xref: /llvm-project/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll (revision a796f597cd445bf6577c09efcac710b568f29d92) 
1; RUN: opt -mtriple amdgcn-- -passes='print<uniformity>' -disable-output %s 2>&1 | FileCheck %s
2
3; CHECK: DIVERGENT: %swizzle = call i32 @llvm.amdgcn.ds.swizzle(i32 %src, i32 100) #0
4define amdgpu_kernel void @ds_swizzle(ptr addrspace(1) %out, i32 %src) #0 {
5  %swizzle = call i32 @llvm.amdgcn.ds.swizzle(i32 %src, i32 100) #0
6  store i32 %swizzle, ptr addrspace(1) %out, align 4
7  ret void
8}
9
10; CHECK: DIVERGENT: %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
11define amdgpu_kernel void @v_permlane16_b32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) #0 {
12  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
13  store i32 %v, ptr addrspace(1) %out
14  ret void
15}
16
17; CHECK: DIVERGENT: %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
18define amdgpu_kernel void @v_permlanex16_b32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) #0 {
19  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
20  store i32 %v, ptr addrspace(1) %out
21  ret void
22}
23
24; CHECK: DIVERGENT: %v = call i32 @llvm.amdgcn.permlane16.var(i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
25define amdgpu_kernel void @v_permlane16_var_b32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) #0 {
26  %v = call i32 @llvm.amdgcn.permlane16.var(i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
27  store i32 %v, ptr addrspace(1) %out
28  ret void
29}
30
31; CHECK: DIVERGENT: %v = call i32 @llvm.amdgcn.permlanex16.var(i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
32define amdgpu_kernel void @v_permlanex16_var_b32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) #0 {
33  %v = call i32 @llvm.amdgcn.permlanex16.var(i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
34  store i32 %v, ptr addrspace(1) %out
35  ret void
36}
37
38; CHECK: DIVERGENT: %tmp0 = call i32 @llvm.amdgcn.update.dpp.i32(i32 %in1, i32 %in2, i32 1, i32 1, i32 1, i1 false) #0
39define amdgpu_kernel void @update_dpp(ptr addrspace(1) %out, i32 %in1, i32 %in2) #0 {
40  %tmp0 = call i32 @llvm.amdgcn.update.dpp.i32(i32 %in1, i32 %in2, i32 1, i32 1, i32 1, i1 false) #0
41  store i32 %tmp0, ptr addrspace(1) %out
42  ret void
43}
44
45; CHECK: DIVERGENT: %tmp0 = call i32 @llvm.amdgcn.mov.dpp.i32(i32 %in, i32 1, i32 1, i32 1, i1 true) #0
46define amdgpu_kernel void @mov_dpp(ptr addrspace(1) %out, i32 %in) #0 {
47  %tmp0 = call i32 @llvm.amdgcn.mov.dpp.i32(i32 %in, i32 1, i32 1, i32 1, i1 true) #0
48  store i32 %tmp0, ptr addrspace(1) %out
49  ret void
50}
51
52; CHECK: DIVERGENT: %tmp0 = call i32 @llvm.amdgcn.mov.dpp8.i32(i32 %in, i32 1) #0
53define amdgpu_kernel void @mov_dpp8(ptr addrspace(1) %out, i32 %in) #0 {
54  %tmp0 = call i32 @llvm.amdgcn.mov.dpp8.i32(i32 %in, i32 1) #0
55  store i32 %tmp0, ptr addrspace(1) %out
56  ret void
57}
58
59; CHECK: DIVERGENT: %tmp0 = call i32 @llvm.amdgcn.writelane.i32(i32 0, i32 1, i32 2)
60define amdgpu_kernel void @writelane(ptr addrspace(1) %out) #0 {
61  %tmp0 = call i32 @llvm.amdgcn.writelane.i32(i32 0, i32 1, i32 2)
62  store i32 %tmp0, ptr addrspace(1) %out
63  ret void
64}
65
66; CHECK: DIVERGENT: %tmp0 = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v16f16(<16 x half> %A, <16 x half> %B, <8 x float> %C)
67define amdgpu_kernel void @wmma_f32_16x16x16_f16(<16 x half> %A, <16 x half> %B, <8 x float> %C, ptr addrspace(1) %out) {
68  %tmp0 = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v16f16(<16 x half> %A, <16 x half> %B, <8 x float> %C)
69  store <8 x float> %tmp0, ptr addrspace(1) %out, align 32
70  ret void
71}
72
73; CHECK: DIVERGENT: %tmp0 = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v16i16(<16 x i16> %A, <16 x i16> %B, <8 x float> %C)
74define amdgpu_kernel void @wmma_f32_16x16x16_ibf16(<16 x i16> %A, <16 x i16> %B, <8 x float> %C, ptr addrspace(1) %out) {
75  %tmp0 = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v16i16(<16 x i16> %A, <16 x i16> %B, <8 x float> %C)
76  store <8 x float> %tmp0, ptr addrspace(1) %out, align 32
77  ret void
78}
79
80; CHECK: DIVERGENT: %tmp0 = call <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v16f16.v16f16(<16 x half> %A, <16 x half> %B, <16 x half> %C, i1 false)
81define amdgpu_kernel void @wmma_f16_16x16x16_f16(<16 x half> %A, <16 x half> %B, <16 x half> %C, ptr addrspace(1) %out) {
82bb:
83  %tmp0 = call <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v16f16.v16f16(<16 x half> %A, <16 x half> %B, <16 x half> %C, i1 false)
