1;RUN: llc < %s -march=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck %s
2;RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck %s
3
4;CHECK-DAG: image_load {{v\[[0-9]+:[0-9]+\]}}, 15, 0, 0, -1
5;CHECK-DAG: image_load_mip {{v\[[0-9]+:[0-9]+\]}}, 3, 0, 0, 0
6;CHECK-DAG: image_load_mip {{v[0-9]+}}, 2, 0, 0, 0
7;CHECK-DAG: image_load_mip {{v[0-9]+}}, 1, 0, 0, 0
8;CHECK-DAG: image_load_mip {{v[0-9]+}}, 4, 0, 0, 0
9;CHECK-DAG: image_load_mip {{v[0-9]+}}, 8, 0, 0, 0
10;CHECK-DAG: image_load_mip {{v\[[0-9]+:[0-9]+\]}}, 5, 0, 0, 0
11;CHECK-DAG: image_load_mip {{v\[[0-9]+:[0-9]+\]}}, 12, 0, 0, -1
12;CHECK-DAG: image_load_mip {{v\[[0-9]+:[0-9]+\]}}, 7, 0, 0, 0
13;CHECK-DAG: image_load_mip {{v[0-9]+}}, 8, 0, 0, -1
14
15define void @test(i32 %a1, i32 %a2, i32 %a3, i32 %a4) {
16   %v1 = insertelement <4 x i32> undef, i32 %a1, i32 0
17   %v2 = insertelement <4 x i32> undef, i32 %a1, i32 1
18   %v3 = insertelement <4 x i32> undef, i32 %a1, i32 2
19   %v4 = insertelement <4 x i32> undef, i32 %a1, i32 3
20   %v5 = insertelement <4 x i32> undef, i32 %a2, i32 0
21   %v6 = insertelement <4 x i32> undef, i32 %a2, i32 1
22   %v10 = insertelement <4 x i32> undef, i32 %a3, i32 1
23   %v11 = insertelement <4 x i32> undef, i32 %a3, i32 2
24   %v15 = insertelement <4 x i32> undef, i32 %a4, i32 2
25   %v16 = insertelement <4 x i32> undef, i32 %a4, i32 3
26   %res1 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v1,
27      <32 x i8> undef, i32 1)
28   %res2 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v2,
29      <32 x i8> undef, i32 2)
30   %res3 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v3,
31      <32 x i8> undef, i32 3)
32   %res4 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v4,
33      <32 x i8> undef, i32 4)
34   %res5 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v5,
35      <32 x i8> undef, i32 5)
36   %res6 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v6,
37      <32 x i8> undef, i32 6)
38   %res10 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v10,
39      <32 x i8> undef, i32 10)
40   %res11 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v11,
41      <32 x i8> undef, i32 11)
42   %res15 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v15,
43      <32 x i8> undef, i32 15)
44   %res16 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v16,
45      <32 x i8> undef, i32 16)
46   %e1 = extractelement <4 x i32> %res1, i32 0
47   %e2 = extractelement <4 x i32> %res2, i32 1
48   %e3 = extractelement <4 x i32> %res3, i32 2
49   %e4 = extractelement <4 x i32> %res4, i32 3
50   %t0 = extractelement <4 x i32> %res5, i32 0
51   %t1 = extractelement <4 x i32> %res5, i32 1
52   %e5 = add i32 %t0, %t1
53   %t2 = extractelement <4 x i32> %res6, i32 0
54   %t3 = extractelement <4 x i32> %res6, i32 2
55   %e6 = add i32 %t2, %t3
56   %t10 = extractelement <4 x i32> %res10, i32 2
57   %t11 = extractelement <4 x i32> %res10, i32 3
58   %e10 = add i32 %t10, %t11
59   %t12 = extractelement <4 x i32> %res11, i32 0
60   %t13 = extractelement <4 x i32> %res11, i32 1
61   %t14 = extractelement <4 x i32> %res11, i32 2
62   %t15 = add i32 %t12, %t13
63   %e11 = add i32 %t14, %t15
