y: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmsub213ps %ymm2, %ymm1, %ymm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fnmsub_ps_y: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmsubps %ymm2, %ymm1, %ymm0, %ymm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fnmsub_ps_y: +; FMA: # BB#0: +; FMA-NEXT: vfnmsub213ps %ymm2, %ymm1, %ymm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fnmsub_ps_y: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmsubps %ymm2, %ymm1, %ymm0, %ymm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fnmsub_ps_y: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmsub213ps %ymm2, %ymm1, %ymm0 +; AVX512-NEXT: retq %x = fmul <8 x float> %a0, %a1 %y = fsub <8 x float> , %x %res = fsub <8 x float> %y, %a2 @@ -129,105 +171,144 @@ define <8 x float> @test_x86_fnmsub_ps_y(<8 x float> %a0, <8 x float> %a1, <8 x } define <4 x double> @test_x86_fmadd_pd_y(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) { -; CHECK_FMA-LABEL: test_x86_fmadd_pd_y: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmadd213pd %ymm2, %ymm1, %ymm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fmadd_pd_y: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmaddpd %ymm2, %ymm1, %ymm0, %ymm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fmadd_pd_y: +; FMA: # BB#0: +; FMA-NEXT: vfmadd213pd %ymm2, %ymm1, %ymm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fmadd_pd_y: +; FMA4: # BB#0: +; FMA4-NEXT: vfmaddpd %ymm2, %ymm1, %ymm0, %ymm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fmadd_pd_y: +; AVX512: # BB#0: +; AVX512-NEXT: vfmadd213pd %ymm2, %ymm1, %ymm0 +; AVX512-NEXT: retq %x = fmul <4 x double> %a0, %a1 %res = fadd <4 x double> %x, %a2 ret <4 x double> %res } define <4 x double> @test_x86_fmsub_pd_y(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) { -; CHECK_FMA-LABEL: test_x86_fmsub_pd_y: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmsub213pd %ymm2, %ymm1, %ymm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fmsub_pd_y: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmsubpd %ymm2, %ymm1, %ymm0, %ymm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fmsub_pd_y: +; FMA: # BB#0: +; FMA-NEXT: vfmsub213pd %ymm2, %ymm1, %ymm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fmsub_pd_y: +; FMA4: # BB#0: +; FMA4-NEXT: vfmsubpd %ymm2, %ymm1, %ymm0, %ymm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fmsub_pd_y: +; AVX512: # BB#0: +; AVX512-NEXT: vfmsub213pd %ymm2, %ymm1, %ymm0 +; AVX512-NEXT: retq %x = fmul <4 x double> %a0, %a1 %res = fsub <4 x double> %x, %a2 ret <4 x double> %res } define <2 x double> @test_x86_fmsub_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) { -; CHECK_FMA-LABEL: test_x86_fmsub_pd: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmsub213pd %xmm2, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fmsub_pd: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmsubpd %xmm2, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fmsub_pd: +; FMA: # BB#0: +; FMA-NEXT: vfmsub213pd %xmm2, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fmsub_pd: +; FMA4: # BB#0: +; FMA4-NEXT: vfmsubpd %xmm2, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fmsub_pd: +; AVX512: # BB#0: +; AVX512-NEXT: vfmsub213pd %xmm2, %xmm1, %xmm0 +; AVX512-NEXT: retq %x = fmul <2 x double> %a0, %a1 %res = fsub <2 x double> %x, %a2 ret <2 x double> %res } define float @test_x86_fnmadd_ss(float %a0, float %a1, float %a2) { -; CHECK_FMA-LABEL: test_x86_fnmadd_ss: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmadd213ss %xmm2, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fnmadd_ss: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmaddss %xmm2, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fnmadd_ss: +; FMA: # BB#0: +; FMA-NEXT: vfnmadd213ss %xmm2, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fnmadd_ss: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmaddss %xmm2, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fnmadd_ss: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmadd213ss %xmm2, %xmm0, %xmm1 +; AVX512-NEXT: vmovaps %zmm1, %zmm0 +; AVX512-NEXT: retq %x = fmul float %a0, %a1 %res = fsub float %a2, %x ret float %res } define double @test_x86_fnmadd_sd(double %a0, double %a1, double %a2) { -; CHECK_FMA-LABEL: test_x86_fnmadd_sd: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmadd213sd %xmm2, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fnmadd_sd: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmaddsd %xmm2, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fnmadd_sd: +; FMA: # BB#0: +; FMA-NEXT: vfnmadd213sd %xmm2, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fnmadd_sd: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmaddsd %xmm2, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fnmadd_sd: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmadd213sd %xmm2, %xmm0, %xmm1 +; AVX512-NEXT: vmovaps %zmm1, %zmm0 +; AVX512-NEXT: