loop_break.ll
11.8 KB
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
; RUN: opt -mtriple=amdgcn-- -S -structurizecfg -si-annotate-control-flow %s | FileCheck -check-prefix=OPT %s
; RUN: llc -march=amdgcn -verify-machineinstrs -disable-block-placement < %s | FileCheck -check-prefix=GCN %s
; Uses llvm.amdgcn.break
; OPT-LABEL: @break_loop(
; OPT: bb1:
; OPT: icmp slt i32
; OPT-NEXT: br i1 %cmp0, label %bb4, label %Flow
; OPT: bb4:
; OPT: load volatile
; OPT: icmp slt i32
; OPT: xor i1 %cmp1
; OPT: br label %Flow
; OPT: Flow:
; OPT: call i64 @llvm.amdgcn.if.break.i64.i64(
; OPT: call i1 @llvm.amdgcn.loop.i64(i64
; OPT: br i1 %{{[0-9]+}}, label %bb9, label %bb1
; OPT: bb9:
; OPT: call void @llvm.amdgcn.end.cf.i64(i64
; GCN-LABEL: {{^}}break_loop:
; GCN: s_mov_b64 [[ACCUM_MASK:s\[[0-9]+:[0-9]+\]]], 0{{$}}
; GCN: [[LOOP_ENTRY:BB[0-9]+_[0-9]+]]: ; %bb1
; GCN: s_add_i32 s6, s6, 1
; GCN: s_or_b64 [[INNER_MASK:s\[[0-9]+:[0-9]+\]]], [[INNER_MASK]], exec
; GCN: s_cmp_gt_i32 s6, -1
; GCN: s_cbranch_scc1 [[FLOW:BB[0-9]+_[0-9]+]]
; GCN: ; %bb4
; GCN: buffer_load_dword
; GCN: v_cmp_ge_i32_e32 vcc
; GCN: s_andn2_b64 [[INNER_MASK]], [[INNER_MASK]], exec
; GCN: s_and_b64 [[BROKEN_MASK:s\[[0-9]+:[0-9]+\]]], vcc, exec
; GCN: s_or_b64 [[INNER_MASK]], [[INNER_MASK]], [[BROKEN_MASK]]
; GCN: [[FLOW]]: ; %Flow
; GCN: ; in Loop: Header=BB0_1 Depth=1
; GCN: s_and_b64 [[AND_MASK:s\[[0-9]+:[0-9]+\]]], exec, [[INNER_MASK]]
; GCN-NEXT: s_or_b64 [[ACCUM_MASK]], [[AND_MASK]], [[ACCUM_MASK]]
; GCN-NEXT: s_andn2_b64 exec, exec, [[ACCUM_MASK]]
; GCN-NEXT: s_cbranch_execnz [[LOOP_ENTRY]]
; GCN: ; %bb.4: ; %bb9
; GCN-NEXT: s_endpgm
define amdgpu_kernel void @break_loop(i32 %arg) #0 {
bb:
%id = call i32 @llvm.amdgcn.workitem.id.x()
%tmp = sub i32 %id, %arg
br label %bb1
bb1:
%lsr.iv = phi i32 [ undef, %bb ], [ %lsr.iv.next, %bb4 ]
%lsr.iv.next = add i32 %lsr.iv, 1
%cmp0 = icmp slt i32 %lsr.iv.next, 0
br i1 %cmp0, label %bb4, label %bb9
bb4:
%load = load volatile i32, i32 addrspace(1)* undef, align 4
%cmp1 = icmp slt i32 %tmp, %load
br i1 %cmp1, label %bb1, label %bb9
bb9:
ret void
}
; OPT-LABEL: @undef_phi_cond_break_loop(
; OPT: bb1:
; OPT-NEXT: %phi.broken = phi i64 [ %0, %Flow ], [ 0, %bb ]
; OPT-NEXT: %lsr.iv = phi i32 [ undef, %bb ], [ %tmp2, %Flow ]
; OPT-NEXT: %lsr.iv.next = add i32 %lsr.iv, 1
; OPT-NEXT: %cmp0 = icmp slt i32 %lsr.iv.next, 0
; OPT-NEXT: br i1 %cmp0, label %bb4, label %Flow
; OPT: bb4:
; OPT-NEXT: %load = load volatile i32, i32 addrspace(1)* undef, align 4
; OPT-NEXT: %cmp1 = icmp sge i32 %tmp, %load
; OPT-NEXT: br label %Flow
; OPT: Flow:
; OPT-NEXT: %tmp2 = phi i32 [ %lsr.iv.next, %bb4 ], [ undef, %bb1 ]
; OPT-NEXT: %tmp3 = phi i1 [ %cmp1, %bb4 ], [ undef, %bb1 ]
