Fix a broadcast related regression on the vector shuffle lowering.

[oota-llvm.git] / test / CodeGen / X86 / sse4a.ll

diff --git a/test/CodeGen/X86/sse4a.ll b/test/CodeGen/X86/sse4a.ll
index 0732353fce4876d0475d67c9fd880151bb4ff16c..165d47639d7acadd6346d0a2b2ffd67641a7ffba 100644 (file)
--- a/test/CodeGen/X86/sse4a.ll
+++ b/test/CodeGen/X86/sse4a.ll
@@ -1,19 +1,56 @@
 ; RUN: llc < %s -mtriple=i686-apple-darwin9 -mattr=sse4a | FileCheck %s
 
-define void @test1(float* %p, <4 x float> %a) nounwind optsize ssp {
+define void @test1(i8* %p, <4 x float> %a) nounwind optsize ssp {
+; CHECK-LABEL: test1:
 ; CHECK: movntss
-entry:
-  tail call void @llvm.x86.sse4a.movnt.ss(float* %p, <4 x float> %a) nounwind
+  tail call void @llvm.x86.sse4a.movnt.ss(i8* %p, <4 x float> %a) nounwind
   ret void
 }
 
-declare void @llvm.x86.sse4a.movnt.ss(float*, <4 x float>)
+declare void @llvm.x86.sse4a.movnt.ss(i8*, <4 x float>)
 
-define void @test2(double* %p, <2 x double> %a) nounwind optsize ssp {
+define void @test2(i8* %p, <2 x double> %a) nounwind optsize ssp {
+; CHECK-LABEL: test2:
 ; CHECK: movntsd
-entry:
-  tail call void @llvm.x86.sse4a.movnt.sd(double* %p, <2 x double> %a) nounwind
+  tail call void @llvm.x86.sse4a.movnt.sd(i8* %p, <2 x double> %a) nounwind
   ret void
 }
 
-declare void @llvm.x86.sse4a.movnt.sd(double*, <2 x double>)
+declare void @llvm.x86.sse4a.movnt.sd(i8*, <2 x double>)
+
+define <2 x i64> @test3(<2 x i64> %x) nounwind uwtable ssp {
+; CHECK-LABEL: test3:
+; CHECK: extrq
+  %1 = tail call <2 x i64> @llvm.x86.sse4a.extrqi(<2 x i64> %x, i8 3, i8 2)
+  ret <2 x i64> %1
+}
+
+declare <2 x i64> @llvm.x86.sse4a.extrqi(<2 x i64>, i8, i8) nounwind
+
+define <2 x i64> @test4(<2 x i64> %x, <2 x i64> %y) nounwind uwtable ssp {
+; CHECK-LABEL: test4:
+; CHECK: extrq
+  %1 = bitcast <2 x i64> %y to <16 x i8>
+  %2 = tail call <2 x i64> @llvm.x86.sse4a.extrq(<2 x i64> %x, <16 x i8> %1) nounwind
+  ret <2 x i64> %2
+}
+
+declare <2 x i64> @llvm.x86.sse4a.extrq(<2 x i64>, <16 x i8>) nounwind
+
+define <2 x i64> @test5(<2 x i64> %x, <2 x i64> %y) nounwind uwtable ssp {
+; CHECK-LABEL: test5:
+; CHECK: insertq
+  %1 = tail call <2 x i64> @llvm.x86.sse4a.insertqi(<2 x i64> %x, <2 x i64> %y, i8 5, i8 6)
+  ret <2 x i64> %1
+}
+
+declare <2 x i64> @llvm.x86.sse4a.insertqi(<2 x i64>, <2 x i64>, i8, i8) nounwind
+
+define <2 x i64> @test6(<2 x i64> %x, <2 x i64> %y) nounwind uwtable ssp {
+; CHECK-LABEL: test6:
+; CHECK: insertq
+  %1 = tail call <2 x i64> @llvm.x86.sse4a.insertq(<2 x i64> %x, <2 x i64> %y) nounwind
+  ret <2 x i64> %1
+}
+
+declare <2 x i64> @llvm.x86.sse4a.insertq(<2 x i64>, <2 x i64>) nounwind