84  store <16 x half> %tmp0, ptr addrspace(1) %out, align 32
85  ret void
86}
87
88; CHECK: DIVERGENT: %tmp0 = call <16 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v16i16.v16i16(<16 x i16> %A, <16 x i16> %B, <16 x i16> %C, i1 false)
89define amdgpu_kernel void @wmma_f16_16x16x16_bf16(<16 x i16> %A, <16 x i16> %B, <16 x i16> %C, ptr addrspace(1) %out) {
90bb:
91  %tmp0 = call <16 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v16i16.v16i16(<16 x i16> %A, <16 x i16> %B, <16 x i16> %C, i1 false)
92  store <16 x i16> %tmp0, ptr addrspace(1) %out, align 32
93  ret void
94}
95
96; CHECK: DIVERGENT: %tmp0 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 false, <4 x i32> %A, i1 false, <4 x i32> %B, <8 x i32> %C, i1 false)
97define amdgpu_kernel void @wmma_i32_16x16x16_ui8(<4 x i32> %A, <4 x i32> %B, <8 x i32> %C, ptr addrspace(1) %out) {
98bb:
99  %tmp0 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 false, <4 x i32> %A, i1 false, <4 x i32> %B, <8 x i32> %C, i1 false)
100  store <8 x i32> %tmp0, ptr addrspace(1) %out, align 32
101  ret void
102}
103
104; CHECK: DIVERGENT: %tmp0 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 false, <2 x i32> %A, i1 false, <2 x i32> %B, <8 x i32> %C, i1 false)
105define amdgpu_kernel void @wmma_i32_16x16x16_ui4(<2 x i32> %A, <2 x i32> %B, <8 x i32> %C, ptr addrspace(1) %out) {
106bb:
107  %tmp0 = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 false, <2 x i32> %A, i1 false, <2 x i32> %B, <8 x i32> %C, i1 false)
108  store <8 x i32> %tmp0, ptr addrspace(1) %out, align 32
109  ret void
110}
111
112; CHECK: DIVERGENT: %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.i16(<8 x half> %A, <16 x half> %B, <8 x float> %C, i16 %Index)
113define amdgpu_kernel void @swmmac_f32_16x16x32_f16(<8 x half> %A, <16 x half> %B, <8 x float> %C, i16 %Index, ptr addrspace(1) %out) {
114bb:
115  %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.i16(<8 x half> %A, <16 x half> %B, <8 x float> %C, i16 %Index)
116    store <8 x float> %tmp0, ptr addrspace(1) %out, align 32
117    ret void
118}
119
120; CHECK: DIVERGENT: %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x float> %C, i16 %Index)
121define amdgpu_kernel void @swmmac_f32_16x16x32_bf16(<8 x i16> %A, <16 x i16> %B, <8 x float> %C, i16 %Index, ptr addrspace(1) %out) {
122bb:
123  %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x float> %C, i16 %Index)
124    store <8 x float> %tmp0, ptr addrspace(1) %out, align 32
125    ret void
126}
127
128; CHECK: DIVERGENT: %tmp0 = call <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.i16(<8 x half> %A, <16 x half> %B, <8 x half> %C, i16 %Index)
129define amdgpu_kernel void @swmmac_f16_16x16x32_f16(<8 x half> %A, <16 x half> %B, <8 x half> %C, i16 %Index, ptr addrspace(1) %out) {
130bb:
131  %tmp0 = call <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.i16(<8 x half> %A, <16 x half> %B, <8 x half> %C, i16 %Index)
132    store <8 x half> %tmp0, ptr addrspace(1) %out, align 32
133    ret void
134}
135
136; CHECK: DIVERGENT: %tmp0 = call <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x i16> %C, i16 %Index)
137define amdgpu_kernel void @swmmac_bf16_16x16x32_bf16(<8 x i16> %A, <16 x i16> %B, <8 x i16> %C, i16 %Index, ptr addrspace(1) %out) {
138bb:
139  %tmp0 = call <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.i16(<8 x i16> %A, <16 x i16> %B, <8 x i16> %C, i16 %Index)
140  store <8 x i16> %tmp0, ptr addrspace(1) %out, align 32
141  ret void
142}
143
144; CHECK: DIVERGENT: %tmp0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.i16(i1 false, <2 x i32> %A, i1 false, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 false)
145define amdgpu_kernel void @swmmac_i32_16x16x32_iu8(<2 x i32> %A, <4 x i32> %B, <8 x i32> %C, i16 %Index, ptr addrspace(1) %out) {
146bb:
147  %tmp0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.i16(i1 false, <2 x i32> %A, i1 false, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 false)
148  store <8 x i32> %tmp0, ptr addrspace(1) %out, align 32
149  ret void
150}
151
152; CHECK: DIVERGENT: %tmp0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.i16(i1 false, i32 %A, i1 false, <2 x i32> %B, <8 x i32> %C, i16 %Index, i1 false)
153define amdgpu_kernel void @swmmac_i32_16x16x32_iu4(i32 %A, <2 x i32> %B, <8 x i32> %C, i16 %Index, ptr addrspace(1) %out) {
154bb:
155  %tmp0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.i16(i1 false, i32 %A, i1 false, <2 x i32> %B, <8 x i32> %C, i16 %Index, i1 false)
