llvm.org GIT mirror llvm / release_60 test / CodeGen / X86 / sse4a-upgrade.ll
release_60

Tree @release_60 (Download .tar.gz)

sse4a-upgrade.ll @release_60raw · history · blame

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=i386-unknown-unknown -mattr=+sse4a | FileCheck %s --check-prefix=X32
; RUN: llc < %s -mtriple=i386-unknown-unknown -mattr=+sse4a,+avx | FileCheck %s --check-prefix=X32
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4a | FileCheck %s --check-prefix=X64
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4a,+avx | FileCheck %s --check-prefix=X64

define void @test_movntss(i8* %p, <4 x float> %a) nounwind optsize ssp {
; X32-LABEL: test_movntss:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movntss %xmm0, (%eax)
; X32-NEXT:    retl
;
; X64-LABEL: test_movntss:
; X64:       # %bb.0:
; X64-NEXT:    movntss %xmm0, (%rdi)
; X64-NEXT:    retq
  tail call void @llvm.x86.sse4a.movnt.ss(i8* %p, <4 x float> %a) nounwind
  ret void
}

declare void @llvm.x86.sse4a.movnt.ss(i8*, <4 x float>)

define void @test_movntsd(i8* %p, <2 x double> %a) nounwind optsize ssp {
; X32-LABEL: test_movntsd:
; X32:       # %bb.0:
; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X32-NEXT:    movntsd %xmm0, (%eax)
; X32-NEXT:    retl
;
; X64-LABEL: test_movntsd:
; X64:       # %bb.0:
; X64-NEXT:    movntsd %xmm0, (%rdi)
; X64-NEXT:    retq
  tail call void @llvm.x86.sse4a.movnt.sd(i8* %p, <2 x double> %a) nounwind
  ret void
}

declare void @llvm.x86.sse4a.movnt.sd(i8*, <2 x double>)