1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
| ; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN %s
; Test that doing a shift of a pointer with a constant add will be
; folded into the constant offset addressing mode even if the add has
; multiple uses. This is relevant to accessing 2 separate, adjacent
; LDS globals.
declare i32 @llvm.amdgcn.workitem.id.x() #1
@lds0 = addrspace(3) global [512 x float] undef, align 4
@lds1 = addrspace(3) global [512 x float] undef, align 4
; Make sure the (add tid, 2) << 2 gets folded into the ds's offset as (tid << 2) + 8
; GCN-LABEL: {{^}}load_shl_base_lds_0:
; GCN: v_lshlrev_b32_e32 [[PTR:v[0-9]+]], 2, {{v[0-9]+}}
; GCN: ds_read_b32 {{v[0-9]+}}, [[PTR]] offset:8
; GCN: s_endpgm
define amdgpu_kernel void @load_shl_base_lds_0(float addrspace(1)* %out, i32 addrspace(1)* %add_use) #0 {
%tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
%idx.0 = add nsw i32 %tid.x, 2
%arrayidx0 = getelementptr inbounds [512 x float], [512 x float] addrspace(3)* @lds0, i32 0, i32 %idx.0
%val0 = load float, float addrspace(3)* %arrayidx0, align 4
store i32 %idx.0, i32 addrspace(1)* %add_use, align 4
store float %val0, float addrspace(1)* %out
ret void
}
; Make sure once the first use is folded into the addressing mode, the
; remaining add use goes through the normal shl + add constant fold.
; GCN-LABEL: {{^}}load_shl_base_lds_1:
; GCN: v_lshlrev_b32_e32 [[OFS:v[0-9]+]], 2, {{v[0-9]+}}
; TODO: integrate into the ds_read_b32 offset using a 16-bit relocation
; GCN: v_add_{{[iu]}}32_e32 [[PTR:v[0-9]+]], vcc, lds0@abs32@lo, [[OFS]]
; GCN: ds_read_b32 [[RESULT:v[0-9]+]], [[PTR]] offset:8
; GCN: v_add_{{[iu]}}32_e32 [[ADDUSE:v[0-9]+]], vcc, 8, v{{[0-9]+}}
; GCN-DAG: buffer_store_dword [[RESULT]]
; GCN-DAG: buffer_store_dword [[ADDUSE]]
; GCN: s_endpgm
define amdgpu_kernel void @load_shl_base_lds_1(float addrspace(1)* %out, i32 addrspace(1)* %add_use) #0 {
%tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
%idx.0 = add nsw i32 %tid.x, 2
%arrayidx0 = getelementptr inbounds [512 x float], [512 x float] addrspace(3)* @lds0, i32 0, i32 %idx.0
%val0 = load float, float addrspace(3)* %arrayidx0, align 4
%shl_add_use = shl i32 %idx.0, 2
store i32 %shl_add_use, i32 addrspace(1)* %add_use, align 4
store float %val0, float addrspace(1)* %out
ret void
}
@maxlds = addrspace(3) global [65536 x i8] undef, align 4
; GCN-LABEL: {{^}}load_shl_base_lds_max_offset
; GCN: ds_read_u8 v{{[0-9]+}}, v{{[0-9]+}} offset:65535
; GCN: s_endpgm
define amdgpu_kernel void @load_shl_base_lds_max_offset(i8 addrspace(1)* %out, i8 addrspace(3)* %lds, i32 addrspace(1)* %add_use) #0 {
%tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
%idx.0 = add nsw i32 %tid.x, 65535
%arrayidx0 = getelementptr inbounds [65536 x i8], [65536 x i8] addrspace(3)* @maxlds, i32 0, i32 %idx.0
%val0 = load i8, i8 addrspace(3)* %arrayidx0
store i32 %idx.0, i32 addrspace(1)* %add_use
store i8 %val0, i8 addrspace(1)* %out
ret void
}
; The two globals are placed adjacent in memory, so the same base
; pointer can be used with an offset into the second one.