156  store <8 x i32> %tmp0, ptr addrspace(1) %out, align 32
157  ret void
158}
159
160; CHECK: DIVERGENT: %tmp0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.i16(i1 false, <2 x i32> %A, i1 false, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 false)
161define amdgpu_kernel void @swmmac_i32_16x16x64_iu4(<2 x i32> %A, <4 x i32> %B, <8 x i32> %C, i16 %Index, ptr addrspace(1) %out) {
162bb:
163  %tmp0 = call <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.i16(i1 false, <2 x i32> %A, i1 false, <4 x i32> %B, <8 x i32> %C, i16 %Index, i1 false)
164  store <8 x i32> %tmp0, ptr addrspace(1) %out, align 32
165  ret void
166}
167
168; CHECK: DIVERGENT: %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
169define amdgpu_kernel void @swmmac_f32_16x16x32_fp8.fp8(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index, ptr addrspace(1) %out) {
170bb:
171  %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
172  store <8 x float> %tmp0, ptr addrspace(1) %out, align 32
173  ret void
174}
175
176; CHECK: DIVERGENT: %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
177define amdgpu_kernel void @swmmac_f32_16x16x32_fp8.bf8(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index, ptr addrspace(1) %out) {
178bb:
179  %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
180  store <8 x float> %tmp0, ptr addrspace(1) %out, align 32
181  ret void
182}
183
184; CHECK: DIVERGENT: %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
185define amdgpu_kernel void @swmmac_f32_16x16x32_bf8.fp8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index, ptr addrspace(1) %out) {
186bb:
187  %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
188  store <8 x float> %tmp0, ptr addrspace(1) %out, align 32
189  ret void
190}
191
192; CHECK: DIVERGENT: %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
193define amdgpu_kernel void @swmmac_f32_16x16x32_bf8.bf8.v8f32.v2i32.v4i32.i16(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index, ptr addrspace(1) %out) {
194bb:
195  %tmp0 = call <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8(<2 x i32> %A, <4 x i32> %B, <8 x float> %C, i16 %Index)
196  store <8 x float> %tmp0, ptr addrspace(1) %out, align 32
197  ret void
198}
199
200; CHECK: DIVERGENT: %tmp0 = call <2 x i32> @llvm.amdgcn.global.load.tr.b64.v2i32(ptr addrspace(1) %addr)
201define amdgpu_kernel void @global_load_tr_b64_v2i32(ptr addrspace(1) %addr, ptr addrspace(1) %out) {
202bb:
203  %tmp0 = call <2 x i32> @llvm.amdgcn.global.load.tr.b64.v2i32(ptr addrspace(1) %addr)
204  store <2 x i32> %tmp0, ptr addrspace(1) %out, align 8
205  ret void
206}
207
208; CHECK: DIVERGENT: %tmp0 = call <8 x i16> @llvm.amdgcn.global.load.tr.b128.v8i16(ptr addrspace(1) %addr)
209define amdgpu_kernel void @global_load_tr_b128_v8i16(ptr addrspace(1) %addr, ptr addrspace(1) %out) {
210bb:
211  %tmp0 = call <8 x i16> @llvm.amdgcn.global.load.tr.b128.v8i16(ptr addrspace(1) %addr)
212  store <8 x i16> %tmp0, ptr addrspace(1) %out, align 16
213  ret void
214}
215
216; CHECK: DIVERGENT: %tmp0 = call <8 x half> @llvm.amdgcn.global.load.tr.b128.v8f16(ptr addrspace(1) %addr)
217define amdgpu_kernel void @global_load_tr_b128_v8f16(ptr addrspace(1) %addr, ptr addrspace(1) %out) {
218bb:
219  %tmp0 = call <8 x half> @llvm.amdgcn.global.load.tr.b128.v8f16(ptr addrspace(1) %addr)
220  store <8 x half> %tmp0, ptr addrspace(1) %out, align 16
221  ret void
222}
223
224; CHECK: DIVERGENT: %tmp0 = call <8 x bfloat> @llvm.amdgcn.global.load.tr.b128.v8bf16(ptr addrspace(1) %addr)
225define amdgpu_kernel void @global_load_tr_b128_v8bf16(ptr addrspace(1) %addr, ptr addrspace(1) %out) {
226bb:
227  %tmp0 = call <8 x bfloat> @llvm.amdgcn.global.load.tr.b128.v8bf16(ptr addrspace(1) %addr)
228  store <8 x bfloat> %tmp0, ptr addrspace(1) %out, align 16
229  ret void
230}
231
232; CHECK: DIVERGENT: %tmp0 = call i32 @llvm.amdgcn.global.load.tr.b64.i32(ptr addrspace(1) %addr)
233define amdgpu_kernel void @global_load_tr_b64_i32(ptr addrspace(1) %addr, ptr addrspace(1) %out) {
234bb:
235  %tmp0 = call i32 @llvm.amdgcn.global.load.tr.b64.i32(ptr addrspace(1) %addr)
236  store i32 %tmp0, ptr addrspace(1) %out, align 4
237  ret void
238}
239
240; CHECK: DIVERGENT: %tmp0 = call <4 x i16> @llvm.amdgcn.global.load.tr.b128.v4i16(ptr addrspace(1) %addr)
241define amdgpu_kernel void @global_load_tr_b128_v4i16(ptr addrspace(1) %addr, ptr addrspace(1) %out) {
242bb:
243  %tmp0 = call <4 x i16> @llvm.amdgcn.global.load.tr.b128.v4i16(ptr addrspace(1) %addr)
244  store <4 x i16> %tmp0, ptr addrspace(1) %out, align 8
245  ret void
246}
247
248; CHECK: DIVERGENT: %tmp0 = call <4 x half> @llvm.amdgcn.global.load.tr.b128.v4f16(ptr addrspace(1) %addr)
249define amdgpu_kernel void @global_load_tr_b128_v4f16(ptr addrspace(1) %addr, ptr addrspace(1) %out) {
250bb:
251  %tmp0 = call <4 x half> @llvm.amdgcn.global.load.tr.b128.v4f16(ptr addrspace(1) %addr)
252  store <4 x half> %tmp0, ptr addrspace(1) %out, align 8
253  ret void
254}
255
256; CHECK: DIVERGENT: %tmp0 = call <4 x bfloat> @llvm.amdgcn.global.load.tr.b128.v4bf16(ptr addrspace(1) %addr)
257define amdgpu_kernel void @global_load_tr_b128_v4bf16(ptr addrspace(1) %addr, ptr addrspace(1) %out) {
258bb:
259  %tmp0 = call <4 x bfloat> @llvm.amdgcn.global.load.tr.b128.v4bf16(ptr addrspace(1) %addr)
260  store <4 x bfloat> %tmp0, ptr addrspace(1) %out, align 8
261  ret void
262}
263
264declare <2 x i32> @llvm.amdgcn.ds.read.tr4.b64.v2i32(ptr addrspace(3))
265
266; CHECK: DIVERGENT: %tmp0 = call <2 x i32> @llvm.amdgcn.ds.read.tr4.b64.v2i32(ptr addrspace(3) %gep)
267define amdgpu_kernel void @ds_read_b64_tr4_v2i32(ptr addrspace(3) %addr, ptr addrspace(1) %out) {
268bb:
269  %gep = getelementptr i64, ptr addrspace(3) %addr, i32 4
270  %tmp0 = call <2 x i32> @llvm.amdgcn.ds.read.tr4.b64.v2i32(ptr addrspace(3) %gep)
271  store <2 x i32> %tmp0, ptr addrspace(1) %out, align 8
272  ret void
273}
274
275declare <3 x i32> @llvm.amdgcn.ds.read.tr6.b96.v3i32(ptr addrspace(3))
276
277; CHECK: DIVERGENT: %tmp0 = call <3 x i32> @llvm.amdgcn.ds.read.tr6.b96.v3i32(ptr addrspace(3) %gep)
278define amdgpu_kernel void @ds_read_b96_tr6_v3i32(ptr addrspace(3) %addr, ptr addrspace(1) %out) {
279bb:
280  %gep = getelementptr i64, ptr addrspace(3) %addr, i32 4
281  %tmp0 = call <3 x i32> @llvm.amdgcn.ds.read.tr6.b96.v3i32(ptr addrspace(3) %gep)
282  store <3 x i32> %tmp0, ptr addrspace(1) %out, align 16
283  ret void
284}
285
286declare <2 x i32> @llvm.amdgcn.ds.read.tr8.b64.v2i32(ptr addrspace(3))
287
288; CHECK: DIVERGENT: %tmp0 = call <2 x i32> @llvm.amdgcn.ds.read.tr8.b64.v2i32(ptr addrspace(3) %gep)
289define amdgpu_kernel void @ds_read_b64_tr8_v2i32(ptr addrspace(3) %addr, ptr addrspace(1) %out) {
290bb:
291  %gep = getelementptr i64, ptr addrspace(3) %addr, i32 4
292  %tmp0 = call <2 x i32> @llvm.amdgcn.ds.read.tr8.b64.v2i32(ptr addrspace(3) %gep)
293  store <2 x i32> %tmp0, ptr addrspace(1) %out, align 8
294  ret void
295}
296
297declare <4 x i16> @llvm.amdgcn.ds.read.tr16.b64.v4i16(ptr addrspace(3))
298
299; CHECK: DIVERGENT: %tmp0 = call <4 x i16> @llvm.amdgcn.ds.read.tr16.b64.v4i16(ptr addrspace(3) %gep)
300define amdgpu_kernel void @ds_read_b64_tr_b16_v4i16(ptr addrspace(3) %addr, ptr addrspace(1) %out) {
301bb:
302  %gep = getelementptr i64, ptr addrspace(3) %addr, i16 4
303  %tmp0 = call <4 x i16> @llvm.amdgcn.ds.read.tr16.b64.v4i16(ptr addrspace(3) %gep)
304  store <4 x i16> %tmp0, ptr addrspace(1) %out, align 16
305  ret void
306}
307
308declare <4 x half> @llvm.amdgcn.ds.read.tr16.b64.v4f16(ptr addrspace(3))
309
310; CHECK: DIVERGENT: %tmp0 = call <4 x half> @llvm.amdgcn.ds.read.tr16.b64.v4f16(ptr addrspace(3) %gep)
311define amdgpu_kernel void @ds_read_b64_tr_b16_v4f16(ptr addrspace(3) %addr, ptr addrspace(1) %out) {
312bb:
313  %gep = getelementptr i64, ptr addrspace(3) %addr, i32 4
314  %tmp0 = call <4 x half> @llvm.amdgcn.ds.read.tr16.b64.v4f16(ptr addrspace(3) %gep)
315  store <4 x half> %tmp0, ptr addrspace(1) %out, align 16
316  ret void
317}
318
319declare <4 x bfloat> @llvm.amdgcn.ds.read.tr16.b64.v4bf16(ptr addrspace(3))
320
321; CHECK: DIVERGENT: %tmp0 = call <4 x bfloat> @llvm.amdgcn.ds.read.tr16.b64.v4bf16(ptr addrspace(3) %gep)
322define amdgpu_kernel void @ds_read_b64_tr_b16_v4bf16(ptr addrspace(3) %addr, ptr addrspace(1) %out) {
323bb:
324  %gep = getelementptr i64, ptr addrspace(3) %addr, i32 4
325  %tmp0 = call <4 x bfloat> @llvm.amdgcn.ds.read.tr16.b64.v4bf16(ptr addrspace(3) %gep)
326  store <4 x bfloat> %tmp0, ptr addrspace(1) %out, align 16
327  ret void
328}
329
330declare <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.f16(<8 x half>, <8 x half>, <4 x float>, i32 immarg, i32 immarg, i32 immarg)
331declare <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.f16(<8 x half>, <8 x half>, <16 x float>, i32 immarg, i32 immarg, i32 immarg)
332
333; CHECK: DIVERGENT: %result = call <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.f16(<8 x half> %arg0, <8 x half> %arg1, <4 x float> %arg2, i32 immarg 0, i32 immarg 0, i32 immarg 0)
