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
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
; RUN: llc -global-isel -mtriple=amdgcn-amd-amdpal -amdgpu-global-isel-risky-select -mcpu=gfx1010 < %s | FileCheck -check-prefix=GFX10 %s
; Simples case, if - then, that requires lane mask merging,
; %phi lane mask will hold %val_A at %A. Lanes that are active in %B
; will overwrite its own lane bit in lane mask with val_B
define amdgpu_ps void @divergent_i1_phi_if_then(ptr addrspace(1) %out, i32 %tid, i32 %cond) {
; GFX10-LABEL: divergent_i1_phi_if_then:
; GFX10: ; %bb.0: ; %A
; GFX10-NEXT: v_cmp_le_u32_e64 s0, 6, v2
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v3
; GFX10-NEXT: s_and_saveexec_b32 s1, vcc_lo
; GFX10-NEXT: ; %bb.1: ; %B
; GFX10-NEXT: v_cmp_gt_u32_e64 s0, 1, v2
; GFX10-NEXT: ; %bb.2: ; %exit
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s1
; GFX10-NEXT: v_cndmask_b32_e64 v2, 2, 1, s0
; GFX10-NEXT: global_store_dword v[0:1], v2, off
; GFX10-NEXT: s_endpgm
A:
%val_A = icmp uge i32 %tid, 6
%cmp = icmp eq i32 %cond, 0
br i1 %cmp, label %B, label %exit
B:
%val_B = icmp ult i32 %tid, 1
br label %exit
exit:
%phi = phi i1 [ %val_A, %A ], [ %val_B, %B ]
%sel = select i1 %phi, i32 1, i32 2
store i32 %sel, ptr addrspace(1) %out
ret void
}
; if - else
define amdgpu_ps void @divergent_i1_phi_if_else(ptr addrspace(1) %out, i32 %tid, i32 %cond) {
; GFX10-LABEL: divergent_i1_phi_if_else:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_and_b32 s0, 1, s0
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v3
; GFX10-NEXT: v_cmp_ne_u32_e64 s0, 0, s0
; GFX10-NEXT: s_and_saveexec_b32 s1, vcc_lo
; GFX10-NEXT: s_xor_b32 s1, exec_lo, s1
; GFX10-NEXT: ; %bb.1: ; %B
; GFX10-NEXT: v_cmp_gt_u32_e64 s0, 2, v2
; GFX10-NEXT: ; implicit-def: $vgpr2
; GFX10-NEXT: ; %bb.2: ; %Flow
; GFX10-NEXT: s_andn2_saveexec_b32 s1, s1
; GFX10-NEXT: ; %bb.3: ; %A
; GFX10-NEXT: v_cmp_le_u32_e64 s0, 1, v2
; GFX10-NEXT: ; %bb.4: ; %exit
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s1
; GFX10-NEXT: v_cndmask_b32_e64 v2, 2, 1, s0
; GFX10-NEXT: global_store_dword v[0:1], v2, off
; GFX10-NEXT: s_endpgm
entry:
%cmp = icmp eq i32 %cond, 0
br i1 %cmp, label %A, label %B
A:
%val_A = icmp uge i32 %tid, 1
br label %exit
B:
%val_B = icmp ult i32 %tid, 2
br label %exit
exit:
%phi = phi i1 [ %val_A, %A ], [ %val_B, %B ]
%sel = select i1 %phi, i32 1, i32 2
store i32 %sel, ptr addrspace(1) %out
ret void
}
; if - break;
; counter = 0;
; do {
; if (a[counter] == 0)
; break;
; if (b[counter] == 0)
; break;
; if (c[counter] == 0)
; break;
; x[counter++]+=1;
; } while (counter<100);
; Tests with multiple break conditions. Divergent phis will be used to track
; if any of the break conditions was reached. We only need to do simple lane
; mask merging (for current loop iteration only). There is an intrinsic,
; if_break, that will merge lane masks across all iterations of the loop.
