path: root/test/CodeGen/X86/xop-intrinsics-x86_64.ll

; RUN: llc < %s -mtriple=x86_64-unknown-unknown -march=x86-64 -mattr=+avx,+fma4,+xop | FileCheck %s

define <2 x double> @test_int_x86_xop_vpermil2pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
  ; CHECK: vpermil2pd
  %res = call <2 x double> @llvm.x86.xop.vpermil2pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 1) ;  [#uses=1]
  ret <2 x double> %res
}
define <2 x double> @test_int_x86_xop_vpermil2pd_mr(<2 x double> %a0, <2 x double>* %a1, <2 x double> %a2) {
  ; CHECK-NOT: vmovaps
  ; CHECK: vpermil2pd
  %vec = load <2 x double>* %a1
  %res = call <2 x double> @llvm.x86.xop.vpermil2pd(<2 x double> %a0, <2 x double> %vec, <2 x double> %a2, i8 1) ;  [#uses=1]
  ret <2 x double> %res
}
define <2 x double> @test_int_x86_xop_vpermil2pd_rm(<2 x double> %a0, <2 x double> %a1, <2 x double>* %a2) {
  ; CHECK-NOT: vmovaps
  ; CHECK: vpermil2pd
  %vec = load <2 x double>* %a2
  %res = call <2 x double> @llvm.x86.xop.vpermil2pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %vec, i8 1) ;  [#uses=1]
  ret <2 x double> %res
}
declare <2 x double> @llvm.x86.xop.vpermil2pd(<2 x double>, <2 x double>, <2 x double>, i8) nounwind readnone

define <4 x double> @test_int_x86_xop_vpermil2pd_256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) {
  ; CHECK: vpermil2pd
  ; CHECK: ymm
  %res = call <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 2) ;
  ret <4 x double> %res
}
define <4 x double> @test_int_x86_xop_vpermil2pd_256_mr(<4 x double> %a0, <4 x double>* %a1, <4 x double> %a2) {
  ; CHECK-NOT: vmovaps
  ; CHECK: vpermil2pd
  ; CHECK: ymm
  %vec = load <4 x double>* %a1
  %res = call <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double> %a0, <4 x double> %vec, <4 x double> %a2, i8 2) ;
  ret <4 x double> %res
}
define <4 x double> @test_int_x86_xop_vpermil2pd_256_rm(<4 x double> %a0, <4 x double> %a1, <4 x double>* %a2) {
  ; CHECK-NOT: vmovaps
  ; CHECK: vpermil2pd
  ; CHECK: ymm
  %vec = load <4 x double>* %a2
  %res = call <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %vec, i8 2) ;
  ret <4 x double> %res
}
declare <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double>, <4 x double>, <4 x double>, i8) nounwind readnone

define <4 x float>