334define amdgpu_kernel void @mfma_f32_16x16x32_f16(<8 x half> %arg0, <8 x half> %arg1, <4 x float> %arg2, ptr addrspace(1) %out) {
335  %result = call <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.f16(<8 x half> %arg0, <8 x half> %arg1, <4 x float> %arg2, i32 immarg 0, i32 immarg 0, i32 immarg 0)
336  store <4 x float> %result, ptr addrspace(1) %out
337  ret void
338}
339
340; CHECK: DIVERGENT: %result = call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.f16(<8 x half> %arg0, <8 x half> %arg1, <16 x float> %arg2, i32 immarg 0, i32 immarg 0, i32 immarg 0)
341define amdgpu_kernel void @mfma_f32_32x32x16_f16(<8 x half> %arg0, <8 x half> %arg1, <16 x float> %arg2, ptr addrspace(1) %out) {
342  %result = call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.f16(<8 x half> %arg0, <8 x half> %arg1, <16 x float> %arg2, i32 immarg 0, i32 immarg 0, i32 immarg 0)
343  store <16 x float> %result, ptr addrspace(1) %out
344  ret void
345}
346
347; CHECK: DIVERGENT:  %result = call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.bf16(<8 x bfloat> %arg0, <8 x bfloat> %arg1, <16 x float> %arg2, i32 immarg 0, i32 immarg 0, i32 immarg 0)
348define amdgpu_kernel void @mfma_f32_32x32x16_bf16(<8 x bfloat> %arg0, <8 x bfloat> %arg1, <16 x float> %arg2, ptr addrspace(1) %out) {
349  %result = call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.bf16(<8 x bfloat> %arg0, <8 x bfloat> %arg1, <16 x float> %arg2, i32 immarg 0, i32 immarg 0, i32 immarg 0)
350  store <16 x float> %result, ptr addrspace(1) %out
351  ret void
352}
353
354declare <4 x float> @llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.v8i32.v8i32(<8 x i32>, <8 x i32>, <4 x float>, i32 immarg, i32 immarg,
355                                                                              i32 immarg, i32, i32 immarg, i32)
356
357; CHECK: DIVERGENT:   %result = call <4 x float> @llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.v8i32.v8i32(<8 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 immarg 0, i32 immarg 0, i32 immarg 0, i32 %arg3, i32 immarg 0, i32 %arg4)
358define amdgpu_kernel void @mfma_scale_f32_16x16x128_f8f6f4(<8 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, i32 %arg4, ptr addrspace(1) %out) {
359  %result = call <4 x float> @llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.v8i32.v8i32(<8 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 immarg 0, i32 immarg 0, i32 immarg 0, i32 %arg3, i32 immarg 0, i32 %arg4)
360  store <4 x float> %result, ptr addrspace(1) %out
361  ret void
362}
363
364declare <16 x float> @llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4(<8 x i32>, <8 x i32>, <16 x float>, i32 immarg, i32 immarg,
365                                                                i32 immarg, i32, i32 immarg, i32)
366
367; CHECK: DIVERGENT: %result = call <16 x float> @llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.v8i32.v8i32(<8 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 immarg 0, i32 immarg 0, i32 immarg 0, i32 %arg3, i32 immarg 0, i32 %arg4)
368define amdgpu_kernel void @mfma_f32_scale_32x32x64_f8f6f4(<8 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 %arg3, i32 %arg4, ptr addrspace(1) %out) {
369  %result = call <16 x float> @llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.v8i32.v8i32(<8 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 immarg 0, i32 immarg 0, i32 immarg 0, i32 %arg3, i32 immarg 0, i32 %arg4)
370  store <16 x float> %result, ptr addrspace(1) %out
371  ret void
372}
373
374declare <4 x i32> @llvm.amdgcn.mfma.i32.16x16x64.i8(<4 x i32>, <4 x i32>, <4 x i32>, i32 immarg, i32 immarg, i32 immarg)
375
376; CHECK: DIVERGENT: %result = call <4 x i32> @llvm.amdgcn.mfma.i32.16x16x64.i8(<4 x i32> %arg0, <4 x i32> %arg1, <4 x i32> %arg2, i32 immarg 0, i32 immarg 0, i32 immarg 0)
377define amdgpu_kernel void @mfma_i32_16x16x64_i8(<4 x i32> %arg0, <4 x i32> %arg1, <4 x i32> %arg2, ptr addrspace(1) %out) {
378  %result = call <4 x i32> @llvm.amdgcn.mfma.i32.16x16x64.i8(<4 x i32> %arg0, <4 x i32> %arg1, <4 x i32> %arg2, i32 immarg 0, i32 immarg 0, i32 immarg 0)
379  store <4 x i32> %result, ptr addrspace(1) %out
380  ret void
381}
382
383declare <16 x i32> @llvm.amdgcn.mfma.i32.32x32x32.i8(<4 x i32>, <4 x i32>, <16 x i32>, i32 immarg, i32 immarg, i32 immarg)
384
385; CHECK: DIVERGENT: %result = call <16 x i32> @llvm.amdgcn.mfma.i32.32x32x32.i8(<4 x i32> %arg0, <4 x i32> %arg1, <16 x i32> %arg2, i32 immarg 0, i32 immarg 0, i32 immarg 0)
