diff options
| author | Evan Cheng <evan.cheng@apple.com> | 2009-12-09 21:00:30 +0000 |
|---|---|---|
| committer | Evan Cheng <evan.cheng@apple.com> | 2009-12-09 21:00:30 +0000 |
| commit | c363094e04df621d41ca570eb2a7bf8826bb8c1a (patch) | |
| tree | f471aca10ab30ffc20bb95bc94a18fed3c50357e /test/CodeGen/X86/splat-scalar-load.ll | |
| parent | 89452f7386540ca83e8991e74f1d74bbe7271922 (diff) | |
| download | llvm-c363094e04df621d41ca570eb2a7bf8826bb8c1a.tar.gz llvm-c363094e04df621d41ca570eb2a7bf8826bb8c1a.tar.bz2 llvm-c363094e04df621d41ca570eb2a7bf8826bb8c1a.tar.xz | |
Optimize splat of a scalar load into a shuffle of a vector load when it's legal. e.g.
vector_shuffle (scalar_to_vector (i32 load (ptr + 4))), undef, <0, 0, 0, 0>
=>
vector_shuffle (v4i32 load ptr), undef, <1, 1, 1, 1>
iff ptr is 16-byte aligned (or can be made into 16-byte aligned).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90984 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen/X86/splat-scalar-load.ll')
| -rw-r--r-- | test/CodeGen/X86/splat-scalar-load.ll | 43 |
1 files changed, 43 insertions, 0 deletions
diff --git a/test/CodeGen/X86/splat-scalar-load.ll b/test/CodeGen/X86/splat-scalar-load.ll new file mode 100644 index 0000000000..6c93efab70 --- /dev/null +++ b/test/CodeGen/X86/splat-scalar-load.ll @@ -0,0 +1,43 @@ +; RUN: llc < %s -march=x86 -mattr=+sse2 | FileCheck %s +; rdar://7434544 + +define <2 x i64> @t1() nounwind ssp { +entry: +; CHECK: t1: +; CHECK: pshufd $0, (%esp), %xmm0 + %array = alloca [8 x float], align 16 + %arrayidx = getelementptr inbounds [8 x float]* %array, i32 0, i32 0 + %tmp2 = load float* %arrayidx + %vecinit = insertelement <4 x float> undef, float %tmp2, i32 0 + %vecinit5 = insertelement <4 x float> %vecinit, float %tmp2, i32 1 + %vecinit7 = insertelement <4 x float> %vecinit5, float %tmp2, i32 2 + %vecinit9 = insertelement <4 x float> %vecinit7, float %tmp2, i32 3 + %0 = bitcast <4 x float> %vecinit9 to <2 x i64> + ret <2 x i64> %0 +} + +define <2 x i64> @t2() nounwind ssp { +entry: +; CHECK: t2: +; CHECK: pshufd $85, (%esp), %xmm0 + %array = alloca [8 x float], align 4 + %arrayidx = getelementptr inbounds [8 x float]* %array, i32 0, i32 1 + %tmp2 = load float* %arrayidx + %vecinit = insertelement <4 x float> undef, float %tmp2, i32 0 + %vecinit5 = insertelement <4 x float> %vecinit, float %tmp2, i32 1 + %vecinit7 = insertelement <4 x float> %vecinit5, float %tmp2, i32 2 + %vecinit9 = insertelement <4 x float> %vecinit7, float %tmp2, i32 3 + %0 = bitcast <4 x float> %vecinit9 to <2 x i64> + ret <2 x i64> %0 +} + +define <4 x float> @t3(float %tmp1, float %tmp2, float %tmp3) nounwind readnone ssp { +entry: +; CHECK: t3: +; CHECK: pshufd $-86, (%esp), %xmm0 + %0 = insertelement <4 x float> undef, float %tmp3, i32 0 + %1 = insertelement <4 x float> %0, float %tmp3, i32 1 + %2 = insertelement <4 x float> %1, float %tmp3, i32 2 + %3 = insertelement <4 x float> %2, float %tmp3, i32 3 + ret <4 x float> %3 +} |