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
|
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -S -aa-pipeline=basic-aa -passes='require<opt-remark-emit>,make-guards-explicit,loop-mssa(licm)' < %s | FileCheck %s
declare void @llvm.experimental.guard(i1,...)
declare void @maythrow()
declare void @llvm.experimental.deoptimize.isVoid(...)
declare i1 @llvm.experimental.widenable.condition()
; Make sure that we do not hoist widenable_cond out of loop.
; Widenable conditions don't actually alias anything or throw, however
; hoisting them is not profitable because it can prevent guard widening
; in loop.
; If we want to widen guards in branch form, we need to generate them
; at the point of widenable_condition call
; (see https://github.com/llvm/llvm-project/issues/60234). Because of
; this, hoisting a widenable condition out of loop can make widening of
; in-loop conditions impossible (e.g. if they depend on loop variant). On
; the other hand, hoisting of the WC doesn't really see profitable,
; because it will be optimized away later anyways. So the test checks that
; LICM will not move it out of loop.
define void @do_not_hoist_widenable_cond(i1 %cond, i32 %N, i32 %M) {
; CHECK-LABEL: @do_not_hoist_widenable_cond(
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[GUARDED:%.*]] ]
; CHECK-NEXT: [[GUARD_COND:%.*]] = icmp slt i32 [[IV]], [[N:%.*]]
; CHECK-NEXT: [[WIDENABLE_COND:%.*]] = call i1 @llvm.experimental.widenable.condition()
; CHECK-NEXT: [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[GUARD_COND]], [[WIDENABLE_COND]]
; CHECK-NEXT: br i1 [[EXIPLICIT_GUARD_COND]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0:![0-9]+]]
; CHECK: deopt:
; CHECK-NEXT: call void (...) @llvm.experimental.deoptimize.isVoid() [ "deopt"() ]
; CHECK-NEXT: ret void
; CHECK: guarded:
; CHECK-NEXT: [[LOOP_COND:%.*]] = icmp slt i32 [[IV]], [[M:%.*]]
; CHECK-NEXT: [[IV_NEXT]] = add i32 [[IV]], 1
; CHECK-NEXT: br i1 [[LOOP_COND]], label [[LOOP]], label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
br label %loop
loop:
%iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]
%guard_cond = icmp slt i32 %iv, %N
call void(i1, ...) @llvm.experimental.guard(i1 %guard_cond) [ "deopt"() ]
%loop_cond = icmp slt i32 %iv, %M
%iv.next = add i32 %iv, 1
br i1 %loop_cond, label %loop, label %exit
exit:
ret void
}
define void @do_not_hoist_widenable_cond_speculate(i1 %cond, i32 %N, i32 %M) {
; CHECK-LABEL: @do_not_hoist_widenable_cond_speculate(
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[GUARDED:%.*]] ]
; CHECK-NEXT: [[GUARD_COND:%.*]] = icmp slt i32 [[IV]], [[N:%.*]]
; CHECK-NEXT: call void @maythrow()
; CHECK-NEXT: [[WIDENABLE_COND:%.*]] = call i1 @llvm.experimental.widenable.condition()
; CHECK-NEXT: [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[GUARD_COND]], [[WIDENABLE_COND]]
; CHECK-NEXT: br i1 [[EXIPLICIT_GUARD_COND]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
; CHECK: deopt:
; CHECK-NEXT: call void (...) @llvm.experimental.deoptimize.isVoid() [ "deopt"() ]
; CHECK-NEXT: ret void
; CHECK: guarded:
; CHECK-NEXT: [[LOOP_COND:%.*]] = icmp slt i32 [[IV]], [[M:%.*]]
; CHECK-NEXT: [[IV_NEXT]] = add i32 [[IV]], 1
; CHECK-NEXT: br i1 [[LOOP_COND]], label [[LOOP]], label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
br label %loop
loop:
%iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]
%guard_cond = icmp slt i32 %iv, %N
call void @maythrow()
call void(i1, ...) @llvm.experimental.guard(i1 %guard_cond) [ "deopt"() ]
%loop_cond = icmp slt i32 %iv, %M
%iv.next = add i32 %iv, 1
br i1 %loop_cond, label %loop, label %exit
exit:
ret void
}
define void @hoist_invariant_load(i1 %cond, ptr %np, i32 %M) {
; CHECK-LABEL: @hoist_invariant_load(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[N:%.*]] = load i32, ptr [[NP:%.*]], align 4
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[GUARDED:%.*]] ]
; CHECK-NEXT: [[GUARD_COND:%.*]] = icmp slt i32 [[IV]], [[N]]
; CHECK-NEXT: [[WIDENABLE_COND:%.*]] = call i1 @llvm.experimental.widenable.condition()
; CHECK-NEXT: [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[GUARD_COND]], [[WIDENABLE_COND]]
; CHECK-NEXT: br i1 [[EXIPLICIT_GUARD_COND]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
; CHECK: deopt:
; CHECK-NEXT: call void (...) @llvm.experimental.deoptimize.isVoid() [ "deopt"() ]
; CHECK-NEXT: ret void
; CHECK: guarded:
; CHECK-NEXT: [[LOOP_COND:%.*]] = icmp slt i32 [[IV]], [[M:%.*]]
; CHECK-NEXT: [[IV_NEXT]] = add i32 [[IV]], 1
; CHECK-NEXT: br i1 [[LOOP_COND]], label [[LOOP]], label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
br label %loop
loop:
%iv = phi i32 [ 0, %entry ], [ %iv.next, %loop ]
%N = load i32, ptr %np
%guard_cond = icmp slt i32 %iv, %N
call void(i1, ...) @llvm.experimental.guard(i1 %guard_cond) [ "deopt"() ]
%loop_cond = icmp slt i32 %iv, %M
%iv.next = add i32 %iv, 1
br i1 %loop_cond, label %loop, label %exit
exit:
ret void
}
; TODO: Hoist widenable condition out of loop since it's the only
; non-invariant operand of its user. If we hoist wc here, we can hoist
; `and` as well, otherwise it prevents us from doing that.
