Home | History | Annotate | Download | only in X86 
      1 ; RUN: llc < %s -mtriple=x86_64-pc-win32 -mcpu=core-avx2 | FileCheck %s
      2 ; RUN: llc < %s -mtriple=x86_64-pc-win32 -mattr=+fma,+fma4 | FileCheck %s
      3 ; RUN: llc < %s -mcpu=bdver2 -mtriple=x86_64-pc-win32 -mattr=-fma4 | FileCheck %s
      4 
      5 define <4 x float> @test_x86_fmadd_ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) {
      6   ; CHECK: fmadd213ss (%r8), %xmm
      7   %res = call <4 x float> @llvm.x86.fma.vfmadd.ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) nounwind
      8   ret <4 x float> %res
      9 }
     10 declare <4 x float> @llvm.x86.fma.vfmadd.ss(<4 x float>, <4 x float>, <4 x float>) nounwind readnone
     11 
     12 define <4 x float> @test_x86_fmadd_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) {
     13   ; CHECK: fmadd213ps
     14   %res = call <4 x float> @llvm.x86.fma.vfmadd.ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) nounwind
     15   ret <4 x float> %res
     16 }
     17 declare <4 x float> @llvm.x86.fma.vfmadd.ps(<4 x float>, <4 x float>, <4 x float>) nounwind readnone
     18 
     19 define <8 x float> @test_x86_fmadd_ps_y(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) {
     20   ; CHECK: fmadd213ps {{.*\(%r.*}}, %ymm
     21   %res = call <8 x float> @llvm.x86.fma.vfmadd.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) nounwind
     22   ret <8 x float> %res
     23 }
     24 declare <8 x float> @llvm.x86.fma.vfmadd.ps.256(<8 x float>, <8 x float>, <8 x float>) nounwind readnone
     25 
     26 define <4 x float> @test_x86_fnmadd_ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) {
     27   ; CHECK: fnmadd213ss (%r8), %xmm
     28   %res = call <4 x float> @llvm.x86.fma.vfnmadd.ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) nounwind
     29   ret <4 x float> %res
     30 }
     31 declare <4 x float> @llvm.x86.fma.vfnmadd.ss(<4 x float>, <4 x float>, <4 x float>) nounwind readnone
     32 
     33 define <4 x float> @test_x86_fnmadd_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) {
     34   ; CHECK: fnmadd213ps
     35   %res = call <4 x float> @llvm.x86.fma.vfnmadd.ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) nounwind
     36   ret <4 x float> %res
     37 }
     38 declare <4 x float> @llvm.x86.fma.vfnmadd.ps(<4 x float>, <4 x float>, <4 x float>) nounwind readnone
     39 
     40 define <8 x float> @test_x86_fnmadd_ps_y(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) {
     41   ; CHECK: fnmadd213ps {{.*\(%r.*}}, %ymm
     42   %res = call <8 x float> @llvm.x86.fma.vfnmadd.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) nounwind
     43   ret <8 x float> %res
     44 }
     45 declare <8 x float> @llvm.x86.fma.vfnmadd.ps.256(<8 x float>, <8 x float>, <8 x float>) nounwind readnone
     46 
     47 
     48 define <4 x float> @test_x86_fmsub_ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) {
     49   ; CHECK: fmsub213ss
     50   %res = call <4 x float> @llvm.x86.fma.vfmsub.ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) nounwind
     51   ret <4 x float> %res
     52 }
     53 declare <4 x float> @llvm.x86.fma.vfmsub.ss(<4 x float>, <4 x float>, <4 x float>) nounwind readnone
     54 
     55 define <4 x float> @test_x86_fmsub_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) {
     56   ; CHECK: fmsub213ps
     57   %res = call <4 x float> @llvm.x86.fma.vfmsub.ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) nounwind
     58   ret <4 x float> %res
     59 }
     60 declare <4 x float> @llvm.x86.fma.vfmsub.ps(<4 x float>, <4 x float>, <4 x float>) nounwind readnone
     61 
     62 define <4 x float> @test_x86_fnmsub_ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) {
     63   ; CHECK: fnmsub213ss
     64   %res = call <4 x float> @llvm.x86.fma.vfnmsub.ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) nounwind
