1;RUN: llc < %s -march=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck %s -check-prefix=CHECK -check-prefix=SICI
2;RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck %s -check-prefix=CHECK -check-prefix=VI
3
4;CHECK-LABEL: {{^}}buffer_load:
5;CHECK: buffer_load_dwordx4 v[0:3], off, s[0:3], 0
6;CHECK: buffer_load_dwordx4 v[4:7], off, s[0:3], 0 glc
7;CHECK: buffer_load_dwordx4 v[8:11], off, s[0:3], 0 slc
8;CHECK: s_waitcnt
9define amdgpu_ps {<4 x float>, <4 x float>, <4 x float>} @buffer_load(<4 x i32> inreg) {
10main_body:
11  %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %0, i32 0, i32 0, i32 0)
12  %data_glc = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %0, i32 0, i32 0, i32 1)
13  %data_slc = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %0, i32 0, i32 0, i32 2)
14  %r0 = insertvalue {<4 x float>, <4 x float>, <4 x float>} undef, <4 x float> %data, 0
15  %r1 = insertvalue {<4 x float>, <4 x float>, <4 x float>} %r0, <4 x float> %data_glc, 1
16  %r2 = insertvalue {<4 x float>, <4 x float>, <4 x float>} %r1, <4 x float> %data_slc, 2
17  ret {<4 x float>, <4 x float>, <4 x float>} %r2
18}
19
20;CHECK-LABEL: {{^}}buffer_load_immoffs:
21;CHECK: buffer_load_dwordx4 v[0:3], off, s[0:3], 0 offset:40
22;CHECK: s_waitcnt
23define amdgpu_ps <4 x float> @buffer_load_immoffs(<4 x i32> inreg) {
24main_body:
25  %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %0, i32 40, i32 0, i32 0)
26  ret <4 x float> %data
27}
28
29;CHECK-LABEL: {{^}}buffer_load_immoffs_large:
30;CHECK: s_movk_i32 [[OFFSET:s[0-9]+]], 0x1ffc
31;CHECK: buffer_load_dwordx4 v[0:3], off, s[0:3], [[OFFSET]] offset:4
32;CHECK: s_waitcnt
33define amdgpu_ps <4 x float> @buffer_load_immoffs_large(<4 x i32> inreg) {
34main_body:
35  %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %0, i32 4, i32 8188, i32 0)
36  ret <4 x float> %data
37}
38
39;CHECK-LABEL: {{^}}buffer_load_ofs:
40;CHECK: buffer_load_dwordx4 v[0:3], v0, s[0:3], 0 offen
41;CHECK: s_waitcnt
42define amdgpu_ps <4 x float> @buffer_load_ofs(<4 x i32> inreg, i32) {
43main_body:
44  %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %0, i32 %1, i32 0, i32 0)
45  ret <4 x float> %data
46}
47
48;CHECK-LABEL: {{^}}buffer_load_ofs_imm:
49;CHECK: buffer_load_dwordx4 v[0:3], v0, s[0:3], 0 offen offset:60
50;CHECK: s_waitcnt
51define amdgpu_ps <4 x float> @buffer_load_ofs_imm(<4 x i32> inreg, i32) {
52main_body:
53  %ofs = add i32 %1, 60
54  %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %0, i32 %ofs, i32 0, i32 0)
55  ret <4 x float> %data
56}
57
58;CHECK-LABEL: {{^}}buffer_load_x1:
59;CHECK: buffer_load_dword v0, v0, s[0:3], 0 offen
60;CHECK: s_waitcnt
61define amdgpu_ps float @buffer_load_x1(<4 x i32> inreg %rsrc, i32 %ofs) {
62main_body:
63  %data = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %ofs, i32 0, i32 0)
64  ret float %data
65}
66
67;CHECK-LABEL: {{^}}buffer_load_x2:
68;CHECK: buffer_load_dwordx2 v[0:1], v0, s[0:3], 0 offen
69;CHECK: s_waitcnt
70define amdgpu_ps <2 x float> @buffer_load_x2(<4 x i32> inreg %rsrc, i32 %ofs) {
71main_body:
72  %data = call <2 x float> @llvm.amdgcn.raw.buffer.load.v2f32(<4 x i32> %rsrc, i32 %ofs, i32 0, i32 0)
73  ret <2 x float> %data
74}
75
76;CHECK-LABEL: {{^}}buffer_load_negative_offset:
77;CHECK: v_add_{{[iu]}}32_e32 [[VOFS:v[0-9]+]], vcc, -16, v0
78;CHECK: buffer_load_dwordx4 v[0:3], [[VOFS]], s[0:3], 0 offen
79define amdgpu_ps <4 x float> @buffer_load_negative_offset(<4 x i32> inreg, i32 %ofs) {
80main_body:
81  %ofs.1 = add i32 %ofs, -16
82  %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %0, i32 %ofs.1, i32 0, i32 0)
83  ret <4 x float> %data
84}
85
86; SI won't merge ds memory operations, because of the signed offset bug, so
87; we only have check lines for VI.