define void @hoist_widenable_cond_1(ptr %p, i32 %x) {
; CHECK-LABEL: @hoist_widenable_cond_1(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[LOAD:%.*]] = load atomic i32, ptr [[P:%.*]] unordered, align 8
; CHECK-NEXT: [[ICMP7:%.*]] = icmp ult i32 0, [[LOAD]]
; CHECK-NEXT: br label [[LOOP_HEADER:%.*]]
; CHECK: loop.header:
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[ENTRY:%.*]] ]
; CHECK-NEXT: [[ICMP:%.*]] = icmp ult i32 [[IV]], [[X:%.*]]
; CHECK-NEXT: br i1 [[ICMP]], label [[LOOP_BODY:%.*]], label [[EXIT:%.*]]
; CHECK: loop.body:
; CHECK-NEXT: [[WIDENABLE_COND:%.*]] = call i1 @llvm.experimental.widenable.condition()
; CHECK-NEXT: [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[ICMP7]], [[WIDENABLE_COND]]
; CHECK-NEXT: br i1 [[EXIPLICIT_GUARD_COND]], label [[GUARDED]], label [[DEOPT:%.*]]
; CHECK: deopt:
; CHECK-NEXT: call void (...) @llvm.experimental.deoptimize.isVoid(i32 13) [ "deopt"() ]
; CHECK-NEXT: ret void
; CHECK: guarded:
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i32 [[IV]], 1
; CHECK-NEXT: br label [[LOOP_HEADER]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
%load = load atomic i32, ptr %p unordered, align 8
br label %loop.header
loop.header:
%iv = phi i32 [ %iv.next, %guarded ], [ 0, %entry ]
%icmp = icmp ult i32 %iv, %x
br i1 %icmp, label %loop.body, label %exit
loop.body:
%icmp7 = icmp ult i32 0, %load
%widenable_cond = call i1 @llvm.experimental.widenable.condition()
%exiplicit_guard_cond = and i1 %icmp7, %widenable_cond
br i1 %exiplicit_guard_cond, label %guarded, label %deopt
deopt:
call void (...) @llvm.experimental.deoptimize.isVoid(i32 13) [ "deopt"() ]
ret void
guarded:
%iv.next = add nuw nsw i32 %iv, 1
br label %loop.header
exit:
ret void
}
; Same as hoist_widenable_cond_1 test, but widenable condition goes before
; icmp - the second operand of wc's user.
; TODO: Hoist widenable condition out of loop.
define void @hoist_widenable_cond_2(ptr %p, i32 %x) {
; CHECK-LABEL: @hoist_widenable_cond_2(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[LOAD:%.*]] = load atomic i32, ptr [[P:%.*]] unordered, align 8
; CHECK-NEXT: [[ICMP7:%.*]] = icmp ult i32 0, [[LOAD]]
; CHECK-NEXT: br label [[LOOP_HEADER:%.*]]
; CHECK: loop.header:
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[ENTRY:%.*]] ]
; CHECK-NEXT: [[ICMP:%.*]] = icmp ult i32 [[IV]], [[X:%.*]]
; CHECK-NEXT: br i1 [[ICMP]], label [[LOOP_BODY:%.*]], label [[EXIT:%.*]]
; CHECK: loop.body:
; CHECK-NEXT: [[WIDENABLE_COND:%.*]] = call i1 @llvm.experimental.widenable.condition()
; CHECK-NEXT: [[EXIPLICIT_GUARD_COND:%.*]] = and i1 [[ICMP7]], [[WIDENABLE_COND]]
; CHECK-NEXT: br i1 [[EXIPLICIT_GUARD_COND]], label [[GUARDED]], label [[DEOPT:%.*]]
; CHECK: deopt:
; CHECK-NEXT: call void (...) @llvm.experimental.deoptimize.isVoid(i32 13) [ "deopt"() ]
; CHECK-NEXT: ret void
; CHECK: guarded:
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i32 [[IV]], 1
; CHECK-NEXT: br label [[LOOP_HEADER]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
%load = load atomic i32, ptr %p unordered, align 8
br label %loop.header
loop.header:
%iv = phi i32 [ %iv.next, %guarded ], [ 0, %entry ]
%icmp = icmp ult i32 %iv, %x
br i1 %icmp, label %loop.body, label %exit
loop.body:
%widenable_cond = call i1 @llvm.experimental.widenable.condition()
%icmp7 = icmp ult i32 0, %load
%exiplicit_guard_cond = and i1 %icmp7, %widenable_cond
br i1 %exiplicit_guard_cond, label %guarded, label %deopt
deopt:
call void (...) @llvm.experimental.deoptimize.isVoid(i32 13) [ "deopt"() ]
ret void
guarded:
%iv.next = add nuw nsw i32 %iv, 1
br label %loop.header
exit:
ret void
}
|
