diff options
| author | Craig Topper <craig.topper@intel.com> | 2018-03-24 01:52:01 +0000 |
|---|---|---|
| committer | Craig Topper <craig.topper@intel.com> | 2018-03-24 01:52:01 +0000 |
| commit | 2c0a62ab9afe781d886e9d37e0811846abc0cffe (patch) | |
| tree | 62b6702267d7dd1b892f01e5cef84ade95a36646 /llvm/test | |
| parent | 8ce0ab39239ded630cf72c9fbe86dbdfb1561ca0 (diff) | |
| download | bcm5719-llvm-2c0a62ab9afe781d886e9d37e0811846abc0cffe.tar.gz bcm5719-llvm-2c0a62ab9afe781d886e9d37e0811846abc0cffe.zip | |
[X86] Add a DAG combine to simplify PMULDQ/PMULUDQ nodes
These nodes only use the lower 32 bits of their inputs so we can use SimplifyDemandedBits to simplify them.
Differential Revision: https://reviews.llvm.org/D44375
llvm-svn: 328405
Diffstat (limited to 'llvm/test')
| -rw-r--r-- | llvm/test/CodeGen/X86/vector-mul.ll | 34 |
1 files changed, 16 insertions, 18 deletions
diff --git a/llvm/test/CodeGen/X86/vector-mul.ll b/llvm/test/CodeGen/X86/vector-mul.ll index 15861353d16..80d18a39384 100644 --- a/llvm/test/CodeGen/X86/vector-mul.ll +++ b/llvm/test/CodeGen/X86/vector-mul.ll @@ -692,17 +692,16 @@ define <2 x i64> @mul_v2i64_15_63(<2 x i64> %a0) nounwind { define <2 x i64> @mul_v2i64_neg_15_63(<2 x i64> %a0) nounwind { ; X86-LABEL: mul_v2i64_neg_15_63: ; X86: # %bb.0: -; X86-NEXT: movdqa %xmm0, %xmm1 -; X86-NEXT: psrlq $32, %xmm1 -; X86-NEXT: movdqa {{.*#+}} xmm2 = [4294967281,4294967295,4294967233,4294967295] -; X86-NEXT: pmuludq %xmm2, %xmm1 -; X86-NEXT: movdqa %xmm2, %xmm3 +; X86-NEXT: movdqa {{.*#+}} xmm1 = [4294967281,4294967295,4294967233,4294967295] +; X86-NEXT: movdqa %xmm0, %xmm2 +; X86-NEXT: pmuludq %xmm1, %xmm2 +; X86-NEXT: movdqa %xmm0, %xmm3 ; X86-NEXT: psrlq $32, %xmm3 -; X86-NEXT: pmuludq %xmm0, %xmm3 -; X86-NEXT: paddq %xmm1, %xmm3 -; X86-NEXT: psllq $32, %xmm3 -; X86-NEXT: pmuludq %xmm2, %xmm0 +; X86-NEXT: pmuludq %xmm1, %xmm3 +; X86-NEXT: pmuludq {{\.LCPI.*}}, %xmm0 ; X86-NEXT: paddq %xmm3, %xmm0 +; X86-NEXT: psllq $32, %xmm0 +; X86-NEXT: paddq %xmm2, %xmm0 ; X86-NEXT: retl ; ; X64-LABEL: mul_v2i64_neg_15_63: @@ -737,17 +736,16 @@ define <2 x i64> @mul_v2i64_neg_15_63(<2 x i64> %a0) nounwind { define <2 x i64> @mul_v2i64_neg_17_65(<2 x i64> %a0) nounwind { ; X86-LABEL: mul_v2i64_neg_17_65: ; X86: # %bb.0: -; X86-NEXT: movdqa %xmm0, %xmm1 -; X86-NEXT: psrlq $32, %xmm1 -; X86-NEXT: movdqa {{.*#+}} xmm2 = [4294967279,4294967295,4294967231,4294967295] -; X86-NEXT: pmuludq %xmm2, %xmm1 -; X86-NEXT: movdqa %xmm2, %xmm3 +; X86-NEXT: movdqa {{.*#+}} xmm1 = [4294967279,4294967295,4294967231,4294967295] +; X86-NEXT: movdqa %xmm0, %xmm2 +; X86-NEXT: pmuludq %xmm1, %xmm2 +; X86-NEXT: movdqa %xmm0, %xmm3 ; X86-NEXT: psrlq $32, %xmm3 -; X86-NEXT: pmuludq %xmm0, %xmm3 -; X86-NEXT: paddq %xmm1, %xmm3 -; X86-NEXT: psllq $32, %xmm3 -; X86-NEXT: pmuludq %xmm2, %xmm0 +; X86-NEXT: pmuludq %xmm1, %xmm3 +; X86-NEXT: pmuludq {{\.LCPI.*}}, %xmm0 ; X86-NEXT: paddq %xmm3, %xmm0 +; X86-NEXT: psllq $32, %xmm0 +; X86-NEXT: paddq %xmm2, %xmm0 ; X86-NEXT: retl ; ; X64-LABEL: mul_v2i64_neg_17_65: |