88; CHECK-LABEL: buffer_load_mmo:
89; VI: v_mov_b32_e32 [[ZERO:v[0-9]+]], 0
90; VI: ds_write2_b32 v{{[0-9]+}}, [[ZERO]], [[ZERO]] offset1:4
91define amdgpu_ps float @buffer_load_mmo(<4 x i32> inreg %rsrc, float addrspace(3)* %lds) {
92entry:
93  store float 0.0, float addrspace(3)* %lds
94  %val = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 0, i32 0, i32 0)
95  %tmp2 = getelementptr float, float addrspace(3)* %lds, i32 4
96  store float 0.0, float addrspace(3)* %tmp2
97  ret float %val
98}
99
100;CHECK-LABEL: {{^}}buffer_load_x1_offen_merged_and:
101;CHECK-NEXT: %bb.
102;CHECK-NEXT: buffer_load_dwordx4 v[{{[0-9]}}:{{[0-9]}}], v0, s[0:3], 0 offen offset:4
103;CHECK-NEXT: buffer_load_dwordx2 v[{{[0-9]}}:{{[0-9]}}], v0, s[0:3], 0 offen offset:28
104;CHECK: s_waitcnt
105define amdgpu_ps void @buffer_load_x1_offen_merged_and(<4 x i32> inreg %rsrc, i32 %a) {
106main_body:
107  %a1 = add i32 %a, 4
108  %a2 = add i32 %a, 8
109  %a3 = add i32 %a, 12
110  %a4 = add i32 %a, 16
111  %a5 = add i32 %a, 28
112  %a6 = add i32 %a, 32
113  %r1 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a1, i32 0, i32 0)
114  %r2 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a2, i32 0, i32 0)
115  %r3 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a3, i32 0, i32 0)
116  %r4 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a4, i32 0, i32 0)
117  %r5 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a5, i32 0, i32 0)
118  %r6 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a6, i32 0, i32 0)
119  call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %r1, float %r2, float %r3, float %r4, i1 true, i1 true)
120  call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %r5, float %r6, float undef, float undef, i1 true, i1 true)
121  ret void
122}
123
124;CHECK-LABEL: {{^}}buffer_load_x1_offen_merged_or:
125;CHECK-NEXT: %bb.
126;CHECK-NEXT: v_lshlrev_b32_e32 v{{[0-9]}}, 6, v0
127;CHECK-NEXT: buffer_load_dwordx4 v[{{[0-9]}}:{{[0-9]}}], v{{[0-9]}}, s[0:3], 0 offen offset:4
128;CHECK-NEXT: buffer_load_dwordx2 v[{{[0-9]}}:{{[0-9]}}], v{{[0-9]}}, s[0:3], 0 offen offset:28
129;CHECK: s_waitcnt
130define amdgpu_ps void @buffer_load_x1_offen_merged_or(<4 x i32> inreg %rsrc, i32 %inp) {
131main_body:
132  %a = shl i32 %inp, 6
133  %a1 = or i32 %a, 4
134  %a2 = or i32 %a, 8
135  %a3 = or i32 %a, 12
136  %a4 = or i32 %a, 16
137  %a5 = or i32 %a, 28
138  %a6 = or i32 %a, 32
139  %r1 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a1, i32 0, i32 0)
140  %r2 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a2, i32 0, i32 0)
141  %r3 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a3, i32 0, i32 0)
142  %r4 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a4, i32 0, i32 0)
143  %r5 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a5, i32 0, i32 0)
144  %r6 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a6, i32 0, i32 0)
145  call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %r1, float %r2, float %r3, float %r4, i1 true, i1 true)
146  call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %r5, float %r6, float undef, float undef, i1 true, i1 true)
147  ret void
148}
149
150;CHECK-LABEL: {{^}}buffer_load_x1_offen_merged_glc_slc:
151;CHECK-NEXT: %bb.
