diff options
author | Chris Lattner <sabre@nondot.org> | 2010-02-16 22:35:06 +0000 |
---|---|---|
committer | Chris Lattner <sabre@nondot.org> | 2010-02-16 22:35:06 +0000 |
commit | 92d3ada814f33e3e1d10f0e5e3a15333b0cca849 (patch) | |
tree | cdda73fc41184e35406bebc258a7cb2c3dff6f6c /test/CodeGen | |
parent | 55f9adf543156f918bebcab3924d4210099e20af (diff) | |
download | external_llvm-92d3ada814f33e3e1d10f0e5e3a15333b0cca849.zip external_llvm-92d3ada814f33e3e1d10f0e5e3a15333b0cca849.tar.gz external_llvm-92d3ada814f33e3e1d10f0e5e3a15333b0cca849.tar.bz2 |
fix rdar://7653908, a crash on a case where we would fold a load
into a roundss intrinsic, producing a cyclic dag. The root cause
of this is badness handling ComplexPattern nodes in the old dagisel
that I noticed through inspection. Eliminate a copy of the of the
code that handled ComplexPatterns by making EmitChildMatchCode call
into EmitMatchCode.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@96408 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen')
-rw-r--r-- | test/CodeGen/X86/vec_ss_load_fold.ll | 25 |
1 files changed, 25 insertions, 0 deletions
diff --git a/test/CodeGen/X86/vec_ss_load_fold.ll b/test/CodeGen/X86/vec_ss_load_fold.ll index 42831f0..a44e08d 100644 --- a/test/CodeGen/X86/vec_ss_load_fold.ll +++ b/test/CodeGen/X86/vec_ss_load_fold.ll @@ -45,3 +45,28 @@ declare <4 x float> @llvm.x86.sse.min.ss(<4 x float>, <4 x float>) declare <4 x float> @llvm.x86.sse.max.ss(<4 x float>, <4 x float>) declare i32 @llvm.x86.sse.cvttss2si(<4 x float>) + + +declare <4 x float> @llvm.x86.sse41.round.ss(<4 x float>, <4 x float>, i32) +declare <4 x float> @f() + +define <4 x float> @test3(<4 x float> %A, float *%b, i32 %C) nounwind { + %a = load float *%b + %B = insertelement <4 x float> undef, float %a, i32 0 + %X = call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> %A, <4 x float> %B, i32 4) + ret <4 x float> %X +; CHECK: test3: +; CHECK: roundss $4, (%eax), %xmm0 +} + +define <4 x float> @test4(<4 x float> %A, float *%b, i32 %C) nounwind { + %a = load float *%b + %B = insertelement <4 x float> undef, float %a, i32 0 + %q = call <4 x float> @f() + %X = call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> %q, <4 x float> %B, i32 4) + ret <4 x float> %X +; CHECK: test4: +; CHECK: movss (%eax), %xmm +; CHECK: call +; CHECK: roundss $4, %xmm{{.*}}, %xmm0 +}
\ No newline at end of file |