1 ; RUN: llc < %s -march=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck %s
2 ; RUN: llc < %s -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs | FileCheck %s
4 ; CHECK: {{^}}test_8_min_char:
5 ; CHECK: buffer_store_byte
6 ; CHECK: buffer_store_byte
7 ; CHECK: buffer_store_byte
8 ; CHECK: buffer_store_byte
9 ; CHECK: buffer_store_byte
10 ; CHECK: buffer_store_byte
11 ; CHECK: buffer_store_byte
12 ; CHECK: buffer_store_byte
15 define amdgpu_kernel void @test_8_min_char(i8 addrspace(1)* nocapture %out, i8 addrspace(1)* nocapture readonly %in0, i8 addrspace(1)* nocapture readonly %in1) #0 {
17 %0 = load i8, i8 addrspace(1)* %in0, align 1
18 %1 = insertelement <8 x i8> undef, i8 %0, i32 0
19 %arrayidx2.i.i = getelementptr inbounds i8, i8 addrspace(1)* %in0, i64 1
20 %2 = load i8, i8 addrspace(1)* %arrayidx2.i.i, align 1
21 %3 = insertelement <8 x i8> %1, i8 %2, i32 1
22 %arrayidx6.i.i = getelementptr inbounds i8, i8 addrspace(1)* %in0, i64 2
23 %4 = load i8, i8 addrspace(1)* %arrayidx6.i.i, align 1
24 %5 = insertelement <8 x i8> %3, i8 %4, i32 2
25 %arrayidx10.i.i = getelementptr inbounds i8, i8 addrspace(1)* %in0, i64 3
26 %6 = load i8, i8 addrspace(1)* %arrayidx10.i.i, align 1
27 %7 = insertelement <8 x i8> %5, i8 %6, i32 3
28 %arrayidx.i.i = getelementptr inbounds i8, i8 addrspace(1)* %in0, i64 4
29 %8 = load i8, i8 addrspace(1)* %arrayidx.i.i, align 1
30 %9 = insertelement <8 x i8> undef, i8 %8, i32 0
31 %arrayidx2.i9.i = getelementptr inbounds i8, i8 addrspace(1)* %in0, i64 5
32 %10 = load i8, i8 addrspace(1)* %arrayidx2.i9.i, align 1
33 %11 = insertelement <8 x i8> %9, i8 %10, i32 1
34 %arrayidx6.i11.i = getelementptr inbounds i8, i8 addrspace(1)* %in0, i64 6
35 %12 = load i8, i8 addrspace(1)* %arrayidx6.i11.i, align 1
36 %13 = insertelement <8 x i8> %11, i8 %12, i32 2
37 %arrayidx10.i13.i = getelementptr inbounds i8, i8 addrspace(1)* %in0, i64 7
38 %14 = load i8, i8 addrspace(1)* %arrayidx10.i13.i, align 1
39 %15 = insertelement <8 x i8> %13, i8 %14, i32 3
40 %vecinit5.i = shufflevector <8 x i8> %7, <8 x i8> %15, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
41 %16 = load i8, i8 addrspace(1)* %in1, align 1
42 %17 = insertelement <8 x i8> undef, i8 %16, i32 0
43 %arrayidx2.i.i4 = getelementptr inbounds i8, i8 addrspace(1)* %in1, i64 1
44 %18 = load i8, i8 addrspace(1)* %arrayidx2.i.i4, align 1
45 %19 = insertelement <8 x i8> %17, i8 %18, i32 1
46 %arrayidx6.i.i5 = getelementptr inbounds i8, i8 addrspace(1)* %in1, i64 2
47 %20 = load i8, i8 addrspace(1)* %arrayidx6.i.i5, align 1
48 %21 = insertelement <8 x i8> %19, i8 %20, i32 2
49 %arrayidx10.i.i6 = getelementptr inbounds i8, i8 addrspace(1)* %in1, i64 3