; OPT-NEXT: %0 = call i64 @llvm.amdgcn.if.break.i64.i64(i1 %tmp3, i64 %phi.broken)
; OPT-NEXT: %1 = call i1 @llvm.amdgcn.loop.i64(i64 %0)
; OPT-NEXT: br i1 %1, label %bb9, label %bb1
; OPT: bb9: ; preds = %Flow
; OPT-NEXT: call void @llvm.amdgcn.end.cf.i64(i64 %0)
; OPT-NEXT: store volatile i32 7
; OPT-NEXT: ret void
define amdgpu_kernel void @undef_phi_cond_break_loop(i32 %arg) #0 {
bb:
%id = call i32 @llvm.amdgcn.workitem.id.x()
%tmp = sub i32 %id, %arg
br label %bb1
bb1: ; preds = %Flow, %bb
%lsr.iv = phi i32 [ undef, %bb ], [ %tmp2, %Flow ]
%lsr.iv.next = add i32 %lsr.iv, 1
%cmp0 = icmp slt i32 %lsr.iv.next, 0
br i1 %cmp0, label %bb4, label %Flow
bb4: ; preds = %bb1
%load = load volatile i32, i32 addrspace(1)* undef, align 4
%cmp1 = icmp sge i32 %tmp, %load
br label %Flow
Flow: ; preds = %bb4, %bb1
%tmp2 = phi i32 [ %lsr.iv.next, %bb4 ], [ undef, %bb1 ]
%tmp3 = phi i1 [ %cmp1, %bb4 ], [ undef, %bb1 ]
br i1 %tmp3, label %bb9, label %bb1
bb9: ; preds = %Flow
store volatile i32 7, i32 addrspace(3)* undef
ret void
}
; FIXME: ConstantExpr compare of address to null folds away
@lds = addrspace(3) global i32 undef
; OPT-LABEL: @constexpr_phi_cond_break_loop(
; OPT: bb1:
; OPT-NEXT: %phi.broken = phi i64 [ %0, %Flow ], [ 0, %bb ]
; OPT-NEXT: %lsr.iv = phi i32 [ undef, %bb ], [ %tmp2, %Flow ]
; OPT-NEXT: %lsr.iv.next = add i32 %lsr.iv, 1
; OPT-NEXT: %cmp0 = icmp slt i32 %lsr.iv.next, 0
; OPT-NEXT: br i1 %cmp0, label %bb4, label %Flow
; OPT: bb4:
; OPT-NEXT: %load = load volatile i32, i32 addrspace(1)* undef, align 4
; OPT-NEXT: %cmp1 = icmp sge i32 %tmp, %load
; OPT-NEXT: br label %Flow
; OPT: Flow:
; OPT-NEXT: %tmp2 = phi i32 [ %lsr.iv.next, %bb4 ], [ undef, %bb1 ]
; OPT-NEXT: %tmp3 = phi i1 [ %cmp1, %bb4 ], [ icmp ne (i32 addrspace(3)* inttoptr (i32 4 to i32 addrspace(3)*), i32 addrspace(3)* @lds), %bb1 ]
; OPT-NEXT: %0 = call i64 @llvm.amdgcn.if.break.i64.i64(i1 %tmp3, i64 %phi.broken)
; OPT-NEXT: %1 = call i1 @llvm.amdgcn.loop.i64(i64 %0)
; OPT-NEXT: br i1 %1, label %bb9, label %bb1
; OPT: bb9: ; preds = %Flow
; OPT-NEXT: call void @llvm.amdgcn.end.cf.i64(i64 %0)
; OPT-NEXT: store volatile i32 7
; OPT-NEXT: ret void
define amdgpu_kernel void @constexpr_phi_cond_break_loop(i32 %arg) #0 {
bb:
%id = call i32 @llvm.amdgcn.workitem.id.x()
%tmp = sub i32 %id, %arg
br label %bb1
bb1: ; preds = %Flow, %bb
%lsr.iv = phi i32 [ undef, %bb ], [ %tmp2, %Flow ]
%lsr.iv.next = add i32 %lsr.iv, 1
%cmp0 = icmp slt i32 %lsr.iv.next, 0
br i1 %cmp0, label %bb4, label %Flow
bb4: ; preds = %bb1
%load = load volatile i32, i32 addrspace(1)* undef, align 4
%cmp1 = icmp sge i32 %tmp, %load
br label %Flow
Flow: ; preds = %bb4, %bb1