386define amdgpu_kernel void @mfma_i32_32x32x32_i8(<4 x i32> %arg0, <4 x i32> %arg1, <16 x i32> %arg2, ptr addrspace(1) %out) {
387  %result = call <16 x i32> @llvm.amdgcn.mfma.i32.32x32x32.i8(<4 x i32> %arg0, <4 x i32> %arg1, <16 x i32> %arg2, i32 immarg 0, i32 immarg 0, i32 immarg 0)
388  store <16 x i32> %result, ptr addrspace(1) %out
389  ret void
390}
391
392declare <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.bf16(<8 x bfloat>, <8 x bfloat>, <4 x float>, i32 immarg, i32 immarg, i32 immarg)
393
394; CHECK: DIVERGENT: %result = call <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.bf16(<8 x bfloat> %arg0, <8 x bfloat> %arg1, <4 x float> %arg2, i32 immarg 0, i32 immarg 0, i32 immarg 0)
395define amdgpu_kernel void @mfma_f32_16x16x32_bf16(<8 x bfloat> %arg0, <8 x bfloat> %arg1, <4 x float> %arg2, ptr addrspace(1) %out) {
396  %result = call <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.bf16(<8 x bfloat> %arg0, <8 x bfloat> %arg1, <4 x float> %arg2, i32 immarg 0, i32 immarg 0, i32 immarg 0)
397  store <4 x float> %result, ptr addrspace(1) %out
398  ret void
399}
400
401declare <4 x float> @llvm.amdgcn.smmfmac.f32.16x16x64.f16(<8 x half>, <16 x half>, <4 x float>, i32, i32 immarg, i32 immarg)
402
403; CHECK: DIVERGENT:   %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x64.f16(<8 x half> %arg0, <16 x half> %arg1, <4 x float> %arg2, i32 %arg3, i32 immarg 0, i32 immarg 0)
404define amdgpu_kernel void @smfmac_f32_16x16x64_f16(<8 x half> %arg0, <16 x half> %arg1, <4 x float> %arg2, i32 %arg3, ptr addrspace(1) %out) {
405  %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x64.f16(<8 x half> %arg0, <16 x half> %arg1, <4 x float> %arg2, i32 %arg3, i32 immarg 0, i32 immarg 0)
406  store <4 x float> %result, ptr addrspace(1) %out
407  ret void
408}
409
410declare <16 x float> @llvm.amdgcn.smfmac.f32.32x32x32.f16(<8 x half>, <16 x half>, <16 x float>, i32, i32 immarg, i32 immarg)
411
412; CHECK: DIVERGENT:   %result = call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x32.f16(<8 x half> %arg0, <16 x half> %arg1, <16 x float> %arg2, i32 %arg3, i32 immarg 0, i32 immarg 0)
413define amdgpu_kernel void @smfmac_f32_32x32x32_f16(<8 x half> %arg0, <16 x half> %arg1, <16 x float> %arg2, i32 %arg3, ptr addrspace(1) %out) {
414  %result = call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x32.f16(<8 x half> %arg0, <16 x half> %arg1, <16 x float> %arg2, i32 %arg3, i32 immarg 0, i32 immarg 0)
415  store <16 x float> %result, ptr addrspace(1) %out
416  ret void
417}
418
419declare <4 x float> @llvm.amdgcn.smmfmac.f32.16x16x64.bf16(<8 x bfloat>, <16 x bfloat>, <4 x float>, i32, i32 immarg, i32 immarg)
420
421; CHECK: DIVERGENT:   %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x64.bf16(<8 x bfloat> %arg0, <16 x bfloat> %arg1, <4 x float> %arg2, i32 %arg3, i32 immarg 0, i32 immarg 0)
422define amdgpu_kernel void @smfmac_f32_16x16x64_bf16(<8 x bfloat> %arg0, <16 x bfloat> %arg1, <4 x float> %arg2, i32 %arg3, ptr addrspace(1) %out) {
423  %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x64.bf16(<8 x bfloat> %arg0, <16 x bfloat> %arg1, <4 x float> %arg2, i32 %arg3, i32 immarg 0, i32 immarg 0)
424  store <4 x float> %result, ptr addrspace(1) %out
425  ret void
426}
427
428declare <4 x i32> @llvm.amdgcn.smfmac.i32.16x16x128.i8(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32)
429
430; CHECK: DIVERGENT:   %result = call <4 x i32> @llvm.amdgcn.smfmac.i32.16x16x128.i8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x i32> %arg2, i32 %arg3, i32 1, i32 2)
431define amdgpu_kernel void @smfmac_i32_16x16x128_i8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x i32> %arg2, i32 %arg3, ptr addrspace(1) %out) {
432  %result = call <4 x i32> @llvm.amdgcn.smfmac.i32.16x16x128.i8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x i32> %arg2, i32 %arg3, i32 1, i32 2)
433  store <4 x i32> %result, ptr addrspace(1) %out
434  ret void
435}
436
437declare <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.bf8(<4 x i32>, <8 x i32>, <4 x float>, i32, i32, i32)
438
439; CHECK: DIVERGENT:  %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.bf8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, i32 1, i32 2)
440define amdgpu_kernel void @smfmac_f32_16x16x128_bf8_bf8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, ptr addrspace(1) %out) {