; TODO: Recover the optimization of using ds_read2st64_b32 using alignment hints
; GCN-LABEL: {{^}}load_shl_base_lds_2:
; GCN: v_lshlrev_b32_e32 [[OFS:v[0-9]+]], 2, {{v[0-9]+}}
; GCN-DAG: v_add_{{[iu]}}32_e32 [[PTR0:v[0-9]+]], vcc, lds0@abs32@lo, [[OFS]]
; GCN-DAG: v_add_{{[iu]}}32_e32 [[PTR1:v[0-9]+]], vcc, lds1@abs32@lo, [[OFS]]
; GCN: s_mov_b32 m0, -1
; GCN-DAG: ds_read_b32 {{v[0-9]+}}, [[PTR0]] offset:256
; GCN-DAG: ds_read_b32 {{v[0-9]+}}, [[PTR1]] offset:256
; TODO: ds_read2st64_b32 {{v\[[0-9]+:[0-9]+\]}}, [[PTR]] offset0:1 offset1:9
; GCN: s_endpgm
define amdgpu_kernel void @load_shl_base_lds_2(float addrspace(1)* %out) #0 {
%tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
%idx.0 = add nsw i32 %tid.x, 64
%arrayidx0 = getelementptr inbounds [512 x float], [512 x float] addrspace(3)* @lds0, i32 0, i32 %idx.0
%val0 = load float, float addrspace(3)* %arrayidx0, align 4
%arrayidx1 = getelementptr inbounds [512 x float], [512 x float] addrspace(3)* @lds1, i32 0, i32 %idx.0
%val1 = load float, float addrspace(3)* %arrayidx1, align 4
%sum = fadd float %val0, %val1
store float %sum, float addrspace(1)* %out, align 4
ret void
}
; GCN-LABEL: {{^}}store_shl_base_lds_0:
; GCN: v_lshlrev_b32_e32 [[PTR:v[0-9]+]], 2, {{v[0-9]+}}
; GCN: ds_write_b32 [[PTR]], {{v[0-9]+}} offset:8
; GCN: s_endpgm
define amdgpu_kernel void @store_shl_base_lds_0(float addrspace(1)* %out, i32 addrspace(1)* %add_use) #0 {
%tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
%idx.0 = add nsw i32 %tid.x, 2
%arrayidx0 = getelementptr inbounds [512 x float], [512 x float] addrspace(3)* @lds0, i32 0, i32 %idx.0
store float 1.0, float addrspace(3)* %arrayidx0, align 4
store i32 %idx.0, i32 addrspace(1)* %add_use, align 4
ret void
}
; --------------------------------------------------------------------------------
; Atomics.
@lds2 = addrspace(3) global [512 x i32] undef, align 4
; define amdgpu_kernel void @atomic_load_shl_base_lds_0(i32 addrspace(1)* %out, i32 addrspace(1)* %add_use) #0 {
; %tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
; %idx.0 = add nsw i32 %tid.x, 2
; %arrayidx0 = getelementptr inbounds [512 x i32], [512 x i32] addrspace(3)* @lds2, i32 0, i32 %idx.0
; %val = load atomic i32, i32 addrspace(3)* %arrayidx0 seq_cst, align 4
; store i32 %val, i32 addrspace(1)* %out, align 4
; store i32 %idx.0, i32 addrspace(1)* %add_use, align 4
; ret void
; }
; GCN-LABEL: {{^}}atomic_cmpxchg_shl_base_lds_0:
; GCN: v_lshlrev_b32_e32 [[PTR:v[0-9]+]], 2, {{v[0-9]+}}
; GCN: ds_cmpst_rtn_b32 {{v[0-9]+}}, [[PTR]], {{v[0-9]+}}, {{v[0-9]+}} offset:8
; GCN: s_endpgm
define amdgpu_kernel void @atomic_cmpxchg_shl_base_lds_0(i32 addrspace(1)* %out, i32 addrspace(1)* %add_use, i32 %swap) #0 {
%tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
%idx.0 = add nsw i32 %tid.x, 2
%arrayidx0 = getelementptr inbounds [512 x i32], [512 x i32] addrspace(3)* @lds2, i32 0, i32 %idx.0
%pair = cmpxchg i32 addrspace(3)* %arrayidx0, i32 7, i32 %swap seq_cst monotonic
%result = extractvalue { i32, i1 } %pair, 0
store i32 %result, i32 addrspace(1)* %out, align 4
store i32 %idx.0, i32 addrspace(1)* %add_use, align 4
ret void
}
; GCN-LABEL: {{^}}atomic_swap_shl_base_lds_0:
; GCN: v_lshlrev_b32_e32 [[PTR:v[0-9]+]], 2, {{v[0-9]+}}
; GCN: ds_wrxchg_rtn_b32 {{v[0-9]+}}, [[PTR]], {{v[0-9]+}} offset:8
; GCN: s_endpgm
define amdgpu_kernel void @atomic_swap_shl_base_lds_0(i32 addrspace(1)* %out, i32 addrspace(1)* %add_use) #0 {
%tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
%idx.0 = add nsw i32 %tid.x, 2
%arrayidx0 = getelementptr inbounds [512 x i32], [512 x i32] addrspace(3)* @lds2, i32 0, i32 %idx.0
%val = atomicrmw xchg i32 addrspace(3)* %arrayidx0, i32 3 seq_cst
store i32 %val, i32 addrspace(1)* %out, align 4
store i32 %idx.0, i32 addrspace(1)* %add_use, align 4
ret void
}
; GCN-LABEL: {{^}}atomic_add_shl_base_lds_0:
; GCN: v_lshlrev_b32_e32 [[PTR:v[0-9]+]], 2, {{v[0-9]+}}
; GCN: ds_add_rtn_u32 {{v[0-9]+}}, [[PTR]], {{v[0-9]+}} offset:8
; GCN: s_endpgm
define amdgpu_kernel void @atomic_add_shl_base_lds_0(i32 addrspace(1)* %out, i32 addrspace(1)* %add_use) #0 {
%tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
%idx.0 = add nsw i32 %tid.x, 2
%arrayidx0 = getelementptr inbounds [512 x i32], [512 x i32] addrspace(3)* @lds2, i32 0, i32 %idx.0
%val = atomicrmw add i32 addrspace(3)* %arrayidx0, i32 3 seq_cst
store i32 %val, i32 addrspace(1)* %out, align 4
store i32 %idx.0, i32 addrspace(1)* %add_use, align 4
ret void
}
; GCN-LABEL: {{^}}atomic_sub_shl_base_lds_0:
; GCN: v_lshlrev_b32_e32 [[PTR:v[0-9]+]], 2, {{v[0-9]+}}
; GCN: ds_sub_rtn_u32 {{v[0-9]+}}, [[PTR]], {{v[0-9]+}} offset:8
; GCN: s_endpgm
define amdgpu_kernel void @atomic_sub_shl_base_lds_0(i32 addrspace(1)* %out, i32 addrspace(1)* %add_use) #0 {
%tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
%idx.0 = add nsw i32 %tid.x, 2
%arrayidx0 = getelementptr inbounds [512 x i32], [512 x i32] addrspace(3)* @lds2, i32 0, i32 %idx.0
%val = atomicrmw sub i32 addrspace(3)* %arrayidx0, i32 3 seq_cst
store i32 %val, i32 addrspace(1)* %out, align 4
store i32 %idx.0, i32 addrspace(1)* %add_use, align 4
ret void
}
; GCN-LABEL: {{^}}atomic_and_shl_base_lds_0:
; GCN: v_lshlrev_b32_e32 [[PTR:v[0-9]+]], 2, {{v[0-9]+}}
; GCN: ds_and_rtn_b32 {{v[0-9]+}}, [[PTR]], {{v[0-9]+}} offset:8
; GCN: s_endpgm
