diff options
| author | Robert Lougher <rob.lougher@gmail.com> | 2014-02-11 15:42:46 +0000 |
|---|---|---|
| committer | Robert Lougher <rob.lougher@gmail.com> | 2014-02-11 15:42:46 +0000 |
| commit | 7d9084ffa13aabb4e3f7f6bd8d8263dce982421e (patch) | |
| tree | a4320aa05e50955bfda075310488e9800abff5a8 /llvm/test/CodeGen/X86/avx-vbroadcast.ll | |
| parent | 2228cd3f070da33be9e8ae456858ac5f2bd59679 (diff) | |
| download | bcm5719-llvm-7d9084ffa13aabb4e3f7f6bd8d8263dce982421e.tar.gz bcm5719-llvm-7d9084ffa13aabb4e3f7f6bd8d8263dce982421e.zip | |
Teach the DAGCombiner how to fold concat_vector nodes when the input is two
BUILD_VECTOR nodes, e.g.:
(concat_vectors (BUILD_VECTOR a1, a2, a3, a4), (BUILD_VECTOR b1, b2, b3, b4))
->
(BUILD_VECTOR a1, a2, a3, a4, b1, b2, b3, b4)
This fixes an issue with AVX, where a sequence was not recognized as a 256-bit
vbroadcast due to the concat_vectors.
llvm-svn: 201158
Diffstat (limited to 'llvm/test/CodeGen/X86/avx-vbroadcast.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/avx-vbroadcast.ll | 63 |
1 files changed, 63 insertions, 0 deletions
diff --git a/llvm/test/CodeGen/X86/avx-vbroadcast.ll b/llvm/test/CodeGen/X86/avx-vbroadcast.ll index 0d403d4bb12..2ebe6fda37a 100644 --- a/llvm/test/CodeGen/X86/avx-vbroadcast.ll +++ b/llvm/test/CodeGen/X86/avx-vbroadcast.ll @@ -141,3 +141,66 @@ entry: ret <4 x float> %t } + +; These tests check that a vbroadcast instruction is used when we have a splat +; formed from a concat_vectors (via the shufflevector) of two BUILD_VECTORs +; (via the insertelements). + +; CHECK-LABEL: splat_concat1 +; CHECK-NOT: vinsertf128 +; CHECK: vbroadcastss (% +; CHECK-NEXT: ret +define <8 x float> @splat_concat1(float* %p) { + %1 = load float* %p, align 4 + %2 = insertelement <4 x float> undef, float %1, i32 0 + %3 = insertelement <4 x float> %2, float %1, i32 1 + %4 = insertelement <4 x float> %3, float %1, i32 2 + %5 = insertelement <4 x float> %4, float %1, i32 3 + %6 = shufflevector <4 x float> %5, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3> + ret <8 x float> %6 +} + +; CHECK-LABEL: splat_concat2 +; CHECK-NOT: vinsertf128 +; CHECK: vbroadcastss (% +; CHECK-NEXT: ret +define <8 x float> @splat_concat2(float* %p) { + %1 = load float* %p, align 4 + %2 = insertelement <4 x float> undef, float %1, i32 0 + %3 = insertelement <4 x float> %2, float %1, i32 1 + %4 = insertelement <4 x float> %3, float %1, i32 2 + %5 = insertelement <4 x float> %4, float %1, i32 3 + %6 = insertelement <4 x float> undef, float %1, i32 0 + %7 = insertelement <4 x float> %6, float %1, i32 1 + %8 = insertelement <4 x float> %7, float %1, i32 2 + %9 = insertelement <4 x float> %8, float %1, i32 3 + %10 = shufflevector <4 x float> %5, <4 x float> %9, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> + ret <8 x float> %10 +} + +; CHECK-LABEL: splat_concat3 +; CHECK-NOT: vinsertf128 +; CHECK: vbroadcastsd (% +; CHECK-NEXT: ret +define <4 x double> @splat_concat3(double* %p) { + %1 = load double* %p, align 8 + %2 = insertelement <2 x double> undef, double %1, i32 0 + %3 = insertelement <2 x double> %2, double %1, i32 1 + %4 = shufflevector <2 x double> %3, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1> + ret <4 x double> %4 +} + +; CHECK-LABEL: splat_concat4 +; CHECK-NOT: vinsertf128 +; CHECK: vbroadcastsd (% +; CHECK-NEXT: ret +define <4 x double> @splat_concat4(double* %p) { + %1 = load double* %p, align 8 + %2 = insertelement <2 x double> undef, double %1, i32 0 + %3 = insertelement <2 x double> %2, double %1, i32 1 + %4 = insertelement <2 x double> undef, double %1, i32 0 + %5 = insertelement <2 x double> %2, double %1, i32 1 + %6 = shufflevector <2 x double> %3, <2 x double> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3> + ret <4 x double> %6 +} + |
