1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.2 | FileCheck %s
4 ; widen a v3i32 to v4i32 to do a vector multiple and a subtraction
6 define void @update(<3 x i32>* %dst, <3 x i32>* %src, i32 %n) nounwind {
8 ; CHECK: # %bb.0: # %entry
9 ; CHECK-NEXT: movq %rdi, -{{[0-9]+}}(%rsp)
10 ; CHECK-NEXT: movq %rsi, -{{[0-9]+}}(%rsp)
11 ; CHECK-NEXT: movl %edx, -{{[0-9]+}}(%rsp)
12 ; CHECK-NEXT: movabsq $4294967297, %rax # imm = 0x100000001
13 ; CHECK-NEXT: movq %rax, -{{[0-9]+}}(%rsp)
14 ; CHECK-NEXT: movl $1, -{{[0-9]+}}(%rsp)
15 ; CHECK-NEXT: movl $0, -{{[0-9]+}}(%rsp)
16 ; CHECK-NEXT: movdqa {{.*#+}} xmm0 = <3,3,3,u>
17 ; CHECK-NEXT: .p2align 4, 0x90
18 ; CHECK-NEXT: .LBB0_1: # %forcond
19 ; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
20 ; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
21 ; CHECK-NEXT: cmpl -{{[0-9]+}}(%rsp), %eax
22 ; CHECK-NEXT: jge .LBB0_3
23 ; CHECK-NEXT: # %bb.2: # %forbody
24 ; CHECK-NEXT: # in Loop: Header=BB0_1 Depth=1
25 ; CHECK-NEXT: movslq -{{[0-9]+}}(%rsp), %rax
26 ; CHECK-NEXT: movq -{{[0-9]+}}(%rsp), %rcx
27 ; CHECK-NEXT: shlq $4, %rax
28 ; CHECK-NEXT: movq -{{[0-9]+}}(%rsp), %rdx
29 ; CHECK-NEXT: movdqa (%rdx,%rax), %xmm1
30 ; CHECK-NEXT: pslld $2, %xmm1
31 ; CHECK-NEXT: psubd %xmm0, %xmm1
32 ; CHECK-NEXT: pextrd $2, %xmm1, 8(%rcx,%rax)
33 ; CHECK-NEXT: movq %xmm1, (%rcx,%rax)
34 ; CHECK-NEXT: incl -{{[0-9]+}}(%rsp)
35 ; CHECK-NEXT: jmp .LBB0_1
36 ; CHECK-NEXT: .LBB0_3: # %afterfor
39 %dst.addr = alloca <3 x i32>*
40 %src.addr = alloca <3 x i32>*
42 %v = alloca <3 x i32>, align 16
43 %i = alloca i32, align 4
44 store <3 x i32>* %dst, <3 x i32>** %dst.addr
45 store <3 x i32>* %src, <3 x i32>** %src.addr
46 store i32 %n, i32* %n.addr
47 store <3 x i32> < i32 1, i32 1, i32 1 >, <3 x i32>* %v
52 %tmp = load i32, i32* %i
53 %tmp1 = load i32, i32* %n.addr
54 %cmp = icmp slt i32 %tmp, %tmp1
55 br i1 %cmp, label %forbody, label %afterfor
58 %tmp2 = load i32, i32* %i
59 %tmp3 = load <3 x i32>*, <3 x i32>** %dst.addr
60 %arrayidx = getelementptr <3 x i32>, <3 x i32>* %tmp3, i32 %tmp2
61 %tmp4 = load i32, i32* %i
62 %tmp5 = load <3 x i32>*, <3 x i32>** %src.addr
63 %arrayidx6 = getelementptr <3 x i32>, <3 x i32>* %tmp5, i32 %tmp4
64 %tmp7 = load <3 x i32>, <3 x i32>* %arrayidx6
65 %mul = mul <3 x i32> %tmp7, < i32 4, i32 4, i32 4 >
66 %sub = sub <3 x i32> %mul, < i32 3, i32 3, i32 3 >
67 store <3 x i32> %sub, <3 x i32>* %arrayidx
71 %tmp8 = load i32, i32* %i
72 %inc = add i32 %tmp8, 1
73 store i32 %inc, i32* %i