define amdgpu_kernel void @atomic_and_shl_base_lds_0(i32 addrspace(1)* %out, i32 addrspace(1)* %add_use) #0 {
%tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
%idx.0 = add nsw i32 %tid.x, 2
%arrayidx0 = getelementptr inbounds [512 x i32], [512 x i32] addrspace(3)* @lds2, i32 0, i32 %idx.0
%val = atomicrmw and i32 addrspace(3)* %arrayidx0, i32 3 seq_cst
store i32 %val, i32 addrspace(1)* %out, align 4
store i32 %idx.0, i32 addrspace(1)* %add_use, align 4
ret void
}
; GCN-LABEL: {{^}}atomic_or_shl_base_lds_0:
; GCN: v_lshlrev_b32_e32 [[PTR:v[0-9]+]], 2, {{v[0-9]+}}
; GCN: ds_or_rtn_b32 {{v[0-9]+}}, [[PTR]], {{v[0-9]+}} offset:8
; GCN: s_endpgm
define amdgpu_kernel void @atomic_or_shl_base_lds_0(i32 addrspace(1)* %out, i32 addrspace(1)* %add_use) #0 {
%tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
%idx.0 = add nsw i32 %tid.x, 2
%arrayidx0 = getelementptr inbounds [512 x i32], [512 x i32] addrspace(3)* @lds2, i32 0, i32 %idx.0
%val = atomicrmw or i32 addrspace(3)* %arrayidx0, i32 3 seq_cst
store i32 %val, i32 addrspace(1)* %out, align 4
store i32 %idx.0, i32 addrspace(1)* %add_use, align 4
ret void
}
; GCN-LABEL: {{^}}atomic_xor_shl_base_lds_0:
; GCN: v_lshlrev_b32_e32 [[PTR:v[0-9]+]], 2, {{v[0-9]+}}
; GCN: ds_xor_rtn_b32 {{v[0-9]+}}, [[PTR]], {{v[0-9]+}} offset:8
; GCN: s_endpgm
define amdgpu_kernel void @atomic_xor_shl_base_lds_0(i32 addrspace(1)* %out, i32 addrspace(1)* %add_use) #0 {
%tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
%idx.0 = add nsw i32 %tid.x, 2
%arrayidx0 = getelementptr inbounds [512 x i32], [512 x i32] addrspace(3)* @lds2, i32 0, i32 %idx.0
%val = atomicrmw xor i32 addrspace(3)* %arrayidx0, i32 3 seq_cst
store i32 %val, i32 addrspace(1)* %out, align 4
store i32 %idx.0, i32 addrspace(1)* %add_use, align 4
ret void
}
; define amdgpu_kernel void @atomic_nand_shl_base_lds_0(i32 addrspace(1)* %out, i32 addrspace(1)* %add_use) #0 {
; %tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
; %idx.0 = add nsw i32 %tid.x, 2
; %arrayidx0 = getelementptr inbounds [512 x i32], [512 x i32] addrspace(3)* @lds2, i32 0, i32 %idx.0
; %val = atomicrmw nand i32 addrspace(3)* %arrayidx0, i32 3 seq_cst
; store i32 %val, i32 addrspace(1)* %out, align 4
; store i32 %idx.0, i32 addrspace(1)* %add_use, align 4
; ret void
; }
; GCN-LABEL: {{^}}atomic_min_shl_base_lds_0:
; GCN: v_lshlrev_b32_e32 [[PTR:v[0-9]+]], 2, {{v[0-9]+}}
; GCN: ds_min_rtn_i32 {{v[0-9]+}}, [[PTR]], {{v[0-9]+}} offset:8
; GCN: s_endpgm
define amdgpu_kernel void @atomic_min_shl_base_lds_0(i32 addrspace(1)* %out, i32 addrspace(1)* %add_use) #0 {
%tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
%idx.0 = add nsw i32 %tid.x, 2
