diff options
| author | Simon Pilgrim <llvm-dev@redking.me.uk> | 2019-08-12 14:01:16 +0000 |
|---|---|---|
| committer | Simon Pilgrim <llvm-dev@redking.me.uk> | 2019-08-12 14:01:16 +0000 |
| commit | 9a452740ec89f856492841af3f9c616f5be96c7c (patch) | |
| tree | 1c724ca43b3528c0c96ab3303c8b8d642bbd84b7 | |
| parent | 489efc68a572c42f71590dd6bc0ea21d82d727b3 (diff) | |
| download | bcm5719-llvm-9a452740ec89f856492841af3f9c616f5be96c7c.tar.gz bcm5719-llvm-9a452740ec89f856492841af3f9c616f5be96c7c.zip | |
[X86][SSE] Add test showing missing demanded elts PSADBW handling
llvm-svn: 368575
| -rw-r--r-- | llvm/test/CodeGen/X86/psadbw.ll | 26 |
1 files changed, 26 insertions, 0 deletions
diff --git a/llvm/test/CodeGen/X86/psadbw.ll b/llvm/test/CodeGen/X86/psadbw.ll index 46916898063..753e88c3dbc 100644 --- a/llvm/test/CodeGen/X86/psadbw.ll +++ b/llvm/test/CodeGen/X86/psadbw.ll @@ -13,5 +13,31 @@ define <2 x i64> @combine_psadbw_shift(<16 x i8> %0, <16 x i8> %1) { ret <2 x i64> %4 } +; Propagate the demanded result elements to the 8 aliasing source elements. +define i64 @combine_psadbw_demandedelt(<16 x i8> %0, <16 x i8> %1) { +; X86-LABEL: combine_psadbw_demandedelt: +; X86: # %bb.0: +; X86-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,3,2] +; X86-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,3,2] +; X86-NEXT: psadbw %xmm0, %xmm1 +; X86-NEXT: movd %xmm1, %eax +; X86-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3] +; X86-NEXT: movd %xmm0, %edx +; X86-NEXT: retl +; +; X64-LABEL: combine_psadbw_demandedelt: +; X64: # %bb.0: +; X64-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,3,2] +; X64-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,3,2] +; X64-NEXT: psadbw %xmm0, %xmm1 +; X64-NEXT: movq %xmm1, %rax +; X64-NEXT: retq + %3 = shufflevector <16 x i8> %0, <16 x i8> %0, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 12, i32 13, i32 14, i32 15, i32 8, i32 9, i32 10, i32 11> + %4 = shufflevector <16 x i8> %1, <16 x i8> %1, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 12, i32 13, i32 14, i32 15, i32 8, i32 9, i32 10, i32 11> + %5 = tail call <2 x i64> @llvm.x86.sse2.psad.bw(<16 x i8> %3, <16 x i8> %4) + %6 = extractelement <2 x i64> %5, i32 0 + ret i64 %6 +} + declare <2 x i64> @llvm.x86.sse2.psad.bw(<16 x i8>, <16 x i8>) |