define amdgpu_cs void @loop_with_1break(ptr addrspace(1) %x, ptr addrspace(1) %a) {
; GFX10-LABEL: loop_with_1break:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_mov_b32 s0, 0
; GFX10-NEXT: v_mov_b32_e32 v4, s0
; GFX10-NEXT: s_branch .LBB2_2
; GFX10-NEXT: .LBB2_1: ; %Flow
; GFX10-NEXT: ; in Loop: Header=BB2_2 Depth=1
; GFX10-NEXT: s_waitcnt_depctr 0xffe3
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s1
; GFX10-NEXT: s_and_b32 s1, exec_lo, s2
; GFX10-NEXT: s_or_b32 s0, s1, s0
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s0
; GFX10-NEXT: s_cbranch_execz .LBB2_4
; GFX10-NEXT: .LBB2_2: ; %A
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: v_ashrrev_i32_e32 v5, 31, v4
; GFX10-NEXT: s_mov_b32 s2, -1
; GFX10-NEXT: v_lshlrev_b64 v[5:6], 2, v[4:5]
; GFX10-NEXT: v_add_co_u32 v7, vcc_lo, v2, v5
; GFX10-NEXT: v_add_co_ci_u32_e32 v8, vcc_lo, v3, v6, vcc_lo
; GFX10-NEXT: global_load_dword v7, v[7:8], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v7
; GFX10-NEXT: s_and_saveexec_b32 s1, vcc_lo
; GFX10-NEXT: s_cbranch_execz .LBB2_1
; GFX10-NEXT: ; %bb.3: ; %loop.body
; GFX10-NEXT: ; in Loop: Header=BB2_2 Depth=1
; GFX10-NEXT: v_add_co_u32 v5, vcc_lo, v0, v5
; GFX10-NEXT: v_add_co_ci_u32_e32 v6, vcc_lo, v1, v6, vcc_lo
; GFX10-NEXT: v_add_nc_u32_e32 v8, 1, v4
; GFX10-NEXT: v_cmp_gt_u32_e64 s2, 0x64, v4
; GFX10-NEXT: global_load_dword v7, v[5:6], off
; GFX10-NEXT: v_mov_b32_e32 v4, v8
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_add_nc_u32_e32 v7, 1, v7
; GFX10-NEXT: global_store_dword v[5:6], v7, off
; GFX10-NEXT: s_branch .LBB2_1
; GFX10-NEXT: .LBB2_4: ; %exit
; GFX10-NEXT: s_endpgm
entry:
br label %A
A:
%counter = phi i32 [ %counter.plus.1, %loop.body ], [ 0, %entry ]
%a.plus.counter = getelementptr inbounds i32, ptr addrspace(1) %a, i32 %counter
%a.val = load i32, ptr addrspace(1) %a.plus.counter
%a.cond = icmp eq i32 %a.val, 0
br i1 %a.cond, label %exit, label %loop.body
loop.body:
%x.plus.counter = getelementptr inbounds i32, ptr addrspace(1) %x, i32 %counter
%x.val = load i32, ptr addrspace(1) %x.plus.counter
%x.val.plus.1 = add i32 %x.val, 1
store i32 %x.val.plus.1, ptr addrspace(1) %x.plus.counter
%counter.plus.1 = add i32 %counter, 1
%x.cond = icmp ult i32 %counter, 100
br i1 %x.cond, label %exit, label %A
exit:
ret void
}
define amdgpu_cs void @loop_with_2breaks(ptr addrspace(1) %x, ptr addrspace(1) %a, ptr addrspace(1) %b) {
; GFX10-LABEL: loop_with_2breaks:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_mov_b32 s0, 0
; GFX10-NEXT: v_mov_b32_e32 v6, s0
; GFX10-NEXT: s_branch .LBB3_3
; GFX10-NEXT: .LBB3_1: ; %Flow3
; GFX10-NEXT: ; in Loop: Header=BB3_3 Depth=1
; GFX10-NEXT: s_waitcnt_depctr 0xffe3
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s3