%tmp2 = phi i32 [ %lsr.iv.next, %bb4 ], [ undef, %bb1 ]
%tmp3 = phi i1 [ %cmp1, %bb4 ], [ icmp ne (i32 addrspace(3)* inttoptr (i32 4 to i32 addrspace(3)*), i32 addrspace(3)* @lds), %bb1 ]
br i1 %tmp3, label %bb9, label %bb1
bb9: ; preds = %Flow
store volatile i32 7, i32 addrspace(3)* undef
ret void
}
; OPT-LABEL: @true_phi_cond_break_loop(
; OPT: bb1:
; OPT-NEXT: %phi.broken = phi i64 [ %0, %Flow ], [ 0, %bb ]
; OPT-NEXT: %lsr.iv = phi i32 [ undef, %bb ], [ %tmp2, %Flow ]
; OPT-NEXT: %lsr.iv.next = add i32 %lsr.iv, 1
; OPT-NEXT: %cmp0 = icmp slt i32 %lsr.iv.next, 0
; OPT-NEXT: br i1 %cmp0, label %bb4, label %Flow
; OPT: bb4:
; OPT-NEXT: %load = load volatile i32, i32 addrspace(1)* undef, align 4
; OPT-NEXT: %cmp1 = icmp sge i32 %tmp, %load
; OPT-NEXT: br label %Flow
; OPT: Flow:
; OPT-NEXT: %tmp2 = phi i32 [ %lsr.iv.next, %bb4 ], [ undef, %bb1 ]
; OPT-NEXT: %tmp3 = phi i1 [ %cmp1, %bb4 ], [ true, %bb1 ]
; OPT-NEXT: %0 = call i64 @llvm.amdgcn.if.break.i64.i64(i1 %tmp3, i64 %phi.broken)
; OPT-NEXT: %1 = call i1 @llvm.amdgcn.loop.i64(i64 %0)
; OPT-NEXT: br i1 %1, label %bb9, label %bb1
; OPT: bb9: ; preds = %Flow
; OPT-NEXT: call void @llvm.amdgcn.end.cf.i64(i64 %0)
; OPT-NEXT: store volatile i32 7
; OPT-NEXT: ret void
define amdgpu_kernel void @true_phi_cond_break_loop(i32 %arg) #0 {
bb:
%id = call i32 @llvm.amdgcn.workitem.id.x()
%tmp = sub i32 %id, %arg
br label %bb1
bb1: ; preds = %Flow, %bb
%lsr.iv = phi i32 [ undef, %bb ], [ %tmp2, %Flow ]
%lsr.iv.next = add i32 %lsr.iv, 1
%cmp0 = icmp slt i32 %lsr.iv.next, 0
br i1 %cmp0, label %bb4, label %Flow
bb4: ; preds = %bb1
%load = load volatile i32, i32 addrspace(1)* undef, align 4
%cmp1 = icmp sge i32 %tmp, %load
br label %Flow
Flow: ; preds = %bb4, %bb1
%tmp2 = phi i32 [ %lsr.iv.next, %bb4 ], [ undef, %bb1 ]
%tmp3 = phi i1 [ %cmp1, %bb4 ], [ true, %bb1 ]
br i1 %tmp3, label %bb9, label %bb1
bb9: ; preds = %Flow
store volatile i32 7, i32 addrspace(3)* undef
ret void
}
; OPT-LABEL: @false_phi_cond_break_loop(
; OPT: bb1:
; OPT-NEXT: %phi.broken = phi i64 [ %0, %Flow ], [ 0, %bb ]
; OPT-NEXT: %lsr.iv = phi i32 [ undef, %bb ], [ %tmp2, %Flow ]
; OPT-NOT: call
; OPT: br i1 %cmp0, label %bb4, label %Flow
; OPT: bb4:
; OPT-NEXT: %load = load volatile i32, i32 addrspace(1)* undef, align 4
; OPT-NEXT: %cmp1 = icmp sge i32 %tmp, %load
; OPT-NEXT: br label %Flow
; OPT: Flow:
; OPT-NEXT: %tmp2 = phi i32 [ %lsr.iv.next, %bb4 ], [ undef, %bb1 ]
; OPT-NEXT: %tmp3 = phi i1 [ %cmp1, %bb4 ], [ false, %bb1 ]
; OPT-NEXT: %0 = call i64 @llvm.amdgcn.if.break.i64.i64(i1 %tmp3, i64 %phi.broken)
; OPT-NEXT: %1 = call i1 @llvm.amdgcn.loop.i64(i64 %0)
; OPT-NEXT: br i1 %1, label %bb9, label %bb1
; OPT: bb9: ; preds = %Flow
