• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=core-avx2 -mattr=+avx | FileCheck %s
2; rdar://11314175: SD Scheduler, BuildSchedUnits assert:
3;                  N->getNodeId() == -1 && "Node already inserted!
4
5; It's hard to test for the ISEL condition because CodeGen optimizes
6; away the bugpointed code. Just ensure the basics are still there.
7;CHECK-LABEL: func:
8;CHECK: vxorps
9;CHECK: vpshufd
10;CHECK: vpbroadcastd
11;CHECK: vinserti128
12;CHECK: vmulps
13;CHECK: vmulps
14;CHECK: ret
15
16define void @func() nounwind ssp {
17  %tmp = load <4 x float>, <4 x float>* null, align 1
18  %tmp14 = getelementptr <4 x float>, <4 x float>* null, i32 2
19  %tmp15 = load <4 x float>, <4 x float>* %tmp14, align 1
20  %tmp16 = shufflevector <4 x float> %tmp, <4 x float> <float 0.000000e+00, float undef, float undef, float undef>, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 4, i32 4, i32 4>
21  %tmp17 = call <8 x float> @llvm.x86.avx.vinsertf128.ps.256(<8 x float> %tmp16, <4 x float> undef, i8 1)
22  %tmp18 = bitcast <4 x float> %tmp to <16 x i8>
23  %tmp19 = shufflevector <16 x i8> %tmp18, <16 x i8> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
24  %tmp20 = bitcast <16 x i8> %tmp19 to <4 x float>
25  %tmp21 = bitcast <4 x float> %tmp15 to <16 x i8>
26  %tmp22 = shufflevector <16 x i8> undef, <16 x i8> %tmp21, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19>
27  %tmp23 = bitcast <16 x i8> %tmp22 to <4 x float>
28  %tmp24 = shufflevector <4 x float> %tmp20, <4 x float> <float 0.000000e+00, float undef, float undef, float undef>, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 4, i32 4, i32 4>
29  %tmp25 = call <8 x float> @llvm.x86.avx.vinsertf128.ps.256(<8 x float> %tmp24, <4 x float> %tmp23, i8 1)
30  %tmp26 = fmul <8 x float> %tmp17, undef
31  %tmp27 = fmul <8 x float> %tmp25, undef
32  %tmp28 = fadd <8 x float> %tmp26, %tmp27
33  %tmp29 = fadd <8 x float> %tmp28, undef
34  %tmp30 = shufflevector <8 x float> %tmp29, <8 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
35  %tmp31 = fmul <4 x float> undef, %tmp30
36  %tmp32 = call <8 x float> @llvm.x86.avx.vinsertf128.ps.256(<8 x float> zeroinitializer, <4 x float> %tmp31, i8 1)
37  %tmp33 = fadd <8 x float> undef, %tmp32
38  %tmp34 = call <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> %tmp33, <8 x float> undef) nounwind
39  %tmp35 = fsub <8 x float> %tmp34, undef
40  %tmp36 = call <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> zeroinitializer, <8 x float> %tmp35) nounwind
41  store <8 x float> %tmp36, <8 x float>* undef, align 32
42  ret void
43}
44
45declare <8 x float> @llvm.x86.avx.vinsertf128.ps.256(<8 x float>, <4 x float>, i8) nounwind readnone
46
47declare <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float>, <8 x float>) nounwind readnone
48