1 ; RUN: opt -mtriple amdgcn-- -passes='print<uniformity>' -disable-output %s 2>&1 | FileCheck %s
3 ; CHECK: DIVERGENT: %swizzle = call i32 @llvm.amdgcn.ds.swizzle(i32 %src, i32 100) #0
4 define 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
10 ; CHECK: DIVERGENT: %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
11 define 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
17 ; CHECK: DIVERGENT: %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
18 define 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
24 ; CHECK: DIVERGENT: %v = call i32 @llvm.amdgcn.permlane16.var(i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
25 define 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
31 ; CHECK: DIVERGENT: %v = call i32 @llvm.amdgcn.permlanex16.var(i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
32 define 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
38 ; CHECK: DIVERGENT: %tmp0 = call i32 @llvm.amdgcn.update.dpp.i32(i32 %in1, i32 %in2, i32 1, i32 1, i32 1, i1 false) #0
39 define 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
45 ; CHECK: DIVERGENT: %tmp0 = call i32 @llvm.amdgcn.mov.dpp.i32(i32 %in, i32 1, i32 1, i32 1, i1 true) #0
46 define 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
52 ; CHECK: DIVERGENT: %tmp0 = call i32 @llvm.amdgcn.mov.dpp8.i32(i32 %in, i32 1) #0
53 define 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
59 ; CHECK: DIVERGENT: %tmp0 = call i32 @llvm.amdgcn.writelane.i32(i32 0, i32 1, i32 2)
60 define 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
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)
67 define 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
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)
74 define 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
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)
81 define amdgpu_kernel void @wmma_f16_16x16x16_f16(<16 x half> %A, <16 x half> %B, <16 x half> %C, ptr addrspace(1) %out) {
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
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)
89 define amdgpu_kernel void @wmma_f16_16x16x16_bf16(<16 x i16> %A, <16 x i16> %B, <16 x i16> %C, ptr addrspace(1) %out) {
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
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)
97 define amdgpu_kernel void @wmma_i32_16x16x16_ui8(<4 x i32> %A, <4 x i32> %B, <8 x i32> %C, ptr addrspace(1) %out) {
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
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)
105 define amdgpu_kernel void @wmma_i32_16x16x16_ui4(<2 x i32> %A, <2 x i32> %B, <8 x i32> %C, ptr addrspace(1) %out) {
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
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)
113 define 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) {
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
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)
121 define 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) {
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
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)
129 define 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) {
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
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)
137 define 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) {
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
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)
145 define 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) {
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
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)
153 define amdgpu_kernel void @swmmac_i32_16x16x32_iu4(i32 %A, <2 x i32> %B, <8 x i32> %C, i16 %Index, ptr addrspace(1) %out) {
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
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)
161 define 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) {
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
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)
169 define 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) {
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
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)