; GFX10-NEXT: .LBB3_2: ; %Flow
; GFX10-NEXT: ; in Loop: Header=BB3_3 Depth=1
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s1
; GFX10-NEXT: s_and_b32 s1, exec_lo, s2
; GFX10-NEXT: s_or_b32 s0, s1, s0
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s0
; GFX10-NEXT: s_cbranch_execz .LBB3_6
; GFX10-NEXT: .LBB3_3: ; %A
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: v_ashrrev_i32_e32 v7, 31, v6
; GFX10-NEXT: s_mov_b32 s2, -1
; GFX10-NEXT: v_lshlrev_b64 v[7:8], 2, v[6:7]
; GFX10-NEXT: v_add_co_u32 v9, vcc_lo, v2, v7
; GFX10-NEXT: v_add_co_ci_u32_e32 v10, vcc_lo, v3, v8, vcc_lo
; GFX10-NEXT: global_load_dword v9, v[9:10], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v9
; GFX10-NEXT: s_and_saveexec_b32 s1, vcc_lo
; GFX10-NEXT: s_cbranch_execz .LBB3_2
; GFX10-NEXT: ; %bb.4: ; %B
; GFX10-NEXT: ; in Loop: Header=BB3_3 Depth=1
; GFX10-NEXT: v_add_co_u32 v9, vcc_lo, v4, v7
; GFX10-NEXT: v_add_co_ci_u32_e32 v10, vcc_lo, v5, v8, vcc_lo
; GFX10-NEXT: global_load_dword v9, v[9:10], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v9
; GFX10-NEXT: s_and_saveexec_b32 s3, vcc_lo
; GFX10-NEXT: s_cbranch_execz .LBB3_1
; GFX10-NEXT: ; %bb.5: ; %loop.body
; GFX10-NEXT: ; in Loop: Header=BB3_3 Depth=1
; GFX10-NEXT: v_add_co_u32 v7, vcc_lo, v0, v7
; GFX10-NEXT: v_add_co_ci_u32_e32 v8, vcc_lo, v1, v8, vcc_lo
; GFX10-NEXT: v_add_nc_u32_e32 v10, 1, v6
; GFX10-NEXT: v_cmp_gt_u32_e64 s2, 0x64, v6
; GFX10-NEXT: global_load_dword v9, v[7:8], off
; GFX10-NEXT: v_mov_b32_e32 v6, v10
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_add_nc_u32_e32 v9, 1, v9
; GFX10-NEXT: global_store_dword v[7:8], v9, off
; GFX10-NEXT: s_branch .LBB3_1
; GFX10-NEXT: .LBB3_6: ; %exit
; GFX10-NEXT: s_endpgm
entry:
br label %A
A:
%counter = phi i32 [ %counter.plus.1, %loop.body ], [ 0, %entry ]
%a.plus.counter = getelementptr inbounds i32, ptr addrspace(1) %a, i32 %counter
%a.val = load i32, ptr addrspace(1) %a.plus.counter
%a.cond = icmp eq i32 %a.val, 0
br i1 %a.cond, label %exit, label %B
B:
%b.plus.counter = getelementptr inbounds i32, ptr addrspace(1) %b, i32 %counter
%b.val = load i32, ptr addrspace(1) %b.plus.counter
%b.cond = icmp eq i32 %b.val, 0
br i1 %b.cond, label %exit, label %loop.body
loop.body:
%x.plus.counter = getelementptr inbounds i32, ptr addrspace(1) %x, i32 %counter
%x.val = load i32, ptr addrspace(1) %x.plus.counter
%x.val.plus.1 = add i32 %x.val, 1
store i32 %x.val.plus.1, ptr addrspace(1) %x.plus.counter
%counter.plus.1 = add i32 %counter, 1
%x.cond = icmp ult i32 %counter, 100
br i1 %x.cond, label %exit, label %A
exit:
ret void
}
define amdgpu_cs void @loop_with_3breaks(ptr addrspace(1) %x, ptr addrspace(1) %a, ptr addrspace(1) %b, ptr addrspace(1) %c) {