64   %t28 = extractelement <4 x i32> %res15, i32 0
65   %t29 = extractelement <4 x i32> %res15, i32 1
66   %t30 = extractelement <4 x i32> %res15, i32 2
67   %t31 = extractelement <4 x i32> %res15, i32 3
68   %t32 = add i32 %t28, %t29
69   %t33 = add i32 %t30, %t31
70   %e15 = add i32 %t32, %t33
71   %e16 = extractelement <4 x i32> %res16, i32 3
72   %s1 = add i32 %e1, %e2
73   %s2 = add i32 %s1, %e3
74   %s3 = add i32 %s2, %e4
75   %s4 = add i32 %s3, %e5
76   %s5 = add i32 %s4, %e6
77   %s9 = add i32 %s5, %e10
78   %s10 = add i32 %s9, %e11
79   %s14 = add i32 %s10, %e15
80   %s15 = add i32 %s14, %e16
81   %s16 = bitcast i32 %s15 to float
82   call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %s16, float %s16, float %s16, float %s16)
83   ret void
84}
85
86; Test that ccordinates are stored in vgprs and not sgprs
87; CHECK: vgpr_coords
88; CHECK: image_load_mip {{v\[[0-9]+:[0-9]+\]}}, 15, 0, 0, 0, 0, 0, 0, 0, {{v\[[0-9]+:[0-9]+\]}}
89define void @vgpr_coords(float addrspace(2)* addrspace(2)* inreg, <16 x i8> addrspace(2)* inreg, <32 x i8> addrspace(2)* inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, float, float, float, float) #0 {
90main_body:
91  %20 = getelementptr float addrspace(2)*, float addrspace(2)* addrspace(2)* %0, i32 0
92  %21 = load float addrspace(2)*, float addrspace(2)* addrspace(2)* %20, !tbaa !2
93  %22 = getelementptr float, float addrspace(2)* %21, i32 0
94  %23 = load float, float addrspace(2)* %22, !tbaa !2, !invariant.load !1
95  %24 = getelementptr float, float addrspace(2)* %21, i32 1
96  %25 = load float, float addrspace(2)* %24, !tbaa !2, !invariant.load !1
97  %26 = getelementptr float, float addrspace(2)* %21, i32 4
98  %27 = load float, float addrspace(2)* %26, !tbaa !2, !invariant.load !1
99  %28 = getelementptr <32 x i8>, <32 x i8> addrspace(2)* %2, i32 0
100  %29 = load <32 x i8>, <32 x i8> addrspace(2)* %28, !tbaa !2
101  %30 = bitcast float %27 to i32
102  %31 = bitcast float %23 to i32
103  %32 = bitcast float %25 to i32
104  %33 = insertelement <4 x i32> undef, i32 %31, i32 0
105  %34 = insertelement <4 x i32> %33, i32 %32, i32 1
106  %35 = insertelement <4 x i32> %34, i32 %30, i32 2
107  %36 = insertelement <4 x i32> %35, i32 undef, i32 3
108  %37 = call <4 x i32> @llvm.SI.imageload.v4i32(<4 x i32> %36, <32 x i8> %29, i32 2)
109  %38 = extractelement <4 x i32> %37, i32 0
110  %39 = extractelement <4 x i32> %37, i32 1
111  %40 = extractelement <4 x i32> %37, i32 2
112  %41 = extractelement <4 x i32> %37, i32 3
113  %42 = bitcast i32 %38 to float
114  %43 = bitcast i32 %39 to float
115  %44 = bitcast i32 %40 to float
116  %45 = bitcast i32 %41 to float
117  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 0, float %42, float %43, float %44, float %45)
118  ret void
119}
120
121declare <4 x i32> @llvm.SI.imageload.(<4 x i32>, <32 x i8>, i32) readnone
122; Function Attrs: nounwind readnone
123declare <4 x i32> @llvm.SI.imageload.v4i32(<4 x i32>, <32 x i8>, i32) #1
124
125declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
126
127attributes #0 = { "ShaderType"="0" }
128attributes #1 = { nounwind readnone }
129
130!0 = !{!"const", null}
131!1 = !{}
132!2 = !{!0, !0, i64 0, i32 1}
133