1 ;RUN: llc < %s -march=r600 -mcpu=verde -verify-machineinstrs | FileCheck %s
3 ;CHECK-DAG: IMAGE_LOAD {{v\[[0-9]+:[0-9]+\]}}, 15, 0, 0, -1
4 ;CHECK-DAG: IMAGE_LOAD_MIP {{v\[[0-9]+:[0-9]+\]}}, 3, 0, 0, 0
5 ;CHECK-DAG: IMAGE_LOAD_MIP {{v[0-9]+}}, 2, 0, 0, 0
6 ;CHECK-DAG: IMAGE_LOAD_MIP {{v[0-9]+}}, 1, 0, 0, 0
7 ;CHECK-DAG: IMAGE_LOAD_MIP {{v[0-9]+}}, 4, 0, 0, 0
8 ;CHECK-DAG: IMAGE_LOAD_MIP {{v[0-9]+}}, 8, 0, 0, 0
9 ;CHECK-DAG: IMAGE_LOAD_MIP {{v\[[0-9]+:[0-9]+\]}}, 5, 0, 0, 0
10 ;CHECK-DAG: IMAGE_LOAD_MIP {{v\[[0-9]+:[0-9]+\]}}, 12, 0, 0, -1
11 ;CHECK-DAG: IMAGE_LOAD_MIP {{v\[[0-9]+:[0-9]+\]}}, 7, 0, 0, 0
12 ;CHECK-DAG: IMAGE_LOAD_MIP {{v[0-9]+}}, 8, 0, 0, -1
14 define void @test(i32 %a1, i32 %a2, i32 %a3, i32 %a4) {
15 %v1 = insertelement <4 x i32> undef, i32 %a1, i32 0
16 %v2 = insertelement <4 x i32> undef, i32 %a1, i32 1
17 %v3 = insertelement <4 x i32> undef, i32 %a1, i32 2
18 %v4 = insertelement <4 x i32> undef, i32 %a1, i32 3
19 %v5 = insertelement <4 x i32> undef, i32 %a2, i32 0
20 %v6 = insertelement <4 x i32> undef, i32 %a2, i32 1
21 %v10 = insertelement <4 x i32> undef, i32 %a3, i32 1
22 %v11 = insertelement <4 x i32> undef, i32 %a3, i32 2
23 %v15 = insertelement <4 x i32> undef, i32 %a4, i32 2
24 %v16 = insertelement <4 x i32> undef, i32 %a4, i32 3
25 %res1 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v1,
26 <32 x i8> undef, i32 1)
27 %res2 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v2,
28 <32 x i8> undef, i32 2)
29 %res3 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v3,
30 <32 x i8> undef, i32 3)
31 %res4 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v4,
32 <32 x i8> undef, i32 4)
33 %res5 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v5,
34 <32 x i8> undef, i32 5)
35 %res6 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v6,
36 <32 x i8> undef, i32 6)
37 %res10 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v10,
38 <32 x i8> undef, i32 10)
39 %res11 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v11,
40 <32 x i8> undef, i32 11)
41 %res15 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v15,
42 <32 x i8> undef, i32 15)
43 %res16 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v16,
44 <32 x i8> undef, i32 16)
45 %e1 = extractelement <4 x i32> %res1, i32 0
46 %e2 = extractelement <4 x i32> %res2, i32 1
47 %e3 = extractelement <4 x i32> %res3, i32 2
48 %e4 = extractelement <4 x i32> %res4, i32 3
49 %t0 = extractelement <4 x i32> %res5, i32 0
50 %t1 = extractelement <4 x i32> %res5, i32 1
51 %e5 = add i32 %t0, %t1
52 %t2 = extractelement <4 x i32> %res6, i32 0
53 %t3 = extractelement <4 x i32> %res6, i32 2
54 %e6 = add i32 %t2, %t3
55 %t10 = extractelement <4 x i32> %res10, i32 2
56 %t11 = extractelement <4 x i32> %res10, i32 3
57 %e10 = add i32 %t10, %t11
58 %t12 = extractelement <4 x i32> %res11, i32 0
59 %t13 = extractelement <4 x i32> %res11, i32 1
60 %t14 = extractelement <4 x i32> %res11, i32 2