; GFX10-LABEL: loop_with_3breaks:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_mov_b32 s0, 0
; GFX10-NEXT: v_mov_b32_e32 v8, s0
; GFX10-NEXT: s_branch .LBB4_4
; GFX10-NEXT: .LBB4_1: ; %Flow5
; GFX10-NEXT: ; in Loop: Header=BB4_4 Depth=1
; GFX10-NEXT: s_waitcnt_depctr 0xffe3
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: .LBB4_2: ; %Flow4
; GFX10-NEXT: ; in Loop: Header=BB4_4 Depth=1
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s3
; GFX10-NEXT: .LBB4_3: ; %Flow
; GFX10-NEXT: ; in Loop: Header=BB4_4 Depth=1
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s1
; GFX10-NEXT: s_and_b32 s1, exec_lo, s2
; GFX10-NEXT: s_or_b32 s0, s1, s0
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s0
; GFX10-NEXT: s_cbranch_execz .LBB4_8
; GFX10-NEXT: .LBB4_4: ; %A
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: v_ashrrev_i32_e32 v9, 31, v8
; GFX10-NEXT: s_mov_b32 s2, -1
; GFX10-NEXT: v_lshlrev_b64 v[9:10], 2, v[8:9]
; GFX10-NEXT: v_add_co_u32 v11, vcc_lo, v2, v9
; GFX10-NEXT: v_add_co_ci_u32_e32 v12, vcc_lo, v3, v10, vcc_lo
; GFX10-NEXT: global_load_dword v11, v[11:12], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v11
; GFX10-NEXT: s_and_saveexec_b32 s1, vcc_lo
; GFX10-NEXT: s_cbranch_execz .LBB4_3
; GFX10-NEXT: ; %bb.5: ; %B
; GFX10-NEXT: ; in Loop: Header=BB4_4 Depth=1
; GFX10-NEXT: v_add_co_u32 v11, vcc_lo, v4, v9
; GFX10-NEXT: v_add_co_ci_u32_e32 v12, vcc_lo, v5, v10, vcc_lo
; GFX10-NEXT: global_load_dword v11, v[11:12], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v11
; GFX10-NEXT: s_and_saveexec_b32 s3, vcc_lo
; GFX10-NEXT: s_cbranch_execz .LBB4_2
; GFX10-NEXT: ; %bb.6: ; %C
; GFX10-NEXT: ; in Loop: Header=BB4_4 Depth=1
; GFX10-NEXT: v_add_co_u32 v11, vcc_lo, v6, v9
; GFX10-NEXT: v_add_co_ci_u32_e32 v12, vcc_lo, v7, v10, vcc_lo
; GFX10-NEXT: global_load_dword v11, v[11:12], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v11
; GFX10-NEXT: s_and_saveexec_b32 s4, vcc_lo
; GFX10-NEXT: s_cbranch_execz .LBB4_1
; GFX10-NEXT: ; %bb.7: ; %loop.body
; GFX10-NEXT: ; in Loop: Header=BB4_4 Depth=1
; GFX10-NEXT: v_add_co_u32 v9, vcc_lo, v0, v9
; GFX10-NEXT: v_add_co_ci_u32_e32 v10, vcc_lo, v1, v10, vcc_lo
; GFX10-NEXT: v_add_nc_u32_e32 v12, 1, v8
; GFX10-NEXT: v_cmp_gt_u32_e64 s2, 0x64, v8
; GFX10-NEXT: global_load_dword v11, v[9:10], off
; GFX10-NEXT: v_mov_b32_e32 v8, v12
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_add_nc_u32_e32 v11, 1, v11
; GFX10-NEXT: global_store_dword v[9:10], v11, off
; GFX10-NEXT: s_branch .LBB4_1
; GFX10-NEXT: .LBB4_8: ; %exit
; GFX10-NEXT: s_endpgm
entry:
br label %A
A:
%counter = phi i32 [ %counter.plus.1, %loop.body ], [ 0, %entry ]
%a.plus.counter = getelementptr inbounds i32, ptr addrspace(1) %a, i32 %counter
