summaryrefslogtreecommitdiffstats
path: root/llvm/test/CodeGen/X86/vec_ss_load_fold.ll
diff options
context:
space:
mode:
authorChris Lattner <sabre@nondot.org>2010-02-16 22:35:06 +0000
committerChris Lattner <sabre@nondot.org>2010-02-16 22:35:06 +0000
commitafac7dad216d38676ca0e9909d5a4e7588410b64 (patch)
treee6dc46988cd3cba40fa4cd091fe5ace2a59822d2 /llvm/test/CodeGen/X86/vec_ss_load_fold.ll
parent0062f7bf599993d5766c70d6cf56476a4e8166b1 (diff)
downloadbcm5719-llvm-afac7dad216d38676ca0e9909d5a4e7588410b64.tar.gz
bcm5719-llvm-afac7dad216d38676ca0e9909d5a4e7588410b64.zip
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. llvm-svn: 96408
Diffstat (limited to 'llvm/test/CodeGen/X86/vec_ss_load_fold.ll')
-rw-r--r--llvm/test/CodeGen/X86/vec_ss_load_fold.ll25
1 files changed, 25 insertions, 0 deletions
diff --git a/llvm/test/CodeGen/X86/vec_ss_load_fold.ll b/llvm/test/CodeGen/X86/vec_ss_load_fold.ll
index 42831f00225..a44e08dc84a 100644
--- a/llvm/test/CodeGen/X86/vec_ss_load_fold.ll
+++ b/llvm/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
OpenPOWER on IntegriCloud