retq %x = fmul double %a0, %a1 %res = fsub double %a2, %x ret double %res } define double @test_x86_fmsub_sd(double %a0, double %a1, double %a2) { -; CHECK_FMA-LABEL: test_x86_fmsub_sd: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmsub213sd %xmm2, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fmsub_sd: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmsubsd %xmm2, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fmsub_sd: +; FMA: # BB#0: +; FMA-NEXT: vfmsub213sd %xmm2, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fmsub_sd: +; FMA4: # BB#0: +; FMA4-NEXT: vfmsubsd %xmm2, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fmsub_sd: +; AVX512: # BB#0: +; AVX512-NEXT: vfmsub213sd %xmm2, %xmm0, %xmm1 +; AVX512-NEXT: vmovaps %zmm1, %zmm0 +; AVX512-NEXT: retq %x = fmul double %a0, %a1 %res = fsub double %x, %a2 ret double %res } define float @test_x86_fnmsub_ss(float %a0, float %a1, float %a2) { -; CHECK_FMA-LABEL: test_x86_fnmsub_ss: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmsub213ss %xmm2, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fnmsub_ss: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmsubss %xmm2, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fnmsub_ss: +; FMA: # BB#0: +; FMA-NEXT: vfnmsub213ss %xmm2, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fnmsub_ss: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmsubss %xmm2, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fnmsub_ss: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmsub213ss %xmm2, %xmm0, %xmm1 +; AVX512-NEXT: vmovaps %zmm1, %zmm0 +; AVX512-NEXT: retq %x = fsub float -0.000000e+00, %a0 %y = fmul float %x, %a1 %res = fsub float %y, %a2 @@ -235,15 +316,22 @@ define float @test_x86_fnmsub_ss(float %a0, float %a1, float %a2) { } define <4 x float> @test_x86_fmadd_ps_load(<4 x float>* %a0, <4 x float> %a1, <4 x float> %a2) { -; CHECK_FMA-LABEL: test_x86_fmadd_ps_load: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmadd132ps (%rdi), %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fmadd_ps_load: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmaddps %xmm1, (%rdi), %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fmadd_ps_load: +; FMA: # BB#0: +; FMA-NEXT: vfmadd132ps (%rdi), %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fmadd_ps_load: +; FMA4: # BB#0: +; FMA4-NEXT: vfmaddps %xmm1, (%rdi), %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fmadd_ps_load: +; AVX512: # BB#0: +; AVX512-NEXT: vmovaps (%rdi), %xmm2 +; AVX512-NEXT: vfmadd213ps %xmm1, %xmm0, %xmm2 +; AVX512-NEXT: vmovaps %zmm2, %zmm0 +; AVX512-NEXT: retq %x = load <4 x float>, <4 x float>* %a0 %y = fmul <4 x float> %x, %a1 %res = fadd <4 x float> %y, %a2 @@ -251,15 +339,22 @@ define <4 x float> @test_x86_fmadd_ps_load(<4 x float>* %a0, <4 x float> %a1, <4 } define <4 x float> @test_x86_fmsub_ps_load(<4 x float>* %a0, <4 x float> %a1, <4 x float> %a2) { -; CHECK_FMA-LABEL: test_x86_fmsub_ps_load: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmsub132ps (%rdi), %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_x86_fmsub_ps_load: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmsubps %xmm1, (%rdi), %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_x86_fmsub_ps_load: +; FMA: # BB#0: +; FMA-NEXT: vfmsub132ps (%rdi), %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fmsub_ps_load: +; FMA4: # BB#0: +; FMA4-NEXT: vfmsubps %xmm1, (%rdi), %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fmsub_ps_load: +; AVX512: # BB#0: +; AVX512-NEXT: vmovaps (%rdi), %xmm2 +; AVX512-NEXT: vfmsub213ps %xmm1, %xmm0, %xmm2 +; AVX512-NEXT: vmovaps %zmm2, %zmm0 +; AVX512-NEXT: retq %x = load <4 x float>, <4 x float>* %a0 %y = fmul <4 x float> %x, %a1 %res = fsub <4 x float> %y, %a2 @@ -271,180 +366,240 @@ define <4 x float> @test_x86_fmsub_ps_load(<4 x float>* %a0, <4 x float> %a1, <4 ; define <4 x float> @test_v4f32_mul_add_x_one_y(<4 x float> %x, <4 x float> %y) { -; CHECK_FMA-LABEL: test_v4f32_mul_add_x_one_y: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmadd213ps %xmm1, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f32_mul_add_x_one_y: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmaddps %xmm1, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f32_mul_add_x_one_y: +; FMA: # BB#0: +; FMA-NEXT: vfmadd213ps %xmm1, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_mul_add_x_one_y: +; FMA4: # BB#0: +; FMA4-NEXT: vfmaddps %xmm1, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_mul_add_x_one_y: +; AVX512: # BB#0: +; AVX512-NEXT: vfmadd213ps %xmm1, %xmm1, %xmm0 +; AVX512-NEXT: retq %a = fadd <4 x float> %x, %m = fmul <4 x float> %a, %y ret <4 x float> %m } define <4 x float> @test_v4f32_mul_y_add_x_one(<4 x float> %x, <4 x float> %y) { -; CHECK_FMA-LABEL: test_v4f32_mul_y_add_x_one: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmadd213ps %xmm1, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f32_mul_y_add_x_one: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmaddps %xmm1, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f32_mul_y_add_x_one: +; FMA: # BB#0: +; FMA-NEXT: vfmadd213ps %xmm1, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_mul_y_add_x_one: +; FMA4: # BB#0: +; FMA4-NEXT: vfmaddps %xmm1, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_mul_y_add_x_one: +; AVX512: # BB#0: +; AVX512-NEXT: vfmadd213ps %xmm1, %xmm1, %xmm0 +; AVX512-NEXT: retq %a = fadd <4 x float> %x, %m = fmul <4 x float> %y, %a ret <4 x float> %m } define <4 x float> @test_v4f32_mul_add_x_negone_y(<4 x float> %x, <4 x float> %y) { -; CHECK_FMA-LABEL: test_v4f32_mul_add_x_negone_y: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmsub213ps %xmm1, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f32_mul_add_x_negone_y: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmsubps %xmm1, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f32_mul_add_x_negone_y: +; FMA: # BB#0: +; FMA-NEXT: vfmsub213ps %xmm1, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_mul_add_x_negone_y: +; FMA4: # BB#0: +; FMA4-NEXT: vfmsubps %xmm1, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_mul_add_x_negone_y: +; AVX512: # BB#0: +; AVX512-NEXT: vfmsub213ps %xmm1, %xmm1, %xmm0 +; AVX512-NEXT: retq %a = fadd <4 x float> %x, %m = fmul <4 x float> %a, %y ret <4 x float> %m } define <4 x float> @test_v4f32_mul_y_add_x_negone(<4 x float> %x, <4 x float> %y) { -; CHECK_FMA-LABEL: test_v4f32_mul_y_add_x_negone: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmsub213ps %xmm1, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f32_mul_y_add_x_negone: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmsubps %xmm1, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f32_mul_y_add_x_negone: +; FMA: # BB#0: +; FMA-NEXT: vfmsub213ps %xmm1, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_mul_y_add_x_negone: +; FMA4: # BB#0: +; FMA4-NEXT: vfmsubps %xmm1, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_mul_y_add_x_negone: +; AVX512: # BB#0: +; AVX512-NEXT: vfmsub213ps %xmm1, %xmm1, %xmm0 +; AVX512-NEXT: retq %a = fadd <4 x float> %x, %m = fmul <4 x float> %y, %a ret <4 x float> %m } define <4 x float> @test_v4f32_mul_sub_one_x_y(<4 x float> %x, <4 x float> %y) { -; CHECK_FMA-LABEL: test_v4f32_mul_sub_one_x_y: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmadd213ps %xmm1, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f32_mul_sub_one_x_y: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmaddps %xmm1, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f32_mul_sub_one_x_y: +; FMA: # BB#0: +; FMA-NEXT: vfnmadd213ps %xmm1, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_mul_sub_one_x_y: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmaddps %xmm1, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_mul_sub_one_x_y: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmadd213ps %xmm1, %xmm1, %xmm0 +; AVX512-NEXT: retq %s = fsub <4 x float> , %x %m = fmul <4 x float> %s, %y ret <4 x float> %m } define <4 x float> @test_v4f32_mul_y_sub_one_x(<4 x float> %x, <4 x float> %y) { -; CHECK_FMA-LABEL: test_v4f32_mul_y_sub_one_x: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmadd213ps %xmm1, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f32_mul_y_sub_one_x: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmaddps %xmm1, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f32_mul_y_sub_one_x: +; FMA: # BB#0: +; FMA-NEXT: vfnmadd213ps %xmm1, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_mul_y_sub_one_x: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmaddps %xmm1, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_mul_y_sub_one_x: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmadd213ps %xmm1, %xmm1, %xmm0 +; AVX512-NEXT: retq %s = fsub <4 x float> , %x %m = fmul <4 x float> %y, %s ret <4 x float> %m } define <4 x float> @test_v4f32_mul_sub_negone_x_y(<4 x float> %x, <4 x float> %y) { -; CHECK_FMA-LABEL: test_v4f32_mul_sub_negone_x_y: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmsub213ps %xmm1, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f32_mul_sub_negone_x_y: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmsubps %xmm1, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f32_mul_sub_negone_x_y: +; FMA: # BB#0: +; FMA-NEXT: vfnmsub213ps %xmm1, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_mul_sub_negone_x_y: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmsubps %xmm1, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_mul_sub_negone_x_y: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmsub213ps %xmm1, %xmm1, %xmm0 +; AVX512-NEXT: retq %s = fsub <4 x float> , %x %m = fmul <4 x float> %s, %y ret <4 x float> %m } define <4 x float> @test_v4f32_mul_y_sub_negone_x(<4 x float> %x, <4 x float> %y) { -; CHECK_FMA-LABEL: test_v4f32_mul_y_sub_negone_x: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmsub213ps %xmm1, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f32_mul_y_sub_negone_x: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmsubps %xmm1, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f32_mul_y_sub_negone_x: +; FMA: # BB#0: +; FMA-NEXT: vfnmsub213ps %xmm1, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_mul_y_sub_negone_x: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmsubps %xmm1, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_mul_y_sub_negone_x: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmsub213ps %xmm1, %xmm1, %xmm0 +; AVX512-NEXT: retq %s = fsub <4 x float> , %x %m = fmul <4 x float> %y, %s ret <4 x float> %m } define <4 x float> @test_v4f32_mul_sub_x_one_y(<4 x float> %x, <4 x float> %y) { -; CHECK_FMA-LABEL: test_v4f32_mul_sub_x_one_y: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmsub213ps %xmm1, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f32_mul_sub_x_one_y: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmsubps %xmm1, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f32_mul_sub_x_one_y: +; FMA: # BB#0: +; FMA-NEXT: vfmsub213ps %xmm1, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_mul_sub_x_one_y: +; FMA4: # BB#0: +; FMA4-NEXT: vfmsubps %xmm1, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_mul_sub_x_one_y: +; AVX512: # BB#0: +; AVX512-NEXT: vfmsub213ps %xmm1, %xmm1, %xmm0 +; AVX512-NEXT: retq %s = fsub <4 x float> %x, %m = fmul <4 x float> %s, %y ret <4 x float> %m } define <4 x float> @test_v4f32_mul_y_sub_x_one(<4 x float> %x, <4 x float> %y) { -; CHECK_FMA-LABEL: test_v4f32_mul_y_sub_x_one: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmsub213ps %xmm1, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f32_mul_y_sub_x_one: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmsubps %xmm1, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f32_mul_y_sub_x_one: +; FMA: # BB#0: +; FMA-NEXT: vfmsub213ps %xmm1, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_mul_y_sub_x_one: +; FMA4: # BB#0: +; FMA4-NEXT: vfmsubps %xmm1, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_mul_y_sub_x_one: +; AVX512: # BB#0: +; AVX512-NEXT: vfmsub213ps %xmm1, %xmm1, %xmm0 +; AVX512-NEXT: retq %s = fsub <4 x float> %x, %m = fmul <4 x float> %y, %s ret <4 x float> %m } define <4 x float> @test_v4f32_mul_sub_x_negone_y(<4 x float> %x, <4 x float> %y) { -; CHECK_FMA-LABEL: test_v4f32_mul_sub_x_negone_y: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmadd213ps %xmm1, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f32_mul_sub_x_negone_y: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmaddps %xmm1, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f32_mul_sub_x_negone_y: +; FMA: # BB#0: +; FMA-NEXT: vfmadd213ps %xmm1, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_mul_sub_x_negone_y: +; FMA4: # BB#0: +; FMA4-NEXT: vfmaddps %xmm1, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_mul_sub_x_negone_y: +; AVX512: # BB#0: +; AVX512-NEXT: vfmadd213ps %xmm1, %xmm1, %xmm0 +; AVX512-NEXT: retq %s = fsub <4 x float> %x, %m = fmul <4 x float> %s, %y ret <4 x float> %m } define <4 x float> @test_v4f32_mul_y_sub_x_negone(<4 x float> %x, <4 x float> %y) { -; CHECK_FMA-LABEL: test_v4f32_mul_y_sub_x_negone: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmadd213ps %xmm1, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f32_mul_y_sub_x_negone: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmaddps %xmm1, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f32_mul_y_sub_x_negone: +; FMA: # BB#0: +; FMA-NEXT: vfmadd213ps %xmm1, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_mul_y_sub_x_negone: +; FMA4: # BB#0: +; FMA4-NEXT: vfmaddps %xmm1, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_mul_y_sub_x_negone: +; AVX512: # BB#0: +; AVX512-NEXT: vfmadd213ps %xmm1, %xmm1, %xmm0 +; AVX512-NEXT: retq %s = fsub <4 x float> %x, %m = fmul <4 x float> %y, %s ret <4 x float> %m @@ -455,17 +610,24 @@ define <4 x float> @test_v4f32_mul_y_sub_x_negone(<4 x float> %x, <4 x float> %y ; define float @test_f32_interp(float %x, float %y, float %t) { -; CHECK_FMA-LABEL: test_f32_interp: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmadd213ss %xmm1, %xmm2, %xmm1 -; CHECK_FMA-NEXT: vfmadd213ss %xmm1, %xmm2, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_f32_interp: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmaddss %xmm1, %xmm1, %xmm2, %xmm1 -; CHECK_FMA4-NEXT: vfmaddss %xmm1, %xmm2, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_f32_interp: +; FMA: # BB#0: +; FMA-NEXT: vfnmadd213ss %xmm1, %xmm2, %xmm1 +; FMA-NEXT: vfmadd213ss %xmm1, %xmm2, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_f32_interp: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmaddss %xmm1, %xmm1, %xmm2, %xmm1 +; FMA4-NEXT: vfmaddss %xmm1, %xmm2, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_f32_interp: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmadd213ss %xmm1, %xmm2, %xmm1 +; AVX512-NEXT: vfmadd213ss %xmm1, %xmm0, %xmm2 +; AVX512-NEXT: vmovaps %zmm2, %zmm0 +; AVX512-NEXT: retq %t1 = fsub float 1.0, %t %tx = fmul float %x, %t %ty = fmul float %y, %t1 @@ -474,17 +636,24 @@ define float @test_f32_interp(float %x, float %y, float %t) { } define <4 x float> @test_v4f32_interp(<4 x float> %x, <4 x float> %y, <4 x float> %t) { -; CHECK_FMA-LABEL: test_v4f32_interp: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmadd213ps %xmm1, %xmm2, %xmm1 -; CHECK_FMA-NEXT: vfmadd213ps %xmm1, %xmm2, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f32_interp: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmaddps %xmm1, %xmm1, %xmm2, %xmm1 -; CHECK_FMA4-NEXT: vfmaddps %xmm1, %xmm2, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f32_interp: +; FMA: # BB#0: +; FMA-NEXT: vfnmadd213ps %xmm1, %xmm2, %xmm1 +; FMA-NEXT: vfmadd213ps %xmm1, %xmm2, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_interp: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmaddps %xmm1, %xmm1, %xmm2, %xmm1 +; FMA4-NEXT: vfmaddps %xmm1, %xmm2, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_interp: +; AVX512: # BB#0: +; AVX512-NEXT: vmovaps %zmm2, %zmm3 +; AVX512-NEXT: vfnmadd213ps %xmm1, %xmm1, %xmm3 +; AVX512-NEXT: vfmadd213ps %xmm3, %xmm2, %xmm0 +; AVX512-NEXT: retq %t1 = fsub <4 x float> , %t %tx = fmul <4 x float> %x, %t %ty = fmul <4 x float> %y, %t1 @@ -493,17 +662,24 @@ define <4 x float> @test_v4f32_interp(<4 x float> %x, <4 x float> %y, <4 x float } define <8 x float> @test_v8f32_interp(<8 x float> %x, <8 x float> %y, <8 x float> %t) { -; CHECK_FMA-LABEL: test_v8f32_interp: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmadd213ps %ymm1, %ymm2, %ymm1 -; CHECK_FMA-NEXT: vfmadd213ps %ymm1, %ymm2, %ymm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v8f32_interp: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmaddps %ymm1, %ymm1, %ymm2, %ymm1 -; CHECK_FMA4-NEXT: vfmaddps %ymm1, %ymm2, %ymm0, %ymm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v8f32_interp: +; FMA: # BB#0: +; FMA-NEXT: vfnmadd213ps %ymm1, %ymm2, %ymm1 +; FMA-NEXT: vfmadd213ps %ymm1, %ymm2, %ymm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v8f32_interp: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmaddps %ymm1, %ymm1, %ymm2, %ymm1 +; FMA4-NEXT: vfmaddps %ymm1, %ymm2, %ymm0, %ymm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v8f32_interp: +; AVX512: # BB#0: +; AVX512-NEXT: vmovaps %zmm2, %zmm3 +; AVX512-NEXT: vfnmadd213ps %ymm1, %ymm1, %ymm3 +; AVX512-NEXT: vfmadd213ps %ymm3, %ymm2, %ymm0 +; AVX512-NEXT: retq %t1 = fsub <8 x float> , %t %tx = fmul <8 x float> %x, %t %ty = fmul <8 x float> %y, %t1 @@ -512,17 +688,24 @@ define <8 x float> @test_v8f32_interp(<8 x float> %x, <8 x float> %y, <8 x float } define double @test_f64_interp(double %x, double %y, double %t) { -; CHECK_FMA-LABEL: test_f64_interp: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmadd213sd %xmm1, %xmm2, %xmm1 -; CHECK_FMA-NEXT: vfmadd213sd %xmm1, %xmm2, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_f64_interp: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmaddsd %xmm1, %xmm1, %xmm2, %xmm1 -; CHECK_FMA4-NEXT: vfmaddsd %xmm1, %xmm2, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_f64_interp: +; FMA: # BB#0: +; FMA-NEXT: vfnmadd213sd %xmm1, %xmm2, %xmm1 +; FMA-NEXT: vfmadd213sd %xmm1, %xmm2, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_f64_interp: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmaddsd %xmm1, %xmm1, %xmm2, %xmm1 +; FMA4-NEXT: vfmaddsd %xmm1, %xmm2, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_f64_interp: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmadd213sd %xmm1, %xmm2, %xmm1 +; AVX512-NEXT: vfmadd213sd %xmm1, %xmm0, %xmm2 +; AVX512-NEXT: vmovaps %zmm2, %zmm0 +; AVX512-NEXT: retq %t1 = fsub double 1.0, %t %tx = fmul double %x, %t %ty = fmul double %y, %t1 @@ -531,17 +714,24 @@ define double @test_f64_interp(double %x, double %y, double %t) { } define <2 x double> @test_v2f64_interp(<2 x double> %x, <2 x double> %y, <2 x double> %t) { -; CHECK_FMA-LABEL: test_v2f64_interp: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmadd213pd %xmm1, %xmm2, %xmm1 -; CHECK_FMA-NEXT: vfmadd213pd %xmm1, %xmm2, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v2f64_interp: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmaddpd %xmm1, %xmm1, %xmm2, %xmm1 -; CHECK_FMA4-NEXT: vfmaddpd %xmm1, %xmm2, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v2f64_interp: +; FMA: # BB#0: +; FMA-NEXT: vfnmadd213pd %xmm1, %xmm2, %xmm1 +; FMA-NEXT: vfmadd213pd %xmm1, %xmm2, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v2f64_interp: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmaddpd %xmm1, %xmm1, %xmm2, %xmm1 +; FMA4-NEXT: vfmaddpd %xmm1, %xmm2, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v2f64_interp: +; AVX512: # BB#0: +; AVX512-NEXT: vmovaps %zmm2, %zmm3 +; AVX512-NEXT: vfnmadd213pd %xmm1, %xmm1, %xmm3 +; AVX512-NEXT: vfmadd213pd %xmm3, %xmm2, %xmm0 +; AVX512-NEXT: retq %t1 = fsub <2 x double> , %t %tx = fmul <2 x double> %x, %t %ty = fmul <2 x double> %y, %t1 @@ -550,17 +740,24 @@ define <2 x double> @test_v2f64_interp(<2 x double> %x, <2 x double> %y, <2 x do } define <4 x double> @test_v4f64_interp(<4 x double> %x, <4 x double> %y, <4 x double> %t) { -; CHECK_FMA-LABEL: test_v4f64_interp: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmadd213pd %ymm1, %ymm2, %ymm1 -; CHECK_FMA-NEXT: vfmadd213pd %ymm1, %ymm2, %ymm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f64_interp: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmaddpd %ymm1, %ymm1, %ymm2, %ymm1 -; CHECK_FMA4-NEXT: vfmaddpd %ymm1, %ymm2, %ymm0, %ymm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f64_interp: +; FMA: # BB#0: +; FMA-NEXT: vfnmadd213pd %ymm1, %ymm2, %ymm1 +; FMA-NEXT: vfmadd213pd %ymm1, %ymm2, %ymm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f64_interp: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmaddpd %ymm1, %ymm1, %ymm2, %ymm1 +; FMA4-NEXT: vfmaddpd %ymm1, %ymm2, %ymm0, %ymm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f64_interp: +; AVX512: # BB#0: +; AVX512-NEXT: vmovaps %zmm2, %zmm3 +; AVX512-NEXT: vfnmadd213pd %ymm1, %ymm1, %ymm3 +; AVX512-NEXT: vfmadd213pd %ymm3, %ymm2, %ymm0 +; AVX512-NEXT: retq %t1 = fsub <4 x double> , %t %tx = fmul <4 x double> %x, %t %ty = fmul <4 x double> %y, %t1 @@ -571,15 +768,20 @@ define <4 x double> @test_v4f64_interp(<4 x double> %x, <4 x double> %y, <4 x do ; (fneg (fma x, y, z)) -> (fma x, -y, -z) define <4 x float> @test_v4f32_fneg_fmadd(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 { -; CHECK_FMA-LABEL: test_v4f32_fneg_fmadd: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmsub213ps %xmm2, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f32_fneg_fmadd: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmsubps %xmm2, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f32_fneg_fmadd: +; FMA: # BB#0: +; FMA-NEXT: vfnmsub213ps %xmm2, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_fneg_fmadd: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmsubps %xmm2, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_fneg_fmadd: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmsub213ps %xmm2, %xmm1, %xmm0 +; AVX512-NEXT: retq %mul = fmul <4 x float> %a0, %a1 %add = fadd <4 x float> %mul, %a2 %neg = fsub <4 x float> , %add @@ -587,15 +789,20 @@ define <4 x float> @test_v4f32_fneg_fmadd(<4 x float> %a0, <4 x float> %a1, <4 x } define <4 x double> @test_v4f64_fneg_fmsub(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 { -; CHECK_FMA-LABEL: test_v4f64_fneg_fmsub: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfnmadd213pd %ymm2, %ymm1, %ymm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f64_fneg_fmsub: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmaddpd %ymm2, %ymm1, %ymm0, %ymm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f64_fneg_fmsub: +; FMA: # BB#0: +; FMA-NEXT: vfnmadd213pd %ymm2, %ymm1, %ymm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f64_fneg_fmsub: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmaddpd %ymm2, %ymm1, %ymm0, %ymm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f64_fneg_fmsub: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmadd213pd %ymm2, %ymm1, %ymm0 +; AVX512-NEXT: retq %mul = fmul <4 x double> %a0, %a1 %sub = fsub <4 x double> %mul, %a2 %neg = fsub <4 x double> , %sub @@ -603,15 +810,20 @@ define <4 x double> @test_v4f64_fneg_fmsub(<4 x double> %a0, <4 x double> %a1, < } define <4 x float> @test_v4f32_fneg_fnmadd(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 { -; CHECK_FMA-LABEL: test_v4f32_fneg_fnmadd: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmsub213ps %xmm2, %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f32_fneg_fnmadd: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmsubps %xmm2, %xmm1, %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f32_fneg_fnmadd: +; FMA: # BB#0: +; FMA-NEXT: vfmsub213ps %xmm2, %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_fneg_fnmadd: +; FMA4: # BB#0: +; FMA4-NEXT: vfmsubps %xmm2, %xmm1, %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_fneg_fnmadd: +; AVX512: # BB#0: +; AVX512-NEXT: vfmsub213ps %xmm2, %xmm1, %xmm0 +; AVX512-NEXT: retq %mul = fmul <4 x float> %a0, %a1 %neg0 = fsub <4 x float> , %mul %add = fadd <4 x float> %neg0, %a2 @@ -620,15 +832,20 @@ define <4 x float> @test_v4f32_fneg_fnmadd(<4 x float> %a0, <4 x float> %a1, <4 } define <4 x double> @test_v4f64_fneg_fnmsub(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 { -; CHECK_FMA-LABEL: test_v4f64_fneg_fnmsub: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmadd213pd %ymm2, %ymm1, %ymm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f64_fneg_fnmsub: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmaddpd %ymm2, %ymm1, %ymm0, %ymm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f64_fneg_fnmsub: +; FMA: # BB#0: +; FMA-NEXT: vfmadd213pd %ymm2, %ymm1, %ymm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f64_fneg_fnmsub: +; FMA4: # BB#0: +; FMA4-NEXT: vfmaddpd %ymm2, %ymm1, %ymm0, %ymm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f64_fneg_fnmsub: +; AVX512: # BB#0: +; AVX512-NEXT: vfmadd213pd %ymm2, %ymm1, %ymm0 +; AVX512-NEXT: retq %mul = fmul <4 x double> %a0, %a1 %neg0 = fsub <4 x double> , %mul %sub = fsub <4 x double> %neg0, %a2 @@ -639,10 +856,20 @@ define <4 x double> @test_v4f64_fneg_fnmsub(<4 x double> %a0, <4 x double> %a1, ; (fma x, c1, (fmul x, c2)) -> (fmul x, c1+c2) define <4 x float> @test_v4f32_fma_x_c1_fmul_x_c2(<4 x float> %x) #0 { -; ALL-LABEL: test_v4f32_fma_x_c1_fmul_x_c2: -; ALL: # BB#0: -; ALL-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 -; ALL-NEXT: retq +; FMA-LABEL: test_v4f32_fma_x_c1_fmul_x_c2: +; FMA: # BB#0: +; FMA-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_fma_x_c1_fmul_x_c2: +; FMA4: # BB#0: +; FMA4-NEXT: vmulps {{.*}}(%rip), %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_fma_x_c1_fmul_x_c2: +; AVX512: # BB#0: +; AVX512-NEXT: vmulps {{.*}}(%rip){1to4}, %xmm0, %xmm0 +; AVX512-NEXT: retq %m0 = fmul <4 x float> %x, %m1 = fmul <4 x float> %x, %a = fadd <4 x float> %m0, %m1 @@ -652,15 +879,21 @@ define <4 x float> @test_v4f32_fma_x_c1_fmul_x_c2(<4 x float> %x) #0 { ; (fma (fmul x, c1), c2, y) -> (fma x, c1*c2, y) define <4 x float> @test_v4f32_fma_fmul_x_c1_c2_y(<4 x float> %x, <4 x float> %y) #0 { -; CHECK_FMA-LABEL: test_v4f32_fma_fmul_x_c1_c2_y: -; CHECK_FMA: # BB#0: -; CHECK_FMA-NEXT: vfmadd132ps {{.*}}(%rip), %xmm1, %xmm0 -; CHECK_FMA-NEXT: retq -; -; CHECK_FMA4-LABEL: test_v4f32_fma_fmul_x_c1_c2_y: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmaddps %xmm1, {{.*}}(%rip), %xmm0, %xmm0 -; CHECK_FMA4-NEXT: retq +; FMA-LABEL: test_v4f32_fma_fmul_x_c1_c2_y: +; FMA: # BB#0: +; FMA-NEXT: vfmadd132ps {{.*}}(%rip), %xmm1, %xmm0 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_v4f32_fma_fmul_x_c1_c2_y: +; FMA4: # BB#0: +; FMA4-NEXT: vfmaddps %xmm1, {{.*}}(%rip), %xmm0, %xmm0 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_v4f32_fma_fmul_x_c1_c2_y: +; AVX512: # BB#0: +; AVX512-NEXT: vfmadd231ps {{.*}}(%rip), %xmm0, %xmm1 +; AVX512-NEXT: vmovaps %zmm1, %zmm0 +; AVX512-NEXT: retq %m0 = fmul <4 x float> %x, %m1 = fmul <4 x float> %m0, %a = fadd <4 x float> %m1, %y diff --git a/test/CodeGen/X86/fma_patterns_wide.ll b/test/CodeGen/X86/fma_patterns_wide.ll index 5d608c91844..5541e05196e 100644 --- a/test/CodeGen/X86/fma_patterns_wide.ll +++ b/test/CodeGen/X86/fma_patterns_wide.ll @@ -1,71 +1,92 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+fma -fp-contract=fast | FileCheck %s -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+fma4,+fma -fp-contract=fast | FileCheck %s -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+fma4 -fp-contract=fast | FileCheck %s --check-prefix=CHECK_FMA4 +; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+fma -fp-contract=fast | FileCheck %s --check-prefix=FMA +; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+fma4,+fma -fp-contract=fast | FileCheck %s --check-prefix=FMA +; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+fma4 -fp-contract=fast | FileCheck %s --check-prefix=FMA4 +; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f -fp-contract=fast | FileCheck %s --check-prefix=AVX512 define <16 x float> @test_x86_fmadd_ps_y_wide(<16 x float> %a0, <16 x float> %a1, <16 x float> %a2) { -; CHECK-LABEL: test_x86_fmadd_ps_y_wide: -; CHECK: # BB#0: -; CHECK-NEXT: vfmadd213ps %ymm4, %ymm2, %ymm0 -; CHECK-NEXT: vfmadd213ps %ymm5, %ymm3, %ymm1 -; CHECK-NEXT: retq +; FMA-LABEL: test_x86_fmadd_ps_y_wide: +; FMA: # BB#0: +; FMA-NEXT: vfmadd213ps %ymm4, %ymm2, %ymm0 +; FMA-NEXT: vfmadd213ps %ymm5, %ymm3, %ymm1 +; FMA-NEXT: retq ; -; CHECK_FMA4-LABEL: test_x86_fmadd_ps_y_wide: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmaddps %ymm4, %ymm2, %ymm0, %ymm0 -; CHECK_FMA4-NEXT: vfmaddps %ymm5, %ymm3, %ymm1, %ymm1 -; CHECK_FMA4-NEXT: retq +; FMA4-LABEL: test_x86_fmadd_ps_y_wide: +; FMA4: # BB#0: +; FMA4-NEXT: vfmaddps %ymm4, %ymm2, %ymm0, %ymm0 +; FMA4-NEXT: vfmaddps %ymm5, %ymm3, %ymm1, %ymm1 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fmadd_ps_y_wide: +; AVX512: # BB#0: +; AVX512-NEXT: vfmadd213ps %zmm2, %zmm1, %zmm0 +; AVX512-NEXT: retq %x = fmul <16 x float> %a0, %a1 %res = fadd <16 x float> %x, %a2 ret <16 x float> %res } define <16 x float> @test_x86_fmsub_ps_y_wide(<16 x float> %a0, <16 x float> %a1, <16 x float> %a2) { -; CHECK-LABEL: test_x86_fmsub_ps_y_wide: -; CHECK: # BB#0: -; CHECK-NEXT: vfmsub213ps %ymm4, %ymm2, %ymm0 -; CHECK-NEXT: vfmsub213ps %ymm5, %ymm3, %ymm1 -; CHECK-NEXT: retq +; FMA-LABEL: test_x86_fmsub_ps_y_wide: +; FMA: # BB#0: +; FMA-NEXT: vfmsub213ps %ymm4, %ymm2, %ymm0 +; FMA-NEXT: vfmsub213ps %ymm5, %ymm3, %ymm1 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fmsub_ps_y_wide: +; FMA4: # BB#0: +; FMA4-NEXT: vfmsubps %ymm4, %ymm2, %ymm0, %ymm0 +; FMA4-NEXT: vfmsubps %ymm5, %ymm3, %ymm1, %ymm1 +; FMA4-NEXT: retq ; -; CHECK_FMA4-LABEL: test_x86_fmsub_ps_y_wide: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmsubps %ymm4, %ymm2, %ymm0, %ymm0 -; CHECK_FMA4-NEXT: vfmsubps %ymm5, %ymm3, %ymm1, %ymm1 -; CHECK_FMA4-NEXT: retq +; AVX512-LABEL: test_x86_fmsub_ps_y_wide: +; AVX512: # BB#0: +; AVX512-NEXT: vfmsub213ps %zmm2, %zmm1, %zmm0 +; AVX512-NEXT: retq %x = fmul <16 x float> %a0, %a1 %res = fsub <16 x float> %x, %a2 ret <16 x float> %res } define <16 x float> @test_x86_fnmadd_ps_y_wide(<16 x float> %a0, <16 x float> %a1, <16 x float> %a2) { -; CHECK-LABEL: test_x86_fnmadd_ps_y_wide: -; CHECK: # BB#0: -; CHECK-NEXT: vfnmadd213ps %ymm4, %ymm2, %ymm0 -; CHECK-NEXT: vfnmadd213ps %ymm5, %ymm3, %ymm1 -; CHECK-NEXT: retq +; FMA-LABEL: test_x86_fnmadd_ps_y_wide: +; FMA: # BB#0: +; FMA-NEXT: vfnmadd213ps %ymm4, %ymm2, %ymm0 +; FMA-NEXT: vfnmadd213ps %ymm5, %ymm3, %ymm1 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fnmadd_ps_y_wide: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmaddps %ymm4, %ymm2, %ymm0, %ymm0 +; FMA4-NEXT: vfnmaddps %ymm5, %ymm3, %ymm1, %ymm1 +; FMA4-NEXT: retq ; -; CHECK_FMA4-LABEL: test_x86_fnmadd_ps_y_wide: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmaddps %ymm4, %ymm2, %ymm0, %ymm0 -; CHECK_FMA4-NEXT: vfnmaddps %ymm5, %ymm3, %ymm1, %ymm1 -; CHECK_FMA4-NEXT: retq +; AVX512-LABEL: test_x86_fnmadd_ps_y_wide: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmadd213ps %zmm2, %zmm1, %zmm0 +; AVX512-NEXT: retq %x = fmul <16 x float> %a0, %a1 %res = fsub <16 x float> %a2, %x ret <16 x float> %res } define <16 x float> @test_x86_fnmsub_ps_y_wide(<16 x float> %a0, <16 x float> %a1, <16 x float> %a2) { -; CHECK-LABEL: test_x86_fnmsub_ps_y_wide: -; CHECK: # BB#0: -; CHECK-NEXT: vfnmsub213ps %ymm4, %ymm2, %ymm0 -; CHECK-NEXT: vfnmsub213ps %ymm5, %ymm3, %ymm1 -; CHECK-NEXT: retq +; FMA-LABEL: test_x86_fnmsub_ps_y_wide: +; FMA: # BB#0: +; FMA-NEXT: vfnmsub213ps %ymm4, %ymm2, %ymm0 +; FMA-NEXT: vfnmsub213ps %ymm5, %ymm3, %ymm1 +; FMA-NEXT: retq ; -; CHECK_FMA4-LABEL: test_x86_fnmsub_ps_y_wide: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfnmsubps %ymm4, %ymm2, %ymm0, %ymm0 -; CHECK_FMA4-NEXT: vfnmsubps %ymm5, %ymm3, %ymm1, %ymm1 -; CHECK_FMA4-NEXT: retq +; FMA4-LABEL: test_x86_fnmsub_ps_y_wide: +; FMA4: # BB#0: +; FMA4-NEXT: vfnmsubps %ymm4, %ymm2, %ymm0, %ymm0 +; FMA4-NEXT: vfnmsubps %ymm5, %ymm3, %ymm1, %ymm1 +; FMA4-NEXT: retq +; +; AVX512-LABEL: test_x86_fnmsub_ps_y_wide: +; AVX512: # BB#0: +; AVX512-NEXT: vfnmsub213ps %zmm2, %zmm1, %zmm0 +; AVX512-NEXT: retq %x = fmul <16 x float> %a0, %a1 %y = fsub <16 x float> , %x %res = fsub <16 x float> %y, %a2 @@ -73,34 +94,44 @@ define <16 x float> @test_x86_fnmsub_ps_y_wide(<16 x float> %a0, <16 x float> %a } define <8 x double> @test_x86_fmadd_pd_y_wide(<8 x double> %a0, <8 x double> %a1, <8 x double> %a2) { -; CHECK-LABEL: test_x86_fmadd_pd_y_wide: -; CHECK: # BB#0: -; CHECK-NEXT: vfmadd213pd %ymm4, %ymm2, %ymm0 -; CHECK-NEXT: vfmadd213pd %ymm5, %ymm3, %ymm1 -; CHECK-NEXT: retq +; FMA-LABEL: test_x86_fmadd_pd_y_wide: +; FMA: # BB#0: +; FMA-NEXT: vfmadd213pd %ymm4, %ymm2, %ymm0 +; FMA-NEXT: vfmadd213pd %ymm5, %ymm3, %ymm1 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fmadd_pd_y_wide: +; FMA4: # BB#0: +; FMA4-NEXT: vfmaddpd %ymm4, %ymm2, %ymm0, %ymm0 +; FMA4-NEXT: vfmaddpd %ymm5, %ymm3, %ymm1, %ymm1 +; FMA4-NEXT: retq ; -; CHECK_FMA4-LABEL: test_x86_fmadd_pd_y_wide: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmaddpd %ymm4, %ymm2, %ymm0, %ymm0 -; CHECK_FMA4-NEXT: vfmaddpd %ymm5, %ymm3, %ymm1, %ymm1 -; CHECK_FMA4-NEXT: retq +; AVX512-LABEL: test_x86_fmadd_pd_y_wide: +; AVX512: # BB#0: +; AVX512-NEXT: vfmadd213pd %zmm2, %zmm1, %zmm0 +; AVX512-NEXT: retq %x = fmul <8 x double> %a0, %a1 %res = fadd <8 x double> %x, %a2 ret <8 x double> %res } define <8 x double> @test_x86_fmsub_pd_y_wide(<8 x double> %a0, <8 x double> %a1, <8 x double> %a2) { -; CHECK-LABEL: test_x86_fmsub_pd_y_wide: -; CHECK: # BB#0: -; CHECK-NEXT: vfmsub213pd %ymm4, %ymm2, %ymm0 -; CHECK-NEXT: vfmsub213pd %ymm5, %ymm3, %ymm1 -; CHECK-NEXT: retq +; FMA-LABEL: test_x86_fmsub_pd_y_wide: +; FMA: # BB#0: +; FMA-NEXT: vfmsub213pd %ymm4, %ymm2, %ymm0 +; FMA-NEXT: vfmsub213pd %ymm5, %ymm3, %ymm1 +; FMA-NEXT: retq +; +; FMA4-LABEL: test_x86_fmsub_pd_y_wide: +; FMA4: # BB#0: +; FMA4-NEXT: vfmsubpd %ymm4, %ymm2, %ymm0, %ymm0 +; FMA4-NEXT: vfmsubpd %ymm5, %ymm3, %ymm1, %ymm1 +; FMA4-NEXT: retq ; -; CHECK_FMA4-LABEL: test_x86_fmsub_pd_y_wide: -; CHECK_FMA4: # BB#0: -; CHECK_FMA4-NEXT: vfmsubpd %ymm4, %ymm2, %ymm0, %ymm0 -; CHECK_FMA4-NEXT: vfmsubpd %ymm5, %ymm3, %ymm1, %ymm1 -; CHECK_FMA4-NEXT: retq +; AVX512-LABEL: test_x86_fmsub_pd_y_wide: +; AVX512: # BB#0: +; AVX512-NEXT: vfmsub213pd %zmm2, %zmm1, %zmm0 +; AVX512-NEXT: retq %x = fmul <8 x double> %a0, %a1 %res = fsub <8 x double> %x, %a2 ret <8 x double> %res -- 2.34.1