%a.val = load i32, ptr addrspace(1) %a.plus.counter
%a.cond = icmp eq i32 %a.val, 0
br i1 %a.cond, label %exit, label %B
B:
%b.plus.counter = getelementptr inbounds i32, ptr addrspace(1) %b, i32 %counter
%b.val = load i32, ptr addrspace(1) %b.plus.counter
%b.cond = icmp eq i32 %b.val, 0
br i1 %b.cond, label %exit, label %C
C:
%c.plus.counter = getelementptr inbounds i32, ptr addrspace(1) %c, i32 %counter
%c.val = load i32, ptr addrspace(1) %c.plus.counter
%c.cond = icmp eq i32 %c.val, 0
br i1 %c.cond, label %exit, label %loop.body
loop.body:
%x.plus.counter = getelementptr inbounds i32, ptr addrspace(1) %x, i32 %counter
%x.val = load i32, ptr addrspace(1) %x.plus.counter
%x.val.plus.1 = add i32 %x.val, 1
store i32 %x.val.plus.1, ptr addrspace(1) %x.plus.counter
%counter.plus.1 = add i32 %counter, 1
%x.cond = icmp ult i32 %counter, 100
br i1 %x.cond, label %exit, label %A
exit:
ret void
}
; Divergent condition if with body, ending with break. This is loop with two
; exits but structurizer will create phi that will track exit from break
; and move break.body after the loop. Loop will then have one exit and phi
; used outside of the loop by condition used to enter the break.body.
define amdgpu_cs void @loop_with_div_break_with_body(ptr addrspace(1) %x, ptr addrspace(1) %a, ptr addrspace(1) %a.break) {
; GFX10-LABEL: loop_with_div_break_with_body:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_mov_b32 s0, 0
; GFX10-NEXT: v_mov_b32_e32 v6, s0
; GFX10-NEXT: s_branch .LBB5_2
; GFX10-NEXT: .LBB5_1: ; %Flow
; GFX10-NEXT: ; in Loop: Header=BB5_2 Depth=1
; GFX10-NEXT: s_waitcnt_depctr 0xffe3
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s1
; GFX10-NEXT: s_and_b32 s1, exec_lo, s2
; GFX10-NEXT: s_or_b32 s0, s1, s0
; GFX10-NEXT: s_and_b32 s1, 1, s3
; GFX10-NEXT: v_cmp_ne_u32_e64 s1, 0, s1
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s0
; GFX10-NEXT: s_cbranch_execz .LBB5_4
; GFX10-NEXT: .LBB5_2: ; %A
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: v_ashrrev_i32_e32 v7, 31, v6
; GFX10-NEXT: s_mov_b32 s2, -1
; GFX10-NEXT: s_mov_b32 s3, 1
; GFX10-NEXT: v_lshlrev_b64 v[7:8], 2, v[6:7]
; GFX10-NEXT: v_add_co_u32 v9, vcc_lo, v2, v7
; GFX10-NEXT: v_add_co_ci_u32_e32 v10, vcc_lo, v3, v8, vcc_lo
; GFX10-NEXT: global_load_dword v9, v[9:10], off
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v9
; GFX10-NEXT: s_and_saveexec_b32 s1, vcc_lo
; GFX10-NEXT: s_cbranch_execz .LBB5_1
; GFX10-NEXT: ; %bb.3: ; %loop.body
; GFX10-NEXT: ; in Loop: Header=BB5_2 Depth=1
; GFX10-NEXT: v_add_co_u32 v7, vcc_lo, v0, v7
; GFX10-NEXT: v_add_co_ci_u32_e32 v8, vcc_lo, v1, v8, vcc_lo
; GFX10-NEXT: v_add_nc_u32_e32 v10, 1, v6
; GFX10-NEXT: v_cmp_gt_u32_e64 s2, 0x64, v6
