1; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX9 %s
2; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX10 %s
3
4; GCN-LABEL: {{^}}load.f32.1d:
5; GFX9: image_load v0, v0, s[0:7] dmask:0x1 unorm a16
6; GFX10: image_load v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm a16
7define amdgpu_ps <4 x float> @load.f32.1d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
8main_body:
9  %x = extractelement <2 x i16> %coords, i32 0
10  %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i16(i32 1, i16 %x, <8 x i32> %rsrc, i32 0, i32 0)
11  ret <4 x float> %v
12}
13
14; GCN-LABEL: {{^}}load.v2f32.1d:
15; GFX9: image_load v[0:1], v0, s[0:7] dmask:0x3 unorm a16
16; GFX10: image_load v[0:1], v0, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm a16
17define amdgpu_ps <4 x float> @load.v2f32.1d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
18main_body:
19  %x = extractelement <2 x i16> %coords, i32 0
20  %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i16(i32 3, i16 %x, <8 x i32> %rsrc, i32 0, i32 0)
21  ret <4 x float> %v
22}
23
24; GCN-LABEL: {{^}}load.v3f32.1d:
25; GFX9: image_load v[0:2], v0, s[0:7] dmask:0x7 unorm a16
26; GFX10: image_load v[0:2], v0, s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_1D unorm a16
27define amdgpu_ps <4 x float> @load.v3f32.1d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
28main_body:
29  %x = extractelement <2 x i16> %coords, i32 0
30  %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i16(i32 7, i16 %x, <8 x i32> %rsrc, i32 0, i32 0)
31  ret <4 x float> %v
32}
33
34; GCN-LABEL: {{^}}load.v4f32.1d:
35; GFX9: image_load v[0:3], v0, s[0:7] dmask:0xf unorm a16
36; GFX10: image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16
37define amdgpu_ps <4 x float> @load.v4f32.1d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
38main_body:
39  %x = extractelement <2 x i16> %coords, i32 0
40  %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i16(i32 15, i16 %x, <8 x i32> %rsrc, i32 0, i32 0)
41  ret <4 x float> %v
42}
43
44; GCN-LABEL: {{^}}load.f32.2d:
45; GFX9: image_load v0, v0, s[0:7] dmask:0x1 unorm a16
46; GFX10: image_load v0, v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D unorm a16
47define amdgpu_ps <4 x float> @load.f32.2d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
48main_body:
49  %x = extractelement <2 x i16> %coords, i32 0
50  %y = extractelement <2 x i16> %coords, i32 1
51  %v = call <4 x float> @llvm.amdgcn.image.load.2d.v4f32.i16(i32 1, i16 %x, i16 %y, <8 x i32> %rsrc, i32 0, i32 0)
52  ret <4 x float> %v
53}
54
55; GCN-LABEL: {{^}}load.v2f32.2d:
56; GFX9: image_load v[0:1], v0, s[0:7] dmask:0x3 unorm a16
57; GFX10: image_load v[0:1], v0, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D unorm a16
58define amdgpu_ps <4 x float> @load.v2f32.2d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
59main_body:
60  %x = extractelement <2 x i16> %coords, i32 0
61  %y = extractelement <2 x i16> %coords, i32 1
62  %v = call <4 x float> @llvm.amdgcn.image.load.2d.v4f32.i16(i32 3, i16 %x, i16 %y, <8 x i32> %rsrc, i32 0, i32 0)
63  ret <4 x float> %v
64}
65
66; GCN-LABEL: {{^}}load.v3f32.2d:
67; GFX9: image_load v[0:2], v0, s[0:7] dmask:0x7 unorm a16
68; GFX10: image_load v[0:2], v0, s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_2D unorm a16
69define amdgpu_ps <4 x float> @load.v3f32.2d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
70main_body:
71  %x = extractelement <2 x i16> %coords, i32 0
72  %y = extractelement <2 x i16> %coords, i32 1