50 %22 = load i8, i8 addrspace(1)* %arrayidx10.i.i6, align 1
51 %23 = insertelement <8 x i8> %21, i8 %22, i32 3
52 %arrayidx.i.i7 = getelementptr inbounds i8, i8 addrspace(1)* %in1, i64 4
53 %24 = load i8, i8 addrspace(1)* %arrayidx.i.i7, align 1
54 %25 = insertelement <8 x i8> undef, i8 %24, i32 0
55 %arrayidx2.i9.i8 = getelementptr inbounds i8, i8 addrspace(1)* %in1, i64 5
56 %26 = load i8, i8 addrspace(1)* %arrayidx2.i9.i8, align 1
57 %27 = insertelement <8 x i8> %25, i8 %26, i32 1
58 %arrayidx6.i11.i9 = getelementptr inbounds i8, i8 addrspace(1)* %in1, i64 6
59 %28 = load i8, i8 addrspace(1)* %arrayidx6.i11.i9, align 1
60 %29 = insertelement <8 x i8> %27, i8 %28, i32 2
61 %arrayidx10.i13.i10 = getelementptr inbounds i8, i8 addrspace(1)* %in1, i64 7
62 %30 = load i8, i8 addrspace(1)* %arrayidx10.i13.i10, align 1
63 %31 = insertelement <8 x i8> %29, i8 %30, i32 3
64 %vecinit5.i11 = shufflevector <8 x i8> %23, <8 x i8> %31, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
65 %cmp.i = icmp slt <8 x i8> %vecinit5.i, %vecinit5.i11
66 %cond.i = select <8 x i1> %cmp.i, <8 x i8> %vecinit5.i, <8 x i8> %vecinit5.i11
67 %32 = extractelement <8 x i8> %cond.i, i32 0
68 store i8 %32, i8 addrspace(1)* %out, align 1
69 %33 = extractelement <8 x i8> %cond.i, i32 1
70 %arrayidx2.i.i.i = getelementptr inbounds i8, i8 addrspace(1)* %out, i64 1
71 store i8 %33, i8 addrspace(1)* %arrayidx2.i.i.i, align 1
72 %34 = extractelement <8 x i8> %cond.i, i32 2
73 %arrayidx.i.i.i = getelementptr inbounds i8, i8 addrspace(1)* %out, i64 2
74 store i8 %34, i8 addrspace(1)* %arrayidx.i.i.i, align 1
75 %35 = extractelement <8 x i8> %cond.i, i32 3
76 %arrayidx2.i6.i.i = getelementptr inbounds i8, i8 addrspace(1)* %out, i64 3
77 store i8 %35, i8 addrspace(1)* %arrayidx2.i6.i.i, align 1
78 %arrayidx.i.i3 = getelementptr inbounds i8, i8 addrspace(1)* %out, i64 4
79 %36 = extractelement <8 x i8> %cond.i, i32 4
80 store i8 %36, i8 addrspace(1)* %arrayidx.i.i3, align 1
81 %37 = extractelement <8 x i8> %cond.i, i32 5
82 %arrayidx2.i.i6.i = getelementptr inbounds i8, i8 addrspace(1)* %out, i64 5
83 store i8 %37, i8 addrspace(1)* %arrayidx2.i.i6.i, align 1
84 %38 = extractelement <8 x i8> %cond.i, i32 6
85 %arrayidx.i.i7.i = getelementptr inbounds i8, i8 addrspace(1)* %out, i64 6
86 store i8 %38, i8 addrspace(1)* %arrayidx.i.i7.i, align 1
87 %39 = extractelement <8 x i8> %cond.i, i32 7
88 %arrayidx2.i6.i8.i = getelementptr inbounds i8, i8 addrspace(1)* %out, i64 7
89 store i8 %39, i8 addrspace(1)* %arrayidx2.i6.i8.i, align 1
93 attributes #0 = { nounwind }
95 !opencl.kernels = !{!0, !1, !2, !3, !4, !5, !6, !7, !8}
100 !3 = !{void (i8 addrspace(1)*, i8 addrspace(1)*, i8 addrspace(1)*)* @test_8_min_char}