; GFX10-NEXT: s_mov_b32 s3, 0
; GFX10-NEXT: global_load_dword v9, v[7:8], off
; GFX10-NEXT: v_mov_b32_e32 v6, v10
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_add_nc_u32_e32 v9, 1, v9
; GFX10-NEXT: global_store_dword v[7:8], v9, off
; GFX10-NEXT: s_branch .LBB5_1
; GFX10-NEXT: .LBB5_4: ; %loop.exit.guard
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s0
; GFX10-NEXT: s_and_saveexec_b32 s0, s1
; GFX10-NEXT: s_xor_b32 s0, exec_lo, s0
; GFX10-NEXT: s_cbranch_execz .LBB5_6
; GFX10-NEXT: ; %bb.5: ; %break.body
; GFX10-NEXT: v_mov_b32_e32 v0, 10
; GFX10-NEXT: global_store_dword v[4:5], v0, off
; GFX10-NEXT: .LBB5_6: ; %exit
; GFX10-NEXT: s_endpgm
entry:
br label %A
A:
%counter = phi i32 [ %counter.plus.1, %loop.body ], [ 0, %entry ]
%a.plus.counter = getelementptr inbounds i32, ptr addrspace(1) %a, i32 %counter
%a.val = load i32, ptr addrspace(1) %a.plus.counter
%a.cond = icmp eq i32 %a.val, 0
br i1 %a.cond, label %break.body, label %loop.body
break.body:
store i32 10, ptr addrspace(1) %a.break
br label %exit
loop.body:
%x.plus.counter = getelementptr inbounds i32, ptr addrspace(1) %x, i32 %counter
%x.val = load i32, ptr addrspace(1) %x.plus.counter
%x.val.plus.1 = add i32 %x.val, 1
store i32 %x.val.plus.1, ptr addrspace(1) %x.plus.counter
%counter.plus.1 = add i32 %counter, 1
%x.cond = icmp ult i32 %counter, 100
br i1 %x.cond, label %exit, label %A
exit:
ret void
}
; Snippet from test generated by the GraphicsFuzz tool, frontend generates ir
; with irreducible control flow graph. FixIrreducible converts it into natural
; loop and in the process creates i1 phi with three incoming values.
; int loop(int x, int y, int a0, int a1, int a2, int a3, int a4) {
; do {
; if (y < a2) {
; do {
; } while (x < a2);
; }
; if (x < a3) {
; return a1;
; }
; } while (y < a2);
; return a0;
; }
; This test is also interesting because it has phi with three incomings
;define amdgpu_ps i32 @irreducible_cfg(i32 %x, i32 %y, i32 %a0, i32 %a1, i32 %a2, i32 %a3) {
;.entry:
; %.y_lt_a2 = icmp sgt i32 %a2, %y
; %.x_lt_a2 = icmp sgt i32 %a2, %x
; %.x_lt_a3 = icmp sgt i32 %a3, %x
; br i1 %.y_lt_a2, label %.preheader, label %.loopexit ; first iteration, jump to inner loop if 'y < a2' or start with 'if (x < a3)'
;
;.preheader: ; if (y < a2),
; br label %.inner_loop
;
;.inner_loop: ; do while x < a2
; br i1 %.x_lt_a2, label %.inner_loop, label %.loopexit
;
;.loopexit: ; if x < a3
; %not.inner_loop = xor i1 %.y_lt_a2, true
; %brmerge = select i1 %.x_lt_a3, i1 true, i1 %not.inner_loop ; exit loop if 'x < a3' or 'loop ends since !(y < a2)'
; %.ret = select i1 %.x_lt_a3, i32 %a1, i32 %a0 ; select retrun value a1 'x < a3' or a0 'loop ends'
; br i1 %brmerge, label %.exit, label %.preheader
;
;.exit:
; ret i32 %.ret
;}
|
