diff options
author | Evan Cheng <evan.cheng@apple.com> | 2008-05-13 08:35:03 +0000 |
---|---|---|
committer | Evan Cheng <evan.cheng@apple.com> | 2008-05-13 08:35:03 +0000 |
commit | 77f0b7a50a08614b5ffd58f1864b68a9a30d0cb0 (patch) | |
tree | bb8e0ed03e63e856988c905af714ffb7bd62205e /test/CodeGen/X86/vec_extract-sse4.ll | |
parent | 6513c1bf90be63b9bd6a43c0db593174075009a0 (diff) | |
download | external_llvm-77f0b7a50a08614b5ffd58f1864b68a9a30d0cb0.zip external_llvm-77f0b7a50a08614b5ffd58f1864b68a9a30d0cb0.tar.gz external_llvm-77f0b7a50a08614b5ffd58f1864b68a9a30d0cb0.tar.bz2 |
Instead of a vector load, shuffle and then extract an element. Load the element from address with an offset.
pshufd $1, (%rdi), %xmm0
movd %xmm0, %eax
=>
movl 4(%rdi), %eax
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51026 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen/X86/vec_extract-sse4.ll')
-rw-r--r-- | test/CodeGen/X86/vec_extract-sse4.ll | 15 |
1 files changed, 8 insertions, 7 deletions
diff --git a/test/CodeGen/X86/vec_extract-sse4.ll b/test/CodeGen/X86/vec_extract-sse4.ll index 1ef5e88..d6726be 100644 --- a/test/CodeGen/X86/vec_extract-sse4.ll +++ b/test/CodeGen/X86/vec_extract-sse4.ll @@ -1,29 +1,30 @@ ; RUN: llvm-as < %s | llc -march=x86 -mattr=+sse41 -o %t -f -; RUN: grep extractps %t | count 1 -; RUN: grep pextrd %t | count 2 -; RUN: grep pshufd %t | count 1 +; RUN: grep extractps %t | count 1 +; RUN: grep pextrd %t | count 1 +; RUN: not grep pshufd %t +; RUN: not grep movss %t -define void @t1(float* %R, <4 x float>* %P1) { +define void @t1(float* %R, <4 x float>* %P1) nounwind { %X = load <4 x float>* %P1 %tmp = extractelement <4 x float> %X, i32 3 store float %tmp, float* %R ret void } -define float @t2(<4 x float>* %P1) { +define float @t2(<4 x float>* %P1) nounwind { %X = load <4 x float>* %P1 %tmp = extractelement <4 x float> %X, i32 2 ret float %tmp } -define void @t3(i32* %R, <4 x i32>* %P1) { +define void @t3(i32* %R, <4 x i32>* %P1) nounwind { %X = load <4 x i32>* %P1 %tmp = extractelement <4 x i32> %X, i32 3 store i32 %tmp, i32* %R ret void } -define i32 @t4(<4 x i32>* %P1) { +define i32 @t4(<4 x i32>* %P1) nounwind { %X = load <4 x i32>* %P1 %tmp = extractelement <4 x i32> %X, i32 3 ret i32 %tmp |