diff options
| author | Nate Begeman <natebegeman@mac.com> | 2010-07-28 00:21:48 +0000 |
|---|---|---|
| committer | Nate Begeman <natebegeman@mac.com> | 2010-07-28 00:21:48 +0000 |
| commit | 53afc8f06af1db2e52498441b82f96bea283a6f6 (patch) | |
| tree | ef666034dcb9cf0d6a2f999a3ea698fcee74297c /llvm/test/CodeGen/X86/vec_shift4.ll | |
| parent | a3659efe0932df57c8b07bdfaa0c804b97ed5820 (diff) | |
| download | bcm5719-llvm-53afc8f06af1db2e52498441b82f96bea283a6f6.tar.gz bcm5719-llvm-53afc8f06af1db2e52498441b82f96bea283a6f6.zip | |
Implement a vectorized algorithm for <16 x i8> << <16 x i8>
This is about 4x faster and smaller than the existing scalarization.
llvm-svn: 109566
Diffstat (limited to 'llvm/test/CodeGen/X86/vec_shift4.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/vec_shift4.ll | 13 |
1 files changed, 12 insertions, 1 deletions
diff --git a/llvm/test/CodeGen/X86/vec_shift4.ll b/llvm/test/CodeGen/X86/vec_shift4.ll index d8f4e4ec689..9ef7fbdb0c5 100644 --- a/llvm/test/CodeGen/X86/vec_shift4.ll +++ b/llvm/test/CodeGen/X86/vec_shift4.ll @@ -1,6 +1,6 @@ ; RUN: llc < %s -march=x86 -mattr=+sse41 | FileCheck %s -define <2 x i64> @shl(<4 x i32> %r, <4 x i32> %a) nounwind readnone ssp { +define <2 x i64> @shl1(<4 x i32> %r, <4 x i32> %a) nounwind readnone ssp { entry: ; CHECK-NOT: shll ; CHECK: pslld @@ -12,3 +12,14 @@ entry: %tmp2 = bitcast <4 x i32> %shl to <2 x i64> ; <<2 x i64>> [#uses=1] ret <2 x i64> %tmp2 } + +define <2 x i64> @shl2(<16 x i8> %r, <16 x i8> %a) nounwind readnone ssp { +entry: +; CHECK-NOT: shlb +; CHECK: pblendvb +; CHECK: pblendvb +; CHECK: pblendvb + %shl = shl <16 x i8> %r, %a ; <<16 x i8>> [#uses=1] + %tmp2 = bitcast <16 x i8> %shl to <2 x i64> ; <<2 x i64>> [#uses=1] + ret <2 x i64> %tmp2 +} |