; OPT-NEXT: call void @llvm.amdgcn.end.cf.i64(i64 %0)
; OPT-NEXT: store volatile i32 7
; OPT-NEXT: ret void
define amdgpu_kernel void @false_phi_cond_break_loop(i32 %arg) #0 {
bb:
%id = call i32 @llvm.amdgcn.workitem.id.x()
%tmp = sub i32 %id, %arg
br label %bb1
bb1: ; preds = %Flow, %bb
%lsr.iv = phi i32 [ undef, %bb ], [ %tmp2, %Flow ]
%lsr.iv.next = add i32 %lsr.iv, 1
%cmp0 = icmp slt i32 %lsr.iv.next, 0
br i1 %cmp0, label %bb4, label %Flow
bb4: ; preds = %bb1
%load = load volatile i32, i32 addrspace(1)* undef, align 4
%cmp1 = icmp sge i32 %tmp, %load
br label %Flow
Flow: ; preds = %bb4, %bb1
%tmp2 = phi i32 [ %lsr.iv.next, %bb4 ], [ undef, %bb1 ]
%tmp3 = phi i1 [ %cmp1, %bb4 ], [ false, %bb1 ]
br i1 %tmp3, label %bb9, label %bb1
bb9: ; preds = %Flow
store volatile i32 7, i32 addrspace(3)* undef
ret void
}
; Swap order of branches in flow block so that the true phi is
; continue.
; OPT-LABEL: @invert_true_phi_cond_break_loop(
; OPT: bb1:
; OPT-NEXT: %phi.broken = phi i64 [ %1, %Flow ], [ 0, %bb ]
; OPT-NEXT: %lsr.iv = phi i32 [ undef, %bb ], [ %tmp2, %Flow ]
; OPT-NEXT: %lsr.iv.next = add i32 %lsr.iv, 1
; OPT-NEXT: %cmp0 = icmp slt i32 %lsr.iv.next, 0
; OPT-NEXT: br i1 %cmp0, label %bb4, label %Flow
; OPT: bb4:
; OPT-NEXT: %load = load volatile i32, i32 addrspace(1)* undef, align 4
; OPT-NEXT: %cmp1 = icmp sge i32 %tmp, %load
; OPT-NEXT: br label %Flow
; OPT: Flow:
; OPT-NEXT: %tmp2 = phi i32 [ %lsr.iv.next, %bb4 ], [ undef, %bb1 ]
; OPT-NEXT: %tmp3 = phi i1 [ %cmp1, %bb4 ], [ true, %bb1 ]
; OPT-NEXT: %0 = xor i1 %tmp3, true
; OPT-NEXT: %1 = call i64 @llvm.amdgcn.if.break.i64.i64(i1 %0, i64 %phi.broken)
; OPT-NEXT: %2 = call i1 @llvm.amdgcn.loop.i64(i64 %1)
; OPT-NEXT: br i1 %2, label %bb9, label %bb1
; OPT: bb9:
; OPT-NEXT: call void @llvm.amdgcn.end.cf.i64(i64 %1)
; OPT-NEXT: store volatile i32 7, i32 addrspace(3)* undef
; OPT-NEXT: ret void
define amdgpu_kernel void @invert_true_phi_cond_break_loop(i32 %arg) #0 {
bb:
%id = call i32 @llvm.amdgcn.workitem.id.x()
%tmp = sub i32 %id, %arg
br label %bb1
bb1: ; preds = %Flow, %bb
%lsr.iv = phi i32 [ undef, %bb ], [ %tmp2, %Flow ]
%lsr.iv.next = add i32 %lsr.iv, 1
%cmp0 = icmp slt i32 %lsr.iv.next, 0
br i1 %cmp0, label %bb4, label %Flow
bb4: ; preds = %bb1
%load = load volatile i32, i32 addrspace(1)* undef, align 4
%cmp1 = icmp sge i32 %tmp, %load
br label %Flow
Flow: ; preds = %bb4, %bb1
%tmp2 = phi i32 [ %lsr.iv.next, %bb4 ], [ undef, %bb1 ]
%tmp3 = phi i1 [ %cmp1, %bb4 ], [ true, %bb1 ]
br i1 %tmp3, label %bb1, label %bb9
bb9: ; preds = %Flow
store volatile i32 7, i32 addrspace(3)* undef
ret void
}
declare i32 @llvm.amdgcn.workitem.id.x() #1
attributes #0 = { nounwind }
attributes #1 = { nounwind readnone }