73  %v = call <4 x float> @llvm.amdgcn.image.load.2d.v4f32.i16(i32 7, i16 %x, i16 %y, <8 x i32> %rsrc, i32 0, i32 0)
74  ret <4 x float> %v
75}
76
77; GCN-LABEL: {{^}}load.v4f32.2d:
78; GFX9: image_load v[0:3], v0, s[0:7] dmask:0xf unorm a16
79; GFX10: image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16
80define amdgpu_ps <4 x float> @load.v4f32.2d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
81main_body:
82  %x = extractelement <2 x i16> %coords, i32 0
83  %y = extractelement <2 x i16> %coords, i32 1
84  %v = call <4 x float> @llvm.amdgcn.image.load.2d.v4f32.i16(i32 15, i16 %x, i16 %y, <8 x i32> %rsrc, i32 0, i32 0)
85  ret <4 x float> %v
86}
87
88; GCN-LABEL: {{^}}load.f32.3d:
89; GFX9: image_load v0, v[0:1], s[0:7] dmask:0x1 unorm a16
90; GFX10: image_load v0, v[0:1], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_3D unorm a16
91define amdgpu_ps <4 x float> @load.f32.3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
92main_body:
93  %x = extractelement <2 x i16> %coords_lo, i32 0
94  %y = extractelement <2 x i16> %coords_lo, i32 1
95  %z = extractelement <2 x i16> %coords_hi, i32 0
96  %v = call <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i16(i32 1, i16 %x, i16 %y, i16 %z, <8 x i32> %rsrc, i32 0, i32 0)
97  ret <4 x float> %v
98}
99
100; GCN-LABEL: {{^}}load.v2f32.3d:
101; GFX9: image_load v[0:1], v[0:1], s[0:7] dmask:0x3 unorm a16
102; GFX10: image_load v[0:1], v[0:1], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_3D unorm a16
103define amdgpu_ps <4 x float> @load.v2f32.3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
104main_body:
105  %x = extractelement <2 x i16> %coords_lo, i32 0
106  %y = extractelement <2 x i16> %coords_lo, i32 1
107  %z = extractelement <2 x i16> %coords_hi, i32 0
108  %v = call <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i16(i32 3, i16 %x, i16 %y, i16 %z, <8 x i32> %rsrc, i32 0, i32 0)
109  ret <4 x float> %v
110}
111
112; GCN-LABEL: {{^}}load.v3f32.3d:
113; GFX9: image_load v[0:2], v[0:1], s[0:7] dmask:0x7 unorm a16
114; GFX10: image_load v[0:2], v[0:1], s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_3D unorm a16
115define amdgpu_ps <4 x float> @load.v3f32.3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
116main_body:
117  %x = extractelement <2 x i16> %coords_lo, i32 0
118  %y = extractelement <2 x i16> %coords_lo, i32 1
119  %z = extractelement <2 x i16> %coords_hi, i32 0
120  %v = call <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i16(i32 7, i16 %x, i16 %y, i16 %z, <8 x i32> %rsrc, i32 0, i32 0)
121  ret <4 x float> %v
122}
123
124; GCN-LABEL: {{^}}load.v4f32.3d:
125; GFX9: image_load v[0:3], v[0:1], s[0:7] dmask:0xf unorm a16
126; GFX10: image_load v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16
127define amdgpu_ps <4 x float> @load.v4f32.3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo, <2 x i16> %coords_hi) {
128main_body:
129  %x = extractelement <2 x i16> %coords_lo, i32 0
130  %y = extractelement <2 x i16> %coords_lo, i32 1
131  %z = extractelement <2 x i16> %coords_hi, i32 0
132  %v = call <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i16(i32 15, i16 %x, i16 %y, i16 %z, <8 x i32> %rsrc, i32 0, i32 0)
133  ret <4 x float> %v
134}
135
136declare <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i16(i32, i16, <8 x i32>, i32, i32) #2
137declare <4 x float> @llvm.amdgcn.image.load.2d.v4f32.i16(i32, i16, i16, <8 x i32>, i32, i32) #2
138declare <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i16(i32, i16, i16, i16, <8 x i32>, i32, i32) #2
139
140attributes #0 = { nounwind }
141attributes #1 = { nounwind readonly }
142