%arrayidx0 = getelementptr inbounds [512 x i32], [512 x i32] addrspace(3)* @lds2, i32 0, i32 %idx.0
%val = atomicrmw min i32 addrspace(3)* %arrayidx0, i32 3 seq_cst
store i32 %val, i32 addrspace(1)* %out, align 4
store i32 %idx.0, i32 addrspace(1)* %add_use, align 4
ret void
}
; GCN-LABEL: {{^}}atomic_max_shl_base_lds_0:
; GCN: v_lshlrev_b32_e32 [[PTR:v[0-9]+]], 2, {{v[0-9]+}}
; GCN: ds_max_rtn_i32 {{v[0-9]+}}, [[PTR]], {{v[0-9]+}} offset:8
; GCN: s_endpgm
define amdgpu_kernel void @atomic_max_shl_base_lds_0(i32 addrspace(1)* %out, i32 addrspace(1)* %add_use) #0 {
%tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
%idx.0 = add nsw i32 %tid.x, 2
%arrayidx0 = getelementptr inbounds [512 x i32], [512 x i32] addrspace(3)* @lds2, i32 0, i32 %idx.0
%val = atomicrmw max i32 addrspace(3)* %arrayidx0, i32 3 seq_cst
store i32 %val, i32 addrspace(1)* %out, align 4
store i32 %idx.0, i32 addrspace(1)* %add_use, align 4
ret void
}
; GCN-LABEL: {{^}}atomic_umin_shl_base_lds_0:
; GCN: v_lshlrev_b32_e32 [[PTR:v[0-9]+]], 2, {{v[0-9]+}}
; GCN: ds_min_rtn_u32 {{v[0-9]+}}, [[PTR]], {{v[0-9]+}} offset:8
; GCN: s_endpgm
define amdgpu_kernel void @atomic_umin_shl_base_lds_0(i32 addrspace(1)* %out, i32 addrspace(1)* %add_use) #0 {
%tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
%idx.0 = add nsw i32 %tid.x, 2
%arrayidx0 = getelementptr inbounds [512 x i32], [512 x i32] addrspace(3)* @lds2, i32 0, i32 %idx.0
%val = atomicrmw umin i32 addrspace(3)* %arrayidx0, i32 3 seq_cst
store i32 %val, i32 addrspace(1)* %out, align 4
store i32 %idx.0, i32 addrspace(1)* %add_use, align 4
ret void
}
; GCN-LABEL: {{^}}atomic_umax_shl_base_lds_0:
; GCN: v_lshlrev_b32_e32 [[PTR:v[0-9]+]], 2, {{v[0-9]+}}
; GCN: ds_max_rtn_u32 {{v[0-9]+}}, [[PTR]], {{v[0-9]+}} offset:8
; GCN: s_endpgm
define amdgpu_kernel void @atomic_umax_shl_base_lds_0(i32 addrspace(1)* %out, i32 addrspace(1)* %add_use) #0 {
%tid.x = tail call i32 @llvm.amdgcn.workitem.id.x() #1
%idx.0 = add nsw i32 %tid.x, 2
%arrayidx0 = getelementptr inbounds [512 x i32], [512 x i32] addrspace(3)* @lds2, i32 0, i32 %idx.0
%val = atomicrmw umax i32 addrspace(3)* %arrayidx0, i32 3 seq_cst
store i32 %val, i32 addrspace(1)* %out, align 4
store i32 %idx.0, i32 addrspace(1)* %add_use, align 4
ret void
}
; GCN-LABEL: {{^}}shl_add_ptr_combine_2use_lds:
; GCN: v_lshlrev_b32_e32 [[SCALE0:v[0-9]+]], 3, v0
; GCN: ds_write_b32 [[SCALE0]], v{{[0-9]+}} offset:32
; GCN: v_lshlrev_b32_e32 [[SCALE1:v[0-9]+]], 4, v0
; GCN: ds_write_b32 [[SCALE1]], v{{[0-9]+}} offset:64
define void @shl_add_ptr_combine_2use_lds(i32 %idx) #0 {
%idx.add = add nuw i32 %idx, 4
%shl0 = shl i32 %idx.add, 3
%shl1 = shl i32 %idx.add, 4
%ptr0 = inttoptr i32 %shl0 to i32 addrspace(3)*
