diff options
| author | Benjamin Kramer <benny.kra@googlemail.com> | 2013-04-18 21:37:45 +0000 |
|---|---|---|
| committer | Benjamin Kramer <benny.kra@googlemail.com> | 2013-04-18 21:37:45 +0000 |
| commit | fcba22decb8f44ff9aa8807f275ca3a32589fba0 (patch) | |
| tree | 7a9672c3b9c41ebcfa9a55d435852a02e71eb1b4 /test | |
| parent | 8bb3b098b98cd62f074bb488d8294498ae8121e3 (diff) | |
| download | llvm-fcba22decb8f44ff9aa8807f275ca3a32589fba0.tar.gz llvm-fcba22decb8f44ff9aa8807f275ca3a32589fba0.tar.bz2 llvm-fcba22decb8f44ff9aa8807f275ca3a32589fba0.tar.xz | |
X86: Add an SSE2 lowering for 64 bit compares when pcmpgtq (SSE4.2) isn't available.
This pattern started popping up in vectorized min/max reductions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179797 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test')
| -rw-r--r-- | test/CodeGen/X86/vec_compare.ll | 136 |
1 files changed, 136 insertions, 0 deletions
diff --git a/test/CodeGen/X86/vec_compare.ll b/test/CodeGen/X86/vec_compare.ll index b6d91a3f77..85d8b2cea3 100644 --- a/test/CodeGen/X86/vec_compare.ll +++ b/test/CodeGen/X86/vec_compare.ll @@ -65,3 +65,139 @@ define <2 x i64> @test6(<2 x i64> %A, <2 x i64> %B) nounwind { %D = sext <2 x i1> %C to <2 x i64> ret <2 x i64> %D } + +define <2 x i64> @test7(<2 x i64> %A, <2 x i64> %B) nounwind { +; CHECK: test7: +; CHECK: pcmpgtd %xmm1 +; CHECK: pshufd $-96 +; CHECK: pcmpeqd +; CHECK: pshufd $-11 +; CHECK: pand +; CHECK: pshufd $-11 +; CHECK: por +; CHECK: ret + %C = icmp sgt <2 x i64> %A, %B + %D = sext <2 x i1> %C to <2 x i64> + ret <2 x i64> %D +} + +define <2 x i64> @test8(<2 x i64> %A, <2 x i64> %B) nounwind { +; CHECK: test8: +; CHECK: pcmpgtd %xmm0 +; CHECK: pshufd $-96 +; CHECK: pcmpeqd +; CHECK: pshufd $-11 +; CHECK: pand +; CHECK: pshufd $-11 +; CHECK: por +; CHECK: ret + %C = icmp slt <2 x i64> %A, %B + %D = sext <2 x i1> %C to <2 x i64> + ret <2 x i64> %D +} + +define <2 x i64> @test9(<2 x i64> %A, <2 x i64> %B) nounwind { +; CHECK: test9: +; CHECK: pcmpgtd %xmm0 +; CHECK: pshufd $-96 +; CHECK: pcmpeqd +; CHECK: pshufd $-11 +; CHECK: pand +; CHECK: pshufd $-11 +; CHECK: por +; CHECK: pcmpeqd +; CHECK: pxor +; CHECK: ret + %C = icmp sge <2 x i64> %A, %B + %D = sext <2 x i1> %C to <2 x i64> + ret <2 x i64> %D +} + +define <2 x i64> @test10(<2 x i64> %A, <2 x i64> %B) nounwind { +; CHECK: test10: +; CHECK: pcmpgtd %xmm1 +; CHECK: pshufd $-96 +; CHECK: pcmpeqd +; CHECK: pshufd $-11 +; CHECK: pand +; CHECK: pshufd $-11 +; CHECK: por +; CHECK: pcmpeqd +; CHECK: pxor +; CHECK: ret + %C = icmp sle <2 x i64> %A, %B + %D = sext <2 x i1> %C to <2 x i64> + ret <2 x i64> %D +} + +define <2 x i64> @test11(<2 x i64> %A, <2 x i64> %B) nounwind { +; CHECK: test11: +; CHECK: pxor +; CHECK: pxor +; CHECK: pcmpgtd %xmm1 +; CHECK: pshufd $-96 +; CHECK: pcmpeqd +; CHECK: pshufd $-11 +; CHECK: pand +; CHECK: pshufd $-11 +; CHECK: por +; CHECK: ret + %C = icmp ugt <2 x i64> %A, %B + %D = sext <2 x i1> %C to <2 x i64> + ret <2 x i64> %D +} + +define <2 x i64> @test12(<2 x i64> %A, <2 x i64> %B) nounwind { +; CHECK: test12: +; CHECK: pxor +; CHECK: pxor +; CHECK: pcmpgtd %xmm0 +; CHECK: pshufd $-96 +; CHECK: pcmpeqd +; CHECK: pshufd $-11 +; CHECK: pand +; CHECK: pshufd $-11 +; CHECK: por +; CHECK: ret + %C = icmp ult <2 x i64> %A, %B + %D = sext <2 x i1> %C to <2 x i64> + ret <2 x i64> %D +} + +define <2 x i64> @test13(<2 x i64> %A, <2 x i64> %B) nounwind { +; CHECK: test13: +; CHECK: pxor +; CHECK: pxor +; CHECK: pcmpgtd %xmm0 +; CHECK: pshufd $-96 +; CHECK: pcmpeqd +; CHECK: pshufd $-11 +; CHECK: pand +; CHECK: pshufd $-11 +; CHECK: por +; CHECK: pcmpeqd +; CHECK: pxor +; CHECK: ret + %C = icmp uge <2 x i64> %A, %B + %D = sext <2 x i1> %C to <2 x i64> + ret <2 x i64> %D +} + +define <2 x i64> @test14(<2 x i64> %A, <2 x i64> %B) nounwind { +; CHECK: test14: +; CHECK: pxor +; CHECK: pxor +; CHECK: pcmpgtd %xmm1 +; CHECK: pshufd $-96 +; CHECK: pcmpeqd +; CHECK: pshufd $-11 +; CHECK: pand +; CHECK: pshufd $-11 +; CHECK: por +; CHECK: pcmpeqd +; CHECK: pxor +; CHECK: ret + %C = icmp ule <2 x i64> %A, %B + %D = sext <2 x i1> %C to <2 x i64> + ret <2 x i64> %D +} |