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 -mtriple=amdgcn-- -passes='amdgpu-atomic-optimizer<strategy=iterative>,verify<domtree>' %s | FileCheck -check-prefix=IR-ITERATIVE %s
; RUN: opt -S -mtriple=amdgcn-- -passes='amdgpu-atomic-optimizer<strategy=dpp>,verify<domtree>' %s | FileCheck -check-prefix=IR-DPP %s
declare i32 @llvm.amdgcn.workitem.id.x()
define amdgpu_kernel void @global_atomic_fadd_uni_value(ptr addrspace(1) %ptr) #0 {
; IR-ITERATIVE-LABEL: @global_atomic_fadd_uni_value(
; IR-ITERATIVE-NEXT: [[TMP1:%.*]] = call i64 @llvm.amdgcn.ballot.i64(i1 true)
; IR-ITERATIVE-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
; IR-ITERATIVE-NEXT: [[TMP3:%.*]] = lshr i64 [[TMP1]], 32
; IR-ITERATIVE-NEXT: [[TMP4:%.*]] = trunc i64 [[TMP3]] to i32
; IR-ITERATIVE-NEXT: [[TMP5:%.*]] = call i32 @llvm.amdgcn.mbcnt.lo(i32 [[TMP2]], i32 0)
; IR-ITERATIVE-NEXT: [[TMP6:%.*]] = call i32 @llvm.amdgcn.mbcnt.hi(i32 [[TMP4]], i32 [[TMP5]])
; IR-ITERATIVE-NEXT: [[TMP7:%.*]] = call i64 @llvm.ctpop.i64(i64 [[TMP1]])
; IR-ITERATIVE-NEXT: [[TMP8:%.*]] = trunc i64 [[TMP7]] to i32
; IR-ITERATIVE-NEXT: [[TMP9:%.*]] = uitofp i32 [[TMP8]] to float
; IR-ITERATIVE-NEXT: [[TMP10:%.*]] = fmul float 4.000000e+00, [[TMP9]]
; IR-ITERATIVE-NEXT: [[TMP11:%.*]] = icmp eq i32 [[TMP6]], 0
; IR-ITERATIVE-NEXT: br i1 [[TMP11]], label [[TMP12:%.*]], label [[TMP14:%.*]]
; IR-ITERATIVE: 12:
; IR-ITERATIVE-NEXT: [[TMP13:%.*]] = atomicrmw fadd ptr addrspace(1) [[PTR:%.*]], float [[TMP10]] seq_cst, align 4
; IR-ITERATIVE-NEXT: br label [[TMP14]]
; IR-ITERATIVE: 14:
; IR-ITERATIVE-NEXT: ret void
;
; IR-DPP-LABEL: @global_atomic_fadd_uni_value(
; IR-DPP-NEXT: [[TMP1:%.*]] = call i64 @llvm.amdgcn.ballot.i64(i1 true)
; IR-DPP-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
; IR-DPP-NEXT: [[TMP3:%.*]] = lshr i64 [[TMP1]], 32
; IR-DPP-NEXT: [[TMP4:%.*]] = trunc i64 [[TMP3]] to i32
; IR-DPP-NEXT: [[TMP5:%.*]] = call i32 @llvm.amdgcn.mbcnt.lo(i32 [[TMP2]], i32 0)
; IR-DPP-NEXT: [[TMP6:%.*]] = call i32 @llvm.amdgcn.mbcnt.hi(i32 [[TMP4]], i32 [[TMP5]])
; IR-DPP-NEXT: [[TMP7:%.*]] = call i64 @llvm.ctpop.i64(i64 [[TMP1]])
; IR-DPP-NEXT: [[TMP8:%.*]] = trunc i64 [[TMP7]] to i32
; IR-DPP-NEXT: [[TMP9:%.*]] = uitofp i32 [[TMP8]] to float
; IR-DPP-NEXT: [[TMP10:%.*]] = fmul float 4.000000e+00, [[TMP9]]
; IR-DPP-NEXT: [[TMP11:%.*]] = icmp eq i32 [[TMP6]], 0
; IR-DPP-NEXT: br i1 [[TMP11]], label [[TMP12:%.*]], label [[TMP14:%.*]]
; IR-DPP: 12:
; IR-DPP-NEXT: [[TMP13:%.*]] = atomicrmw fadd ptr addrspace(1) [[PTR:%.*]], float [[TMP10]] seq_cst, align 4
; IR-DPP-NEXT: br label [[TMP14]]
; IR-DPP: 14:
; IR-DPP-NEXT: ret void
;
%result = atomicrmw fadd ptr addrspace(1) %ptr, float 4.0 seq_cst
ret void
}
define amdgpu_kernel void @global_atomic_fadd_div_value(ptr addrspace(1) %ptr) #0 {
; IR-ITERATIVE-LABEL: @global_atomic_fadd_div_value(
; IR-ITERATIVE-NEXT: [[ID_X:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()
; IR-ITERATIVE-NEXT: [[DIVVALUE:%.*]] = bitcast i32 [[ID_X]] to float
; IR-ITERATIVE-NEXT: [[TMP1:%.*]] = call i64 @llvm.amdgcn.ballot.i64(i1 true)
; IR-ITERATIVE-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
; IR-ITERATIVE-NEXT: [[TMP3:%.*]] = lshr i64 [[TMP1]], 32
; IR-ITERATIVE-NEXT: [[TMP4:%.*]] = trunc i64 [[TMP3]] to i32
; IR-ITERATIVE-NEXT: [[TMP5:%.*]] = call i32 @llvm.amdgcn.mbcnt.lo(i32 [[TMP2]], i32 0)
; IR-ITERATIVE-NEXT: [[TMP6:%.*]] = call i32 @llvm.amdgcn.mbcnt.hi(i32 [[TMP4]], i32 [[TMP5]])