%ptr1 = inttoptr i32 %shl1 to i32 addrspace(3)*
store volatile i32 9, i32 addrspace(3)* %ptr0
store volatile i32 10, i32 addrspace(3)* %ptr1
ret void
}
; GCN-LABEL: {{^}}shl_add_ptr_combine_2use_max_lds_offset:
; GCN-DAG: v_lshlrev_b32_e32 [[SCALE0:v[0-9]+]], 3, v0
; GCN-DAG: v_lshlrev_b32_e32 [[SCALE1:v[0-9]+]], 4, v0
; GCN-DAG: ds_write_b32 [[SCALE0]], v{{[0-9]+}} offset:65528
; GCN-DAG: v_add_{{[iu]}}32_e32 [[ADD1:v[0-9]+]], vcc, 0x1fff0, [[SCALE1]]
; GCN: ds_write_b32 [[ADD1]], v{{[0-9]+$}}
define void @shl_add_ptr_combine_2use_max_lds_offset(i32 %idx) #0 {
%idx.add = add nuw i32 %idx, 8191
%shl0 = shl i32 %idx.add, 3
%shl1 = shl i32 %idx.add, 4
%ptr0 = inttoptr i32 %shl0 to i32 addrspace(3)*
%ptr1 = inttoptr i32 %shl1 to i32 addrspace(3)*
store volatile i32 9, i32 addrspace(3)* %ptr0
store volatile i32 10, i32 addrspace(3)* %ptr1
ret void
}
; GCN-LABEL: {{^}}shl_add_ptr_combine_2use_both_max_lds_offset:
; GCN: v_add_{{[iu]}}32_e32 [[ADD:v[0-9]+]], vcc, 0x1000, v0
; GCN-DAG: v_lshlrev_b32_e32 [[SCALE0:v[0-9]+]], 4, [[ADD]]
; GCN-DAG: v_lshlrev_b32_e32 [[SCALE1:v[0-9]+]], 5, [[ADD]]
; GCN-DAG: ds_write_b32 [[SCALE0]], v{{[0-9]+$}}
; GCN: ds_write_b32 [[SCALE1]], v{{[0-9]+$}}
define void @shl_add_ptr_combine_2use_both_max_lds_offset(i32 %idx) #0 {
%idx.add = add nuw i32 %idx, 4096
%shl0 = shl i32 %idx.add, 4
%shl1 = shl i32 %idx.add, 5
%ptr0 = inttoptr i32 %shl0 to i32 addrspace(3)*
%ptr1 = inttoptr i32 %shl1 to i32 addrspace(3)*
store volatile i32 9, i32 addrspace(3)* %ptr0
store volatile i32 10, i32 addrspace(3)* %ptr1
ret void
}
; GCN-LABEL: {{^}}shl_add_ptr_combine_2use_private:
; GCN: v_lshlrev_b32_e32 [[SCALE0:v[0-9]+]], 2, v0
; GCN: buffer_store_dword v{{[0-9]+}}, [[SCALE0]], s[0:3], s33 offen offset:16
; GCN: v_lshlrev_b32_e32 [[SCALE1:v[0-9]+]], 3, v0
; GCN: buffer_store_dword v{{[0-9]+}}, [[SCALE1]], s[0:3], s33 offen offset:32
define void @shl_add_ptr_combine_2use_private(i16 zeroext %idx.arg) #0 {
%idx = zext i16 %idx.arg to i32
%idx.add = add nuw i32 %idx, 4
%shl0 = shl i32 %idx.add, 2
%shl1 = shl i32 %idx.add, 3
%ptr0 = inttoptr i32 %shl0 to i32 addrspace(5)*
%ptr1 = inttoptr i32 %shl1 to i32 addrspace(5)*
store volatile i32 9, i32 addrspace(5)* %ptr0
store volatile i32 10, i32 addrspace(5)* %ptr1
ret void
}
; GCN-LABEL: {{^}}shl_add_ptr_combine_2use_max_private_offset:
; GCN-DAG: v_lshlrev_b32_e32 [[SCALE0:v[0-9]+]], 3, v0
; GCN-DAG: v_lshlrev_b32_e32 [[SCALE1:v[0-9]+]], 4, v0
; GCN-DAG: buffer_store_dword v{{[0-9]+}}, [[SCALE0]], s[0:3], s33 offen offset:4088
; GCN-DAG: v_add_{{[iu]}}32_e32 [[ADD:v[0-9]+]], vcc, 0x1ff0, [[SCALE1]]
; GCN: buffer_store_dword v{{[0-9]+}}, [[ADD]], s[0:3], s33 offen{{$}}