177 define 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) {
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
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)
185 define 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) {
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
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)
193 define 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) {
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
200 ; CHECK: DIVERGENT: %tmp0 = call <2 x i32> @llvm.amdgcn.global.load.tr.b64.v2i32(ptr addrspace(1) %addr)
201 define amdgpu_kernel void @global_load_tr_b64_v2i32(ptr addrspace(1) %addr, ptr addrspace(1) %out) {
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
208 ; CHECK: DIVERGENT: %tmp0 = call <8 x i16> @llvm.amdgcn.global.load.tr.b128.v8i16(ptr addrspace(1) %addr)
209 define amdgpu_kernel void @global_load_tr_b128_v8i16(ptr addrspace(1) %addr, ptr addrspace(1) %out) {
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
216 ; CHECK: DIVERGENT: %tmp0 = call <8 x half> @llvm.amdgcn.global.load.tr.b128.v8f16(ptr addrspace(1) %addr)
217 define amdgpu_kernel void @global_load_tr_b128_v8f16(ptr addrspace(1) %addr, ptr addrspace(1) %out) {
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
224 ; CHECK: DIVERGENT: %tmp0 = call <8 x bfloat> @llvm.amdgcn.global.load.tr.b128.v8bf16(ptr addrspace(1) %addr)
225 define amdgpu_kernel void @global_load_tr_b128_v8bf16(ptr addrspace(1) %addr, ptr addrspace(1) %out) {
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
232 ; CHECK: DIVERGENT: %tmp0 = call i32 @llvm.amdgcn.global.load.tr.b64.i32(ptr addrspace(1) %addr)
233 define amdgpu_kernel void @global_load_tr_b64_i32(ptr addrspace(1) %addr, ptr addrspace(1) %out) {
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
240 ; CHECK: DIVERGENT: %tmp0 = call <4 x i16> @llvm.amdgcn.global.load.tr.b128.v4i16(ptr addrspace(1) %addr)
241 define amdgpu_kernel void @global_load_tr_b128_v4i16(ptr addrspace(1) %addr, ptr addrspace(1) %out) {
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
248 ; CHECK: DIVERGENT: %tmp0 = call <4 x half> @llvm.amdgcn.global.load.tr.b128.v4f16(ptr addrspace(1) %addr)
249 define amdgpu_kernel void @global_load_tr_b128_v4f16(ptr addrspace(1) %addr, ptr addrspace(1) %out) {
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
256 ; CHECK: DIVERGENT: %tmp0 = call <4 x bfloat> @llvm.amdgcn.global.load.tr.b128.v4bf16(ptr addrspace(1) %addr)
257 define amdgpu_kernel void @global_load_tr_b128_v4bf16(ptr addrspace(1) %addr, ptr addrspace(1) %out) {
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
264 declare <2 x i32> @llvm.amdgcn.ds.read.tr4.b64.v2i32(ptr addrspace(3))
266 ; CHECK: DIVERGENT: %tmp0 = call <2 x i32> @llvm.amdgcn.ds.read.tr4.b64.v2i32(ptr addrspace(3) %gep)
267 define amdgpu_kernel void @ds_read_b64_tr4_v2i32(ptr addrspace(3) %addr, ptr addrspace(1) %out) {
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
275 declare <3 x i32> @llvm.amdgcn.ds.read.tr6.b96.v3i32(ptr addrspace(3))
277 ; CHECK: DIVERGENT: %tmp0 = call <3 x i32> @llvm.amdgcn.ds.read.tr6.b96.v3i32(ptr addrspace(3) %gep)
278 define amdgpu_kernel void @ds_read_b96_tr6_v3i32(ptr addrspace(3) %addr, ptr addrspace(1) %out) {
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
286 declare <2 x i32> @llvm.amdgcn.ds.read.tr8.b64.v2i32(ptr addrspace(3))
288 ; CHECK: DIVERGENT: %tmp0 = call <2 x i32> @llvm.amdgcn.ds.read.tr8.b64.v2i32(ptr addrspace(3) %gep)
289 define amdgpu_kernel void @ds_read_b64_tr8_v2i32(ptr addrspace(3) %addr, ptr addrspace(1) %out) {
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
297 declare <4 x i16> @llvm.amdgcn.ds.read.tr16.b64.v4i16(ptr addrspace(3))
299 ; CHECK: DIVERGENT: %tmp0 = call <4 x i16> @llvm.amdgcn.ds.read.tr16.b64.v4i16(ptr addrspace(3) %gep)
300 define amdgpu_kernel void @ds_read_b64_tr_b16_v4i16(ptr addrspace(3) %addr, ptr addrspace(1) %out) {
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
308 declare <4 x half> @llvm.amdgcn.ds.read.tr16.b64.v4f16(ptr addrspace(3))
310 ; CHECK: DIVERGENT: %tmp0 = call <4 x half> @llvm.amdgcn.ds.read.tr16.b64.v4f16(ptr addrspace(3) %gep)
311 define amdgpu_kernel void @ds_read_b64_tr_b16_v4f16(ptr addrspace(3) %addr, ptr addrspace(1) %out) {