441  %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.bf8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, i32 1, i32 2)
442  store <4 x float> %result, ptr addrspace(1) %out
443  ret void
444}
445
446declare <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.fp8(<4 x i32>, <8 x i32>, <4 x float>, i32, i32, i32)
447
448; CHECK: DIVERGENT:  %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.fp8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, i32 1, i32 2)
449define amdgpu_kernel void @smfmac_f32_16x16x128_bf8_fp8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, ptr addrspace(1) %out) {
450  %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.fp8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, i32 1, i32 2)
451  store <4 x float> %result, ptr addrspace(1) %out
452  ret void
453}
454
455declare <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.fp8.bf8(<4 x i32>, <8 x i32>, <4 x float>, i32, i32, i32)
456
457; CHECK: DIVERGENT:  %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.fp8.bf8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, i32 1, i32 2)
458define amdgpu_kernel void @smfmac_f32_16x16x128_fp8_bf8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, ptr addrspace(1) %out) {
459  %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.fp8.bf8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, i32 1, i32 2)
460  store <4 x float> %result, ptr addrspace(1) %out
461  ret void
462}
463
464declare <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.fp8.fp8(<4 x i32>, <8 x i32>, <4 x float>, i32, i32, i32)
465
466; CHECK: DIVERGENT:  %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.fp8.fp8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, i32 1, i32 2)
467define amdgpu_kernel void @smfmac_f32_16x16x128_fp8_fp8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, ptr addrspace(1) %out) {
468  %result = call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.fp8.fp8(<4 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 %arg3, i32 1, i32 2)
469  store <4 x float> %result, ptr addrspace(1) %out
470  ret void
471}
472
473declare <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.bf8.bf8(<4 x i32>, <8 x i32>, <16 x float>, i32, i32, i32)
474
475; CHECK: DIVERGENT:  %result = call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.bf8.bf8(<4 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 %arg3, i32 1, i32 2)
476define amdgpu_kernel void @smfmac_f32_32x32x64_bf8_bf8(<4 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 %arg3, ptr addrspace(1) %out) {
477  %result = call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.bf8.bf8(<4 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 %arg3, i32 1, i32 2)
478  store <16 x float> %result, ptr addrspace(1) %out
479  ret void
480}
481
482declare <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.bf8.fp8(<4 x i32>, <8 x i32>, <16 x float>, i32, i32, i32)
483
484; CHECK: DIVERGENT:  %result = call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.bf8.fp8(<4 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 %arg3, i32 1, i32 2)
485define amdgpu_kernel void @smfmac_f32_32x32x64_bf8_fp8(<4 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 %arg3, ptr addrspace(1) %out) {
486  %result = call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.bf8.fp8(<4 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 %arg3, i32 1, i32 2)
487  store <16 x float> %result, ptr addrspace(1) %out
488  ret void
489}
490
491declare <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.fp8.bf8(<4 x i32>, <8 x i32>, <16 x float>, i32, i32, i32)
492
493; CHECK: DIVERGENT:  %result = call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.fp8.bf8(<4 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 %arg3, i32 1, i32 2)
494define amdgpu_kernel void @smfmac_f32_32x32x64_fp8_bf8(<4 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 %arg3, ptr addrspace(1) %out) {
495  %result = call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.fp8.bf8(<4 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 %arg3, i32 1, i32 2)
496  store <16 x float> %result, ptr addrspace(1) %out
497  ret void
498}
499
500declare <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.fp8.fp8(<4 x i32>, <8 x i32>, <16 x float>, i32, i32, i32)
501
502; CHECK: DIVERGENT:  %result = call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.fp8.fp8(<4 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 %arg3, i32 1, i32 2)
503define amdgpu_kernel void @smfmac_f32_32x32x64_fp8_fp8(<4 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 %arg3, ptr addrspace(1) %out) {
504  %result = call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.fp8.fp8(<4 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 %arg3, i32 1, i32 2)
505  store <16 x float> %result, ptr addrspace(1) %out
506  ret void
507}
508
509; CHECK: DIVERGENT:   %v = call { i32, i32 } @llvm.amdgcn.permlane16.swap(i32 %src0, i32 %src1, i1 false, i1 false)
510define amdgpu_kernel void @v_permlane16_swap(ptr addrspace(1) %out, i32 %src0, i32 %src1) #0 {
511  %v = call { i32, i32 } @llvm.amdgcn.permlane16.swap(i32 %src0, i32 %src1, i1 false, i1 false)
512  store { i32, i32 } %v, ptr addrspace(1) %out
513  ret void
514}
515
516; CHECK: DIVERGENT:   %v = call { i32, i32 } @llvm.amdgcn.permlane32.swap(i32 %src0, i32 %src1, i1 false, i1 false)
517define amdgpu_kernel void @v_permlane32_swap(ptr addrspace(1) %out, i32 %src0, i32 %src1) #0 {
518  %v = call { i32, i32 } @llvm.amdgcn.permlane32.swap(i32 %src0, i32 %src1, i1 false, i1 false)
519  store { i32, i32 } %v, ptr addrspace(1) %out
520  ret void
521}
522
523
524
525declare i32 @llvm.amdgcn.ds.swizzle(i32, i32) #1
526declare i32 @llvm.amdgcn.permlane16.i32(i32, i32, i32, i32, i1, i1) #1
527declare i32 @llvm.amdgcn.permlanex16.i32(i32, i32, i32, i32, i1, i1) #1
528declare i32 @llvm.amdgcn.permlane16.var(i32, i32, i32, i1, i1) #1
529declare i32 @llvm.amdgcn.permlanex16.var(i32, i32, i32, i1, i1) #1
530declare i32 @llvm.amdgcn.mov.dpp.i32(i32, i32, i32, i32, i1) #1
531declare i32 @llvm.amdgcn.mov.dpp8.i32(i32, i32) #1
532declare i32 @llvm.amdgcn.update.dpp.i32(i32, i32, i32, i32, i32, i1) #1
533declare i32 @llvm.amdgcn.writelane.i32(i32, i32, i32) #1
534declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v16f16(<16 x half>, <16 x half> , <8 x float>) #1
535declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v16i16(<16 x i16>, <16 x i16> , <8 x float>) #1
536declare <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v16f16.v16f16(<16 x half>, <16 x half> , <16 x half>, i1 immarg) #1
537declare <16 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v16i16.v16i16(<16 x i16>, <16 x i16> , <16 x i16>, i1 immarg) #1
538declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8.v8i32.v4i32(i1 immarg, <4 x i32>, i1 immarg, <4 x i32> , <8 x i32>, i1 immarg) #1
539declare <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4.v8i32.v2i32(i1 immarg, <2 x i32>, i1 immarg, <2 x i32> , <8 x i32>, i1 immarg) #1
540declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.i16(<8 x half>, <16 x half>, <8 x float>, i16)
541declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.i16(<8 x i16>, <16 x i16>, <8 x float>, i16)
542declare <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.i16(<8 x half>, <16 x half>, <8 x half>, i16)
543declare <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.i16(<8 x i16>, <16 x i16>, <8 x i16>, i16)
544declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu8.v8i32.v2i32.v4i32.i16(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i16, i1)
545declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x32.iu4.v8i32.i32.v2i32.i16(i1 immarg, i32, i1 immarg, <2 x i32>, <8 x i32>, i16, i1)
546declare <8 x i32> @llvm.amdgcn.swmmac.i32.16x16x64.iu4.v8i32.v2i32.v4i32.i16(i1 immarg, <2 x i32>, i1 immarg, <4 x i32>, <8 x i32>, i16, i1)
547declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8(<2 x i32>, <4 x i32>, <8 x float>, i16)
548declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8(<2 x i32>, <4 x i32>, <8 x float>, i16)
549declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8(<2 x i32>, <4 x i32>, <8 x float>, i16)
550declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8(<2 x i32>, <4 x i32>, <8 x float>, i16)
551
552declare <2 x i32> @llvm.amdgcn.global.load.tr.b64.v2i32(ptr addrspace(1))
553declare <8 x i16> @llvm.amdgcn.global.load.tr.b128.v8i16(ptr addrspace(1))
554declare <8 x half> @llvm.amdgcn.global.load.tr.b128.v8f16(ptr addrspace(1))
555declare <8 x bfloat> @llvm.amdgcn.global.load.tr.b128.v8bf16(ptr addrspace(1))
556declare i32 @llvm.amdgcn.global.load.tr.b64.i32(ptr addrspace(1))
557declare <4 x i16> @llvm.amdgcn.global.load.tr.b128.v4i16(ptr addrspace(1))
558declare <4 x half> @llvm.amdgcn.global.load.tr.b128.v4f16(ptr addrspace(1))
559declare <4 x bfloat> @llvm.amdgcn.global.load.tr.b128.v4bf16(ptr addrspace(1))
560
561attributes #0 = { nounwind convergent }
562attributes #1 = { nounwind readnone convergent }
563