152;CHECK-NEXT: buffer_load_dwordx2 v[{{[0-9]}}:{{[0-9]}}], v0, s[0:3], 0 offen offset:4{{$}}
153;CHECK-NEXT: buffer_load_dwordx2 v[{{[0-9]}}:{{[0-9]}}], v0, s[0:3], 0 offen offset:12 glc{{$}}
154;CHECK-NEXT: buffer_load_dwordx2 v[{{[0-9]}}:{{[0-9]}}], v0, s[0:3], 0 offen offset:28 glc slc{{$}}
155;CHECK: s_waitcnt
156define amdgpu_ps void @buffer_load_x1_offen_merged_glc_slc(<4 x i32> inreg %rsrc, i32 %a) {
157main_body:
158  %a1 = add i32 %a, 4
159  %a2 = add i32 %a, 8
160  %a3 = add i32 %a, 12
161  %a4 = add i32 %a, 16
162  %a5 = add i32 %a, 28
163  %a6 = add i32 %a, 32
164  %r1 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a1, i32 0, i32 0)
165  %r2 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a2, i32 0, i32 0)
166  %r3 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a3, i32 0, i32 1)
167  %r4 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a4, i32 0, i32 1)
168  %r5 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a5, i32 0, i32 3)
169  %r6 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %a6, i32 0, i32 3)
170  call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %r1, float %r2, float %r3, float %r4, i1 true, i1 true)
171  call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %r5, float %r6, float undef, float undef, i1 true, i1 true)
172  ret void
173}
174
175;CHECK-LABEL: {{^}}buffer_load_x2_offen_merged_and:
176;CHECK-NEXT: %bb.
177;CHECK-NEXT: buffer_load_dwordx4 v[{{[0-9]}}:{{[0-9]}}], v0, s[0:3], 0 offen offset:4
178;CHECK: s_waitcnt
179define amdgpu_ps void @buffer_load_x2_offen_merged_and(<4 x i32> inreg %rsrc, i32 %a) {
180main_body:
181  %a1 = add i32 %a, 4
182  %a2 = add i32 %a, 12
183  %vr1 = call <2 x float> @llvm.amdgcn.raw.buffer.load.v2f32(<4 x i32> %rsrc, i32 %a1, i32 0, i32 0)
184  %vr2 = call <2 x float> @llvm.amdgcn.raw.buffer.load.v2f32(<4 x i32> %rsrc, i32 %a2, i32 0, i32 0)
185  %r1 = extractelement <2 x float> %vr1, i32 0
186  %r2 = extractelement <2 x float> %vr1, i32 1
187  %r3 = extractelement <2 x float> %vr2, i32 0
188  %r4 = extractelement <2 x float> %vr2, i32 1
189  call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %r1, float %r2, float %r3, float %r4, i1 true, i1 true)
190  ret void
191}
192
193;CHECK-LABEL: {{^}}buffer_load_x2_offen_merged_or:
194;CHECK-NEXT: %bb.
195;CHECK-NEXT: v_lshlrev_b32_e32 v{{[0-9]}}, 4, v0
196;CHECK-NEXT: buffer_load_dwordx4 v[{{[0-9]}}:{{[0-9]}}], v{{[0-9]}}, s[0:3], 0 offen offset:4
197;CHECK: s_waitcnt
198define amdgpu_ps void @buffer_load_x2_offen_merged_or(<4 x i32> inreg %rsrc, i32 %inp) {
199main_body:
200  %a = shl i32 %inp, 4
201  %a1 = add i32 %a, 4
202  %a2 = add i32 %a, 12
203  %vr1 = call <2 x float> @llvm.amdgcn.raw.buffer.load.v2f32(<4 x i32> %rsrc, i32 %a1, i32 0, i32 0)
204  %vr2 = call <2 x float> @llvm.amdgcn.raw.buffer.load.v2f32(<4 x i32> %rsrc, i32 %a2, i32 0, i32 0)
205  %r1 = extractelement <2 x float> %vr1, i32 0
206  %r2 = extractelement <2 x float> %vr1, i32 1
207  %r3 = extractelement <2 x float> %vr2, i32 0
208  %r4 = extractelement <2 x float> %vr2, i32 1
209  call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %r1, float %r2, float %r3, float %r4, i1 true, i1 true)
210  ret void
211}
212
213;CHECK-LABEL: {{^}}buffer_load_x1_offset_merged:
214;CHECK-NEXT: %bb.
