1 ; RUN: llc -mtriple=i686-unknown-unknown -mattr=+sse4.1 < %s | FileCheck %s -check-prefix=X32
2 ; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 < %s | FileCheck %s -check-prefix=X64
4 ; Test for case where insertps was folding the load of the insertion element, but a later optimization
5 ; was then manipulating the load.
7 define <4 x float> @insertps_unfold(ptr %v0, ptr %v1) {
8 ; X32-LABEL: insertps_unfold:
10 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
11 ; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx
12 ; X32-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
13 ; X32-NEXT: movaps (%eax), %xmm0
14 ; X32-NEXT: insertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
15 ; X32-NEXT: addps %xmm1, %xmm0
18 ; X64-LABEL: insertps_unfold:
20 ; X64-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
21 ; X64-NEXT: movaps (%rdi), %xmm0
22 ; X64-NEXT: insertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]
23 ; X64-NEXT: addps %xmm1, %xmm0
25 %a = getelementptr inbounds <4 x float>, ptr %v1, i64 0, i64 1
26 %b = load float, ptr %a, align 4
27 %c = insertelement <4 x float> undef, float %b, i32 0
28 %d = load <4 x float>, ptr %v1, align 16
29 %e = load <4 x float>, ptr %v0, align 16
30 %f = shufflevector <4 x float> %e, <4 x float> %d, <4 x i32> <i32 0, i32 1, i32 2, i32 5>
31 %g = fadd <4 x float> %c, %f