     65   ret <4 x float> %res
     66 }
     67 declare <4 x float> @llvm.x86.fma.vfnmsub.ss(<4 x float>, <4 x float>, <4 x float>) nounwind readnone
     68 
     69 define <4 x float> @test_x86_fnmsub_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) {
     70   ; CHECK: fnmsub213ps
     71   %res = call <4 x float> @llvm.x86.fma.vfnmsub.ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) nounwind
     72   ret <4 x float> %res
     73 }
     74 declare <4 x float> @llvm.x86.fma.vfnmsub.ps(<4 x float>, <4 x float>, <4 x float>) nounwind readnone
     75 
     76 ;;;;
     77 
     78 define <2 x double> @test_x86_fmadd_sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
     79   ; CHECK: fmadd213sd
     80   %res = call <2 x double> @llvm.x86.fma.vfmadd.sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) nounwind
     81   ret <2 x double> %res
     82 }
     83 declare <2 x double> @llvm.x86.fma.vfmadd.sd(<2 x double>, <2 x double>, <2 x double>) nounwind readnone
     84 
     85 define <2 x double> @test_x86_fmadd_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
     86   ; CHECK: fmadd213pd
     87   %res = call <2 x double> @llvm.x86.fma.vfmadd.pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) nounwind
     88   ret <2 x double> %res
     89 }
     90 declare <2 x double> @llvm.x86.fma.vfmadd.pd(<2 x double>, <2 x double>, <2 x double>) nounwind readnone
     91 
     92 define <2 x double> @test_x86_fnmadd_sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
     93   ; CHECK: fnmadd213sd
     94   %res = call <2 x double> @llvm.x86.fma.vfnmadd.sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) nounwind
     95   ret <2 x double> %res
     96 }
     97 declare <2 x double> @llvm.x86.fma.vfnmadd.sd(<2 x double>, <2 x double>, <2 x double>) nounwind readnone
     98 
     99 define <2 x double> @test_x86_fnmadd_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
    100   ; CHECK: fnmadd213pd
    101   %res = call <2 x double> @llvm.x86.fma.vfnmadd.pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) nounwind
    102   ret <2 x double> %res
    103 }
    104 declare <2 x double> @llvm.x86.fma.vfnmadd.pd(<2 x double>, <2 x double>, <2 x double>) nounwind readnone
    105 
    106 
    107 
    108 define <2 x double> @test_x86_fmsub_sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
    109   ; CHECK: fmsub213sd
    110   %res = call <2 x double> @llvm.x86.fma.vfmsub.sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) nounwind
    111   ret <2 x double> %res
    112 }
    113 declare <2 x double> @llvm.x86.fma.vfmsub.sd(<2 x double>, <2 x double>, <2 x double>) nounwind readnone
    114 
    115 define <2 x double> @test_x86_fmsub_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
    116   ; CHECK: fmsub213pd
    117   %res = call <2 x double> @llvm.x86.fma.vfmsub.pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) nounwind
    118   ret <2 x double> %res
    119 }
    120 declare <2 x double> @llvm.x86.fma.vfmsub.pd(<2 x double>, <2 x double>, <2 x double>) nounwind readnone
    121 
    122 define <2 x double> @test_x86_fnmsub_sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
    123   ; CHECK: fnmsub213sd
    124   %res = call <2 x double> @llvm.x86.fma.vfnmsub.sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) nounwind
    125   ret <2 x double> %res
    126 }
    127 declare <2 x double> @llvm.x86.fma.vfnmsub.sd(<2 x double>, <2 x double>, <2 x double>) nounwind readnone
    128 
    129 define <2 x double> @test_x86_fnmsub_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
    130   ; CHECK: fnmsub213pd
    131   %res = call <2 x double> @llvm.x86.fma.vfnmsub.pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) nounwind
    132   ret <2 x double> %res
    133 }
    134 declare <2 x double> @llvm.x86.fma.vfnmsub.pd(<2 x double>, <2 x double>, <2 x double>) nounwind readnone
    135