215;CHECK-NEXT: buffer_load_dwordx4 v[{{[0-9]}}:{{[0-9]}}], off, s[0:3], 0 offset:4
216;CHECK-NEXT: buffer_load_dwordx2 v[{{[0-9]}}:{{[0-9]}}], off, s[0:3], 0 offset:28
217;CHECK: s_waitcnt
218define amdgpu_ps void @buffer_load_x1_offset_merged(<4 x i32> inreg %rsrc) {
219main_body:
220  %r1 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 4, i32 0, i32 0)
221  %r2 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 8, i32 0, i32 0)
222  %r3 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 12, i32 0, i32 0)
223  %r4 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 16, i32 0, i32 0)
224  %r5 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 28, i32 0, i32 0)
225  %r6 = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 32, i32 0, i32 0)
226  call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %r1, float %r2, float %r3, float %r4, i1 true, i1 true)
227  call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %r5, float %r6, float undef, float undef, i1 true, i1 true)
228  ret void
229}
230
231;CHECK-LABEL: {{^}}buffer_load_x2_offset_merged:
232;CHECK-NEXT: %bb.
233;CHECK-NEXT: buffer_load_dwordx4 v[{{[0-9]}}:{{[0-9]}}], off, s[0:3], 0 offset:4
234;CHECK: s_waitcnt
235define amdgpu_ps void @buffer_load_x2_offset_merged(<4 x i32> inreg %rsrc) {
236main_body:
237  %vr1 = call <2 x float> @llvm.amdgcn.raw.buffer.load.v2f32(<4 x i32> %rsrc, i32 4, i32 0, i32 0)
238  %vr2 = call <2 x float> @llvm.amdgcn.raw.buffer.load.v2f32(<4 x i32> %rsrc, i32 12, i32 0, i32 0)
239  %r1 = extractelement <2 x float> %vr1, i32 0
240  %r2 = extractelement <2 x float> %vr1, i32 1
241  %r3 = extractelement <2 x float> %vr2, i32 0
242  %r4 = extractelement <2 x float> %vr2, i32 1
243  call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %r1, float %r2, float %r3, float %r4, i1 true, i1 true)
244  ret void
245}
246
247;CHECK-LABEL: {{^}}buffer_load_int:
248;CHECK: buffer_load_dwordx4 v[0:3], off, s[0:3], 0
249;CHECK: buffer_load_dwordx2 v[4:5], off, s[0:3], 0 glc
250;CHECK: buffer_load_dword v6, off, s[0:3], 0 slc
251;CHECK: s_waitcnt
252define amdgpu_ps {<4 x float>, <2 x float>, float} @buffer_load_int(<4 x i32> inreg) {
253main_body:
254  %data = call <4 x i32> @llvm.amdgcn.raw.buffer.load.v4i32(<4 x i32> %0, i32 0, i32 0, i32 0)
255  %data_glc = call <2 x i32> @llvm.amdgcn.raw.buffer.load.v2i32(<4 x i32> %0, i32 0, i32 0, i32 1)
256  %data_slc = call i32 @llvm.amdgcn.raw.buffer.load.i32(<4 x i32> %0, i32 0, i32 0, i32 2)
257  %fdata = bitcast <4 x i32> %data to <4 x float>
258  %fdata_glc = bitcast <2 x i32> %data_glc to <2 x float>
259  %fdata_slc = bitcast i32 %data_slc to float
260  %r0 = insertvalue {<4 x float>, <2 x float>, float} undef, <4 x float> %fdata, 0
261  %r1 = insertvalue {<4 x float>, <2 x float>, float} %r0, <2 x float> %fdata_glc, 1
262  %r2 = insertvalue {<4 x float>, <2 x float>, float} %r1, float %fdata_slc, 2
263  ret {<4 x float>, <2 x float>, float} %r2
264}
265
266;CHECK-LABEL: {{^}}raw_buffer_load_ubyte:
267;CHECK-NEXT: %bb.
