diff options
| author | Evan Cheng <evan.cheng@apple.com> | 2008-02-18 23:04:32 +0000 |
|---|---|---|
| committer | Evan Cheng <evan.cheng@apple.com> | 2008-02-18 23:04:32 +0000 |
| commit | 6200c225e0ab09dfa66154c39dacfbefb0f9a99e (patch) | |
| tree | 0ad268b8d99a79cc3b51f86ab6a3958a11f32ca6 /llvm/test | |
| parent | a6f63d94a5c0a9adaef66ee43d6d4938e66b83de (diff) | |
| download | bcm5719-llvm-6200c225e0ab09dfa66154c39dacfbefb0f9a99e.tar.gz bcm5719-llvm-6200c225e0ab09dfa66154c39dacfbefb0f9a99e.zip | |
- When DAG combiner is folding a bit convert into a BUILD_VECTOR, it should check if it's essentially a SCALAR_TO_VECTOR. Avoid turning (v8i16) <10, u, u, u> to <10, 0, u, u, u, u, u, u>. Instead, simply convert it to a SCALAR_TO_VECTOR of the proper type.
- X86 now normalize SCALAR_TO_VECTOR to (BIT_CONVERT (v4i32 SCALAR_TO_VECTOR)). Get rid of X86ISD::S2VEC.
llvm-svn: 47290
Diffstat (limited to 'llvm/test')
| -rw-r--r-- | llvm/test/CodeGen/X86/vec_shift2.ll | 17 |
1 files changed, 17 insertions, 0 deletions
diff --git a/llvm/test/CodeGen/X86/vec_shift2.ll b/llvm/test/CodeGen/X86/vec_shift2.ll new file mode 100644 index 00000000000..b73f5f49000 --- /dev/null +++ b/llvm/test/CodeGen/X86/vec_shift2.ll @@ -0,0 +1,17 @@ +; RUN: llvm-as < %s | llc -march=x86 -mattr=+sse2 | not grep CPI + +define <2 x i64> @t1(<2 x i64> %b1, <2 x i64> %c) nounwind { + %tmp1 = bitcast <2 x i64> %b1 to <8 x i16> + %tmp2 = tail call <8 x i16> @llvm.x86.sse2.psrl.w( <8 x i16> %tmp1, <8 x i16> bitcast (<4 x i32> < i32 14, i32 undef, i32 undef, i32 undef > to <8 x i16>) ) nounwind readnone + %tmp3 = bitcast <8 x i16> %tmp2 to <2 x i64> + ret <2 x i64> %tmp3 +} + +define <4 x i32> @t2(<2 x i64> %b1, <2 x i64> %c) nounwind { + %tmp1 = bitcast <2 x i64> %b1 to <4 x i32> + %tmp2 = tail call <4 x i32> @llvm.x86.sse2.psll.d( <4 x i32> %tmp1, <4 x i32> < i32 14, i32 undef, i32 undef, i32 undef > ) nounwind readnone + ret <4 x i32> %tmp2 +} + +declare <8 x i16> @llvm.x86.sse2.psrl.w(<8 x i16>, <8 x i16>) nounwind readnone +declare <4 x i32> @llvm.x86.sse2.psll.d(<4 x i32>, <4 x i32>) nounwind readnone |