61 %t15 = add i32 %t12, %t13
62 %e11 = add i32 %t14, %t15
63 %t28 = extractelement <4 x i32> %res15, i32 0
64 %t29 = extractelement <4 x i32> %res15, i32 1
65 %t30 = extractelement <4 x i32> %res15, i32 2
66 %t31 = extractelement <4 x i32> %res15, i32 3
67 %t32 = add i32 %t28, %t29
68 %t33 = add i32 %t30, %t31
69 %e15 = add i32 %t32, %t33
70 %e16 = extractelement <4 x i32> %res16, i32 3
71 %s1 = add i32 %e1, %e2
72 %s2 = add i32 %s1, %e3
73 %s3 = add i32 %s2, %e4
74 %s4 = add i32 %s3, %e5
75 %s5 = add i32 %s4, %e6
76 %s9 = add i32 %s5, %e10
77 %s10 = add i32 %s9, %e11
78 %s14 = add i32 %s10, %e15
79 %s15 = add i32 %s14, %e16
80 %s16 = bitcast i32 %s15 to float
81 call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %s16, float %s16, float %s16, float %s16)
85 ; Test that ccordinates are stored in vgprs and not sgprs
87 ; CHECK: IMAGE_LOAD_MIP {{v\[[0-9]+:[0-9]+\]}}, 15, 0, 0, 0, 0, 0, 0, 0, {{v\[[0-9]+:[0-9]+\]}}
88 define 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 {
90 %20 = getelementptr float addrspace(2)* addrspace(2)* %0, i32 0
91 %21 = load float addrspace(2)* addrspace(2)* %20, !tbaa !2
92 %22 = getelementptr float addrspace(2)* %21, i32 0
93 %23 = load float addrspace(2)* %22, !tbaa !2, !invariant.load !1
94 %24 = getelementptr float addrspace(2)* %21, i32 1
95 %25 = load float addrspace(2)* %24, !tbaa !2, !invariant.load !1
96 %26 = getelementptr float addrspace(2)* %21, i32 4
97 %27 = load float addrspace(2)* %26, !tbaa !2, !invariant.load !1
98 %28 = getelementptr <32 x i8> addrspace(2)* %2, i32 0
99 %29 = load <32 x i8> addrspace(2)* %28, !tbaa !2
100 %30 = bitcast float %27 to i32
101 %31 = bitcast float %23 to i32
102 %32 = bitcast float %25 to i32
103 %33 = insertelement <4 x i32> undef, i32 %31, i32 0
104 %34 = insertelement <4 x i32> %33, i32 %32, i32 1
105 %35 = insertelement <4 x i32> %34, i32 %30, i32 2
106 %36 = insertelement <4 x i32> %35, i32 undef, i32 3
107 %37 = call <4 x i32> @llvm.SI.imageload.v4i32(<4 x i32> %36, <32 x i8> %29, i32 2)
108 %38 = extractelement <4 x i32> %37, i32 0
109 %39 = extractelement <4 x i32> %37, i32 1
110 %40 = extractelement <4 x i32> %37, i32 2
111 %41 = extractelement <4 x i32> %37, i32 3
112 %42 = bitcast i32 %38 to float
113 %43 = bitcast i32 %39 to float
114 %44 = bitcast i32 %40 to float
115 %45 = bitcast i32 %41 to float
116 call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 0, float %42, float %43, float %44, float %45)
120 declare <4 x i32> @llvm.SI.imageload.(<4 x i32>, <32 x i8>, i32) readnone
121 ; Function Attrs: nounwind readnone
122 declare <4 x i32> @llvm.SI.imageload.v4i32(<4 x i32>, <32 x i8>, i32) #1
124 declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
126 attributes #0 = { "ShaderType"="0" }
127 attributes #1 = { nounwind readnone }
129 !0 = metadata !{metadata !"const", null}
131 !2 = metadata !{metadata !0, metadata !0, i64 0, i32 1}