268;CHECK-NEXT: buffer_load_ubyte v{{[0-9]}}, off, s[0:3], 0
269;CHECK: s_waitcnt vmcnt(0)
270;CHECK-NEXT: v_cvt_f32_ubyte0_e32 v0, v0
271;CHECK-NEXT: ; return to shader part epilog
272define amdgpu_ps float @raw_buffer_load_ubyte(<4 x i32> inreg %rsrc) {
273main_body:
274  %tmp = call i8 @llvm.amdgcn.raw.buffer.load.i8(<4 x i32> %rsrc, i32 0, i32 0, i32 0)
275  %tmp2 = zext i8 %tmp to i32
276  %val = uitofp i32 %tmp2 to float
277  ret float %val
278}
279
280;CHECK-LABEL: {{^}}raw_buffer_load_ushort:
281;CHECK-NEXT: %bb.
282;CHECK-NEXT: buffer_load_ushort v{{[0-9]}}, off, s[0:3], 0
283;CHECK-NEXT: s_waitcnt vmcnt(0)
284;CHECK-NEXT: v_cvt_f32_u32_e32 v0, v0
285;CHECK-NEXT: ; return to shader part epilog
286define amdgpu_ps float @raw_buffer_load_ushort(<4 x i32> inreg %rsrc) {
287main_body:
288  %tmp = call i16 @llvm.amdgcn.raw.buffer.load.i16(<4 x i32> %rsrc, i32 0, i32 0, i32 0)
289  %tmp2 = zext i16 %tmp to i32
290  %val = uitofp i32 %tmp2 to float
291  ret float %val
292}
293
294;CHECK-LABEL: {{^}}raw_buffer_load_sbyte:
295;CHECK-NEXT: %bb.
296;CHECK-NEXT: buffer_load_sbyte v{{[0-9]}}, off, s[0:3], 0
297;CHECK-NEXT: s_waitcnt vmcnt(0)
298;CHECK-NEXT: v_cvt_f32_i32_e32 v0, v0
299;CHECK-NEXT: ; return to shader part epilog
300define amdgpu_ps float @raw_buffer_load_sbyte(<4 x i32> inreg %rsrc) {
301main_body:
302  %tmp = call i8 @llvm.amdgcn.raw.buffer.load.i8(<4 x i32> %rsrc, i32 0, i32 0, i32 0)
303  %tmp2 = sext i8 %tmp to i32
304  %val = sitofp i32 %tmp2 to float
305  ret float %val
306}
307
308;CHECK-LABEL: {{^}}raw_buffer_load_sshort:
309;CHECK-NEXT: %bb.
310;CHECK-NEXT: buffer_load_sshort v{{[0-9]}}, off, s[0:3], 0
311;CHECK-NEXT: s_waitcnt vmcnt(0)
312;CHECK-NEXT: v_cvt_f32_i32_e32 v0, v0
313;CHECK-NEXT: ; return to shader part epilog
314define amdgpu_ps float @raw_buffer_load_sshort(<4 x i32> inreg %rsrc) {
315main_body:
316  %tmp = call i16 @llvm.amdgcn.raw.buffer.load.i16(<4 x i32> %rsrc, i32 0, i32 0, i32 0)
317  %tmp2 = sext i16 %tmp to i32
318  %val = sitofp i32 %tmp2 to float
319  ret float %val
320}
321
322declare float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32>, i32, i32, i32) #0
323declare <2 x float> @llvm.amdgcn.raw.buffer.load.v2f32(<4 x i32>, i32, i32, i32) #0
324declare <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32>, i32, i32, i32) #0
325declare i32 @llvm.amdgcn.raw.buffer.load.i32(<4 x i32>, i32, i32, i32) #0
326declare <2 x i32> @llvm.amdgcn.raw.buffer.load.v2i32(<4 x i32>, i32, i32, i32) #0
327declare <4 x i32> @llvm.amdgcn.raw.buffer.load.v4i32(<4 x i32>, i32, i32, i32) #0
328declare void @llvm.amdgcn.exp.f32(i32, i32, float, float, float, float, i1, i1) #0
329declare i8 @llvm.amdgcn.raw.buffer.load.i8(<4 x i32>, i32, i32, i32) #0
330declare i16 @llvm.amdgcn.raw.buffer.load.i16(<4 x i32>, i32, i32, i32) #0
331
332attributes #0 = { nounwind readonly }
333