define void @shl_add_ptr_combine_2use_max_private_offset(i16 zeroext %idx.arg) #0 {
%idx = zext i16 %idx.arg to i32
%idx.add = add nuw i32 %idx, 511
%shl0 = shl i32 %idx.add, 3
%shl1 = shl i32 %idx.add, 4
%ptr0 = inttoptr i32 %shl0 to i32 addrspace(5)*
%ptr1 = inttoptr i32 %shl1 to i32 addrspace(5)*
store volatile i32 9, i32 addrspace(5)* %ptr0
store volatile i32 10, i32 addrspace(5)* %ptr1
ret void
}
; GCN-LABEL: {{^}}shl_add_ptr_combine_2use_both_max_private_offset:
; GCN: v_add_{{[iu]}}32_e32 [[ADD:v[0-9]+]], vcc, 0x100, v0
; GCN-DAG: v_lshlrev_b32_e32 [[SCALE0:v[0-9]+]], 4, [[ADD]]
; GCN-DAG: v_lshlrev_b32_e32 [[SCALE1:v[0-9]+]], 5, [[ADD]]
; GCN-DAG: buffer_store_dword v{{[0-9]+}}, [[SCALE0]], s[0:3], s33 offen{{$}}
; GCN: buffer_store_dword v{{[0-9]+}}, [[SCALE1]], s[0:3], s33 offen{{$}}
define void @shl_add_ptr_combine_2use_both_max_private_offset(i16 zeroext %idx.arg) #0 {
%idx = zext i16 %idx.arg to i32
%idx.add = add nuw i32 %idx, 256
%shl0 = shl i32 %idx.add, 4
%shl1 = shl i32 %idx.add, 5
%ptr0 = inttoptr i32 %shl0 to i32 addrspace(5)*
%ptr1 = inttoptr i32 %shl1 to i32 addrspace(5)*
store volatile i32 9, i32 addrspace(5)* %ptr0
store volatile i32 10, i32 addrspace(5)* %ptr1
ret void
}
; FIXME: This or should fold into an offset on the write
; GCN-LABEL: {{^}}shl_or_ptr_combine_2use_lds:
; GCN: v_lshlrev_b32_e32 [[SCALE0:v[0-9]+]], 3, v0
; GCN: v_or_b32_e32 [[SCALE1:v[0-9]+]], 32, [[SCALE0]]
; GCN: ds_write_b32 [[SCALE1]], v{{[0-9]+}}
; GCN: v_lshlrev_b32_e32 [[SCALE1:v[0-9]+]], 4, v0
; GCN: ds_write_b32 [[SCALE1]], v{{[0-9]+}} offset:64
define void @shl_or_ptr_combine_2use_lds(i32 %idx) #0 {
%idx.add = or i32 %idx, 4
%shl0 = shl i32 %idx.add, 3
%shl1 = shl i32 %idx.add, 4
%ptr0 = inttoptr i32 %shl0 to i32 addrspace(3)*
%ptr1 = inttoptr i32 %shl1 to i32 addrspace(3)*
store volatile i32 9, i32 addrspace(3)* %ptr0
store volatile i32 10, i32 addrspace(3)* %ptr1
ret void
}
; GCN-LABEL: {{^}}shl_or_ptr_combine_2use_max_lds_offset:
; GCN-DAG: v_lshlrev_b32_e32 [[SCALE0:v[0-9]+]], 3, v0
; GCN-DAG: v_lshlrev_b32_e32 [[SCALE1:v[0-9]+]], 4, v0
; GCN-DAG: ds_write_b32 [[SCALE0]], v{{[0-9]+}} offset:65528
; GCN-DAG: v_or_b32_e32 [[ADD1:v[0-9]+]], 0x1fff0, [[SCALE1]]
; GCN: ds_write_b32 [[ADD1]], v{{[0-9]+$}}
define void @shl_or_ptr_combine_2use_max_lds_offset(i32 %idx) #0 {
%idx.add = or i32 %idx, 8191
%shl0 = shl i32 %idx.add, 3
%shl1 = shl i32 %idx.add, 4
%ptr0 = inttoptr i32 %shl0 to i32 addrspace(3)*
%ptr1 = inttoptr i32 %shl1 to i32 addrspace(3)*
store volatile i32 9, i32 addrspace(3)* %ptr0
store volatile i32 10, i32 addrspace(3)* %ptr1
ret void
}
attributes #0 = { nounwind }
attributes #1 = { nounwind readnone }
|