1 files changed, 12 insertions, 8 deletions
diff --git a/test/CodeGen/X86/sse4a-intrinsics-fast-isel.ll b/test/CodeGen/X86/sse4a-intrinsics-fast-isel.ll
index f93a16a5eb3d..f45abf1d85df 100644
--- a/test/CodeGen/X86/sse4a-intrinsics-fast-isel.ll
+++ b/test/CodeGen/X86/sse4a-intrinsics-fast-isel.ll
@@ -1,6 +1,8 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -fast-isel -mtriple=i386-unknown-unknown -mattr=+sse4a | FileCheck %s --check-prefix=ALL  --check-prefix=X32
-; RUN: llc < %s -fast-isel -mtriple=x86_64-unknown-unknown -mattr=+sse4a | FileCheck %s --check-prefix=ALL  --check-prefix=X64
+; RUN: llc < %s -fast-isel -mtriple=i386-unknown-unknown -mattr=+sse4a | FileCheck %s --check-prefix=X32
+; RUN: llc < %s -fast-isel -mtriple=i386-unknown-unknown -mattr=+sse4a,+avx | FileCheck %s --check-prefix=X32
+; RUN: llc < %s -fast-isel -mtriple=x86_64-unknown-unknown -mattr=+sse4a | FileCheck %s --check-prefix=X64
+; RUN: llc < %s -fast-isel -mtriple=x86_64-unknown-unknown -mattr=+sse4a,+avx | FileCheck %s --check-prefix=X64
 
 ; NOTE: This should use IR equivalent to what is generated by clang/test/CodeGen/sse4a-builtins.c
 
@@ -65,7 +67,7 @@ define <2 x i64> @test_mm_insert_si64(<2 x i64> %x, <2 x i64> %y) {
 }
 declare <2 x i64> @llvm.x86.sse4a.insertq(<2 x i64>, <2 x i64>) nounwind readnone
 
-define void @test_stream_sd(i8* %p, <2 x double> %a) {
+define void @test_stream_sd(double* %p, <2 x double> %a) {
 ; X32-LABEL: test_stream_sd:
 ; X32:       # BB#0:
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
@@ -76,12 +78,12 @@ define void @test_stream_sd(i8* %p, <2 x double> %a) {
 ; X64:       # BB#0:
 ; X64-NEXT:    movntsd %xmm0, (%rdi)
 ; X64-NEXT:    retq
-  call void @llvm.x86.sse4a.movnt.sd(i8* %p, <2 x double> %a)
+  %1 = extractelement <2 x double> %a, i64 0
+  store double %1, double* %p, align 1, !nontemporal !1
   ret void
 }
-declare void @llvm.x86.sse4a.movnt.sd(i8*, <2 x double>) nounwind readnone
 
-define void @test_mm_stream_ss(i8* %p, <4 x float> %a) {
+define void @test_mm_stream_ss(float* %p, <4 x float> %a) {
 ; X32-LABEL: test_mm_stream_ss:
 ; X32:       # BB#0:
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
@@ -92,7 +94,9 @@ define void @test_mm_stream_ss(i8* %p, <4 x float> %a) {
 ; X64:       # BB#0:
 ; X64-NEXT:    movntss %xmm0, (%rdi)
 ; X64-NEXT:    retq
-  call void @llvm.x86.sse4a.movnt.ss(i8* %p, <4 x float> %a)
+  %1 = extractelement <4 x float> %a, i64 0
+  store float %1, float* %p, align 1, !nontemporal !1
   ret void
 }
-declare void @llvm.x86.sse4a.movnt.ss(i8*, <4 x float>) nounwind readnone
+
+!1 = !{i32 1}