1;RUN: llc < %s -march=r600 -mcpu=verde | FileCheck %s
2
3;CHECK-DAG: IMAGE_LOAD_MIP {{VGPR[0-9]+_VGPR[0-9]+_VGPR[0-9]+_VGPR[0-9]+}}, 15, 0, 0, -1
4;CHECK-DAG: IMAGE_LOAD_MIP {{VGPR[0-9]+_VGPR[0-9]+}}, 3, 0, 0, 0
5;CHECK-DAG: IMAGE_LOAD_MIP {{VGPR[0-9]+}}, 2, 0, 0, 0
6;CHECK-DAG: IMAGE_LOAD_MIP {{VGPR[0-9]+}}, 1, 0, 0, 0
7;CHECK-DAG: IMAGE_LOAD_MIP {{VGPR[0-9]+}}, 4, 0, 0, 0
8;CHECK-DAG: IMAGE_LOAD_MIP {{VGPR[0-9]+}}, 8, 0, 0, 0
9;CHECK-DAG: IMAGE_LOAD_MIP {{VGPR[0-9]+_VGPR[0-9]+}}, 5, 0, 0, 0
10;CHECK-DAG: IMAGE_LOAD_MIP {{VGPR[0-9]+_VGPR[0-9]+}}, 12, 0, 0, -1
11;CHECK-DAG: IMAGE_LOAD_MIP {{VGPR[0-9]+_VGPR[0-9]+_VGPR[0-9]+}}, 7, 0, 0, 0
12;CHECK-DAG: IMAGE_LOAD_MIP {{VGPR[0-9]+}}, 8, 0, 0, -1
13
14define 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      <8 x i32> undef, i32 1)
27   %res2 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v2,
28      <8 x i32> undef, i32 2)
29   %res3 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v3,
30      <8 x i32> undef, i32 3)
31   %res4 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v4,
32      <8 x i32> undef, i32 4)
33   %res5 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v5,
34      <8 x i32> undef, i32 5)
35   %res6 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v6,
36      <8 x i32> undef, i32 6)
37   %res10 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v10,
38      <8 x i32> undef, i32 10)
39   %res11 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v11,
40      <8 x i32> undef, i32 11)
41   %res15 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v15,
42      <8 x i32> undef, i32 15)
43   %res16 = call <4 x i32> @llvm.SI.imageload.(<4 x i32> %v16,
44      <8 x i32> 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)
82   ret void
83}
84
85declare <4 x i32> @llvm.SI.imageload.(<4 x i32>, <8 x i32>, i32) readnone
86
87declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)
88