diff options
| author | Nadav Rotem <nrotem@apple.com> | 2012-09-16 07:39:07 +0000 |
|---|---|---|
| committer | Nadav Rotem <nrotem@apple.com> | 2012-09-16 07:39:07 +0000 |
| commit | 638e4c13cb25e8ad1044c1d581fb387ec8d15033 (patch) | |
| tree | b6e61b7c40456f0778feab86ac8529afa3011633 /test/CodeGen/X86/pmovext.ll | |
| parent | 3c0e5c9ecedb00d3c36fb2747b642bd3e38d0260 (diff) | |
| download | llvm-638e4c13cb25e8ad1044c1d581fb387ec8d15033.tar.gz llvm-638e4c13cb25e8ad1044c1d581fb387ec8d15033.tar.bz2 llvm-638e4c13cb25e8ad1044c1d581fb387ec8d15033.tar.xz | |
The PMOVZXWD family of functions had patterns extends narrow vector types to wide vector types.
It had patterns for zext-loading and extending. This commit adds patterns for loading a wide type, performing a bitcast,
and extending. This is an odd pattern, but it is commonly used when writing code with intrinsics.
rdar://11897677
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163995 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen/X86/pmovext.ll')
| -rw-r--r-- | test/CodeGen/X86/pmovext.ll | 22 |
1 files changed, 22 insertions, 0 deletions
diff --git a/test/CodeGen/X86/pmovext.ll b/test/CodeGen/X86/pmovext.ll new file mode 100644 index 0000000000..c4530465c7 --- /dev/null +++ b/test/CodeGen/X86/pmovext.ll @@ -0,0 +1,22 @@ +; RUN: llc < %s -march=x86-64 -mcpu=corei7 | FileCheck %s + +; rdar://11897677 + +;CHECK: intrin_pmov +;CHECK: pmovzxbw (%rsi), %xmm0 +;CHECK-NEXT: movdqu +;CHECK-NEXT: ret +define void @intrin_pmov(i16* noalias %dest, i8* noalias %src) nounwind uwtable ssp { + %1 = bitcast i8* %src to <2 x i64>* + %2 = load <2 x i64>* %1, align 16 + %3 = bitcast <2 x i64> %2 to <16 x i8> + %4 = tail call <8 x i16> @llvm.x86.sse41.pmovzxbw(<16 x i8> %3) nounwind + %5 = bitcast i16* %dest to i8* + %6 = bitcast <8 x i16> %4 to <16 x i8> + tail call void @llvm.x86.sse2.storeu.dq(i8* %5, <16 x i8> %6) nounwind + ret void +} + +declare <8 x i16> @llvm.x86.sse41.pmovzxbw(<16 x i8>) nounwind readnone + +declare void @llvm.x86.sse2.storeu.dq(i8*, <16 x i8>) nounwind |