; IR-ITERATIVE-NEXT: [[TMP7:%.*]] = call i64 @llvm.amdgcn.ballot.i64(i1 true)
; IR-ITERATIVE-NEXT: br label [[COMPUTELOOP:%.*]]
; IR-ITERATIVE: 8:
; IR-ITERATIVE-NEXT: [[TMP9:%.*]] = atomicrmw fadd ptr addrspace(1) [[PTR:%.*]], float [[TMP14:%.*]] seq_cst, align 4
; IR-ITERATIVE-NEXT: br label [[TMP10:%.*]]
; IR-ITERATIVE: 10:
; IR-ITERATIVE-NEXT: ret void
; IR-ITERATIVE: ComputeLoop:
; IR-ITERATIVE-NEXT: [[ACCUMULATOR:%.*]] = phi float [ -0.000000e+00, [[TMP0:%.*]] ], [ [[TMP14]], [[COMPUTELOOP]] ]
; IR-ITERATIVE-NEXT: [[ACTIVEBITS:%.*]] = phi i64 [ [[TMP7]], [[TMP0]] ], [ [[TMP17:%.*]], [[COMPUTELOOP]] ]
; IR-ITERATIVE-NEXT: [[TMP11:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true)
; IR-ITERATIVE-NEXT: [[TMP12:%.*]] = trunc i64 [[TMP11]] to i32
; IR-ITERATIVE-NEXT: [[TMP13:%.*]] = call float @llvm.amdgcn.readlane.f32(float [[DIVVALUE]], i32 [[TMP12]])
; IR-ITERATIVE-NEXT: [[TMP14]] = fadd float [[ACCUMULATOR]], [[TMP13]]
; IR-ITERATIVE-NEXT: [[TMP15:%.*]] = shl i64 1, [[TMP11]]
; IR-ITERATIVE-NEXT: [[TMP16:%.*]] = xor i64 [[TMP15]], -1
; IR-ITERATIVE-NEXT: [[TMP17]] = and i64 [[ACTIVEBITS]], [[TMP16]]
; IR-ITERATIVE-NEXT: [[TMP18:%.*]] = icmp eq i64 [[TMP17]], 0
; IR-ITERATIVE-NEXT: br i1 [[TMP18]], label [[COMPUTEEND:%.*]], label [[COMPUTELOOP]]
; IR-ITERATIVE: ComputeEnd:
; IR-ITERATIVE-NEXT: [[TMP19:%.*]] = icmp eq i32 [[TMP6]], 0
; IR-ITERATIVE-NEXT: br i1 [[TMP19]], label [[TMP8:%.*]], label [[TMP10]]
;
; IR-DPP-LABEL: @global_atomic_fadd_div_value(
; IR-DPP-NEXT: [[ID_X:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()
; IR-DPP-NEXT: [[DIVVALUE:%.*]] = bitcast i32 [[ID_X]] to float
; IR-DPP-NEXT: [[TMP1:%.*]] = call i64 @llvm.amdgcn.ballot.i64(i1 true)
; IR-DPP-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
; IR-DPP-NEXT: [[TMP3:%.*]] = lshr i64 [[TMP1]], 32
; IR-DPP-NEXT: [[TMP4:%.*]] = trunc i64 [[TMP3]] to i32
; IR-DPP-NEXT: [[TMP5:%.*]] = call i32 @llvm.amdgcn.mbcnt.lo(i32 [[TMP2]], i32 0)
; IR-DPP-NEXT: [[TMP6:%.*]] = call i32 @llvm.amdgcn.mbcnt.hi(i32 [[TMP4]], i32 [[TMP5]])
; IR-DPP-NEXT: [[TMP7:%.*]] = call float @llvm.amdgcn.set.inactive.f32(float [[DIVVALUE]], float -0.000000e+00)
; IR-DPP-NEXT: [[TMP8:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP7]], i32 273, i32 15, i32 15, i1 false)
; IR-DPP-NEXT: [[TMP9:%.*]] = fadd float [[TMP7]], [[TMP8]]
; IR-DPP-NEXT: [[TMP10:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP9]], i32 274, i32 15, i32 15, i1 false)
; IR-DPP-NEXT: [[TMP11:%.*]] = fadd float [[TMP9]], [[TMP10]]
; IR-DPP-NEXT: [[TMP12:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP11]], i32 276, i32 15, i32 15, i1 false)
; IR-DPP-NEXT: [[TMP13:%.*]] = fadd float [[TMP11]], [[TMP12]]
; IR-DPP-NEXT: [[TMP14:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP13]], i32 280, i32 15, i32 15, i1 false)
; IR-DPP-NEXT: [[TMP15:%.*]] = fadd float [[TMP13]], [[TMP14]]
; IR-DPP-NEXT: [[TMP16:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP15]], i32 322, i32 10, i32 15, i1 false)
; IR-DPP-NEXT: [[TMP17:%.*]] = fadd float [[TMP15]], [[TMP16]]
; IR-DPP-NEXT: [[TMP18:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP17]], i32 323, i32 12, i32 15, i1 false)
; IR-DPP-NEXT: [[TMP19:%.*]] = fadd float [[TMP17]], [[TMP18]]
; IR-DPP-NEXT: [[TMP20:%.*]] = call float @llvm.amdgcn.readlane.f32(float [[TMP19]], i32 63)
; IR-DPP-NEXT: [[TMP21:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP20]])
; IR-DPP-NEXT: [[TMP22:%.