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
319 declare <4 x bfloat> @llvm.amdgcn.ds.read.tr16.b64.v4bf16(ptr addrspace(3))
321 ; CHECK: DIVERGENT: %tmp0 = call <4 x bfloat> @llvm.amdgcn.ds.read.tr16.b64.v4bf16(ptr addrspace(3) %gep)
322 define amdgpu_kernel void @ds_read_b64_tr_b16_v4bf16(ptr addrspace(3) %addr, ptr addrspace(1) %out) {
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
330 declare <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.f16(<8 x half>, <8 x half>, <4 x float>, i32 immarg, i32 immarg, i32 immarg)
331 declare <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.f16(<8 x half>, <8 x half>, <16 x float>, i32 immarg, i32 immarg, i32 immarg)
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)
334 define 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
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)
341 define 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
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)
348 define 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
354 declare <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)
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)
358 define 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
364 declare <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)
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)
368 define 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
374 declare <4 x i32> @llvm.amdgcn.mfma.i32.16x16x64.i8(<4 x i32>, <4 x i32>, <4 x i32>, i32 immarg, i32 immarg, i32 immarg)
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)
377 define 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
383 declare <16 x i32> @llvm.amdgcn.mfma.i32.32x32x32.i8(<4 x i32>, <4 x i32>, <16 x i32>, i32 immarg, i32 immarg, i32 immarg)
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)
386 define 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
392 declare <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.bf16(<8 x bfloat>, <8 x bfloat>, <4 x float>, i32 immarg, i32 immarg, i32 immarg)
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)
395 define 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
401 declare <4 x float> @llvm.amdgcn.smmfmac.f32.16x16x64.f16(<8 x half>, <16 x half>, <4 x float>, i32, i32 immarg, i32 immarg)
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)
404 define 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
410 declare <16 x float> @llvm.amdgcn.smfmac.f32.32x32x32.f16(<8 x half>, <16 x half>, <16 x float>, i32, i32 immarg, i32 immarg)
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)
413 define 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
419 declare <4 x float> @llvm.amdgcn.smmfmac.f32.16x16x64.bf16(<8 x bfloat>, <16 x bfloat>, <4 x float>, i32, i32 immarg, i32 immarg)
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)
422 define 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
428 declare <4 x i32> @llvm.amdgcn.smfmac.i32.16x16x128.i8(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32)
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)
431 define 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
437 declare <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.bf8(<4 x i32>, <8 x i32>, <4 x float>, i32, i32, i32)
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)
440 define 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
446 declare <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.bf8.fp8(<4 x i32>, <8 x i32>, <4 x float>, i32, i32, i32)
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)
449 define 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
455 declare <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.fp8.bf8(<4 x i32>, <8 x i32>, <4 x float>, i32, i32, i32)
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)
458 define 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
464 declare <4 x float> @llvm.amdgcn.smfmac.f32.16x16x128.fp8.fp8(<4 x i32>, <8 x i32>, <4 x float>, i32, i32, i32)
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)
467 define 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
473 declare <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.bf8.bf8(<4 x i32>, <8 x i32>, <16 x float>, i32, i32, i32)
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)
476 define 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
482 declare <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.bf8.fp8(<4 x i32>, <8 x i32>, <16 x float>, i32, i32, i32)
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)
485 define 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
491 declare <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.fp8.bf8(<4 x i32>, <8 x i32>, <16 x float>, i32, i32, i32)
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)
494 define 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
500 declare <16 x float> @llvm.amdgcn.smfmac.f32.32x32x64.fp8.fp8(<4 x i32>, <8 x i32>, <16 x float>, i32, i32, i32)
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)
503 define 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
509 ; CHECK: DIVERGENT: %v = call { i32, i32 } @llvm.amdgcn.permlane16.swap(i32 %src0, i32 %src1, i1 false, i1 false)
510 define 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
516 ; CHECK: DIVERGENT: %v = call { i32, i32 } @llvm.amdgcn.permlane32.swap(i32 %src0, i32 %src1, i1 false, i1 false)
517 define 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
525 declare i32 @llvm.amdgcn.ds.swizzle(i32, i32) #1
526 declare i32 @llvm.amdgcn.permlane16.i32(i32, i32, i32, i32, i1, i1) #1
527 declare i32 @llvm.amdgcn.permlanex16.i32(i32, i32, i32, i32, i1, i1) #1
528 declare i32 @llvm.amdgcn.permlane16.var(i32, i32, i32, i1, i1) #1
529 declare i32 @llvm.amdgcn.permlanex16.var(i32, i32, i32, i1, i1) #1
530 declare i32 @llvm.amdgcn.mov.dpp.i32(i32, i32, i32, i32, i1) #1
531 declare i32 @llvm.amdgcn.mov.dpp8.i32(i32, i32) #1
532 declare i32 @llvm.amdgcn.update.dpp.i32(i32, i32, i32, i32, i32, i1) #1
533 declare i32 @llvm.amdgcn.writelane.i32(i32, i32, i32) #1
534 declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v16f16(<16 x half>, <16 x half> , <8 x float>) #1
535 declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v16i16(<16 x i16>, <16 x i16> , <8 x float>) #1
536 declare <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v16f16.v16f16(<16 x half>, <16 x half> , <16 x half>, i1 immarg) #1
537 declare <16 x i16> @llvm.amdgcn.wmma.bf16.16x16x16.bf16.v16i16.v16i16(<16 x i16>, <16 x i16> , <16 x i16>, i1 immarg) #1
538 declare <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
539 declare <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
540 declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.f16.v8f32.v8f16.v16f16.i16(<8 x half>, <16 x half>, <8 x float>, i16)
541 declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf16.v8f32.v8i16.v16i16.i16(<8 x i16>, <16 x i16>, <8 x float>, i16)
542 declare <8 x half> @llvm.amdgcn.swmmac.f16.16x16x32.f16.v8f16.v8f16.v16f16.i16(<8 x half>, <16 x half>, <8 x half>, i16)
543 declare <8 x i16> @llvm.amdgcn.swmmac.bf16.16x16x32.bf16.v8i16.v8i16.v16i16.i16(<8 x i16>, <16 x i16>, <8 x i16>, i16)
544 declare <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)
545 declare <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)
546 declare <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)
547 declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.fp8(<2 x i32>, <4 x i32>, <8 x float>, i16)
548 declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.fp8.bf8(<2 x i32>, <4 x i32>, <8 x float>, i16)
549 declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.fp8(<2 x i32>, <4 x i32>, <8 x float>, i16)
550 declare <8 x float> @llvm.amdgcn.swmmac.f32.16x16x32.bf8.bf8(<2 x i32>, <4 x i32>, <8 x float>, i16)
552 declare <2 x i32> @llvm.amdgcn.global.load.tr.b64.v2i32(ptr addrspace(1))
553 declare <8 x i16> @llvm.amdgcn.global.load.tr.b128.v8i16(ptr addrspace(1))
554 declare <8 x half> @llvm.amdgcn.global.load.tr.b128.v8f16(ptr addrspace(1))
555 declare <8 x bfloat> @llvm.amdgcn.global.load.tr.b128.v8bf16(ptr addrspace(1))
556 declare i32 @llvm.amdgcn.global.load.tr.b64.i32(ptr addrspace(1))
557 declare <4 x i16> @llvm.amdgcn.global.load.tr.b128.v4i16(ptr addrspace(1))
558 declare <4 x half> @llvm.amdgcn.global.load.tr.b128.v4f16(ptr addrspace(1))
559 declare <4 x bfloat> @llvm.amdgcn.global.load.tr.b128.v4bf16(ptr addrspace(1))
561 attributes #0 = { nounwind convergent }
562 attributes #1 = { nounwind readnone convergent }