1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=X86
3 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=X64
5 %v4_varying_complex = type { <4 x float>, <4 x float> }
7 define void @FFT(ptr noalias nocapture %destination, ptr noalias %re, ptr noalias nocapture %ptr_cast_for_load) nounwind {
9 ; X86: # %bb.0: # %begin
10 ; X86-NEXT: pushl %ebx
11 ; X86-NEXT: pushl %edi
12 ; X86-NEXT: pushl %esi
13 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
14 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
15 ; X86-NEXT: movl {{[0-9]+}}(%esp), %edx
16 ; X86-NEXT: movdqu (%edx), %xmm0
17 ; X86-NEXT: pslld $4, %xmm0
18 ; X86-NEXT: movd %xmm0, %edx
19 ; X86-NEXT: pextrd $1, %xmm0, %esi
20 ; X86-NEXT: pextrd $2, %xmm0, %edi
21 ; X86-NEXT: pextrd $3, %xmm0, %ebx
22 ; X86-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
23 ; X86-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
24 ; X86-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
25 ; X86-NEXT: movss {{.*#+}} xmm3 = mem[0],zero,zero,zero
26 ; X86-NEXT: movss %xmm0, 128(%eax)
27 ; X86-NEXT: movss %xmm1, 164(%eax)
28 ; X86-NEXT: movss %xmm2, 200(%eax)
29 ; X86-NEXT: movss %xmm3, 236(%eax)
36 ; X64: # %bb.0: # %begin
37 ; X64-NEXT: movdqu (%rdx), %xmm0
38 ; X64-NEXT: pslld $4, %xmm0
39 ; X64-NEXT: movd %xmm0, %eax
40 ; X64-NEXT: movslq %eax, %r8
41 ; X64-NEXT: pextrd $1, %xmm0, %ecx
42 ; X64-NEXT: movslq %ecx, %rcx
43 ; X64-NEXT: pextrd $2, %xmm0, %edx
44 ; X64-NEXT: movslq %edx, %rdx
45 ; X64-NEXT: pextrd $3, %xmm0, %eax
47 ; X64-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
48 ; X64-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
49 ; X64-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
50 ; X64-NEXT: movss {{.*#+}} xmm3 = mem[0],zero,zero,zero
51 ; X64-NEXT: movss %xmm0, 128(%rdi)
52 ; X64-NEXT: movss %xmm1, 164(%rdi)
53 ; X64-NEXT: movss %xmm2, 200(%rdi)
54 ; X64-NEXT: movss %xmm3, 236(%rdi)
57 %ptr_masked_load79 = load <4 x i32>, ptr %ptr_cast_for_load, align 4
58 %mul__bitReversedProgramIndex_load = shl <4 x i32> %ptr_masked_load79, <i32 4, i32 4, i32 4, i32 4>
60 %offset32_1 = extractelement <4 x i32> %mul__bitReversedProgramIndex_load, i32 0
61 %ptroffset_1 = sext i32 %offset32_1 to i64
62 %offset32_2 = extractelement <4 x i32> %mul__bitReversedProgramIndex_load, i32 1
63 %ptroffset_2 = sext i32 %offset32_2 to i64
64 %offset32_3 = extractelement <4 x i32> %mul__bitReversedProgramIndex_load, i32 2
65 %ptroffset_3 = sext i32 %offset32_3 to i64
66 %offset32_4 = extractelement <4 x i32> %mul__bitReversedProgramIndex_load, i32 3
67 %ptroffset_4 = sext i32 %offset32_4 to i64
69 %ptrcast_1 = getelementptr float, ptr %re, i64 %ptroffset_1
70 %val_1 = load float, ptr %ptrcast_1, align 4
71 %ptrcast_2 = getelementptr float, ptr %re, i64 %ptroffset_2
72 %val_2 = load float, ptr %ptrcast_2, align 4
73 %ptrcast_3 = getelementptr float, ptr %re, i64 %ptroffset_3
74 %val_3 = load float, ptr %ptrcast_3, align 4
75 %ptrcast_4 = getelementptr float, ptr %re, i64 %ptroffset_4
76 %val_4 = load float, ptr %ptrcast_4, align 4
78 %ptrcast1_1 = getelementptr inbounds %v4_varying_complex, ptr %destination, i64 4, i32 0, i64 0
79 store float %val_1, ptr %ptrcast1_1, align 4
80 %finalptr_2 = getelementptr i8, ptr %destination, i64 164
81 store float %val_2, ptr %finalptr_2, align 4
82 %finalptr_3 = getelementptr i8, ptr %destination, i64 200
83 store float %val_3, ptr %finalptr_3, align 4
84 %finalptr_4 = getelementptr i8, ptr %destination, i64 236
85 store float %val_4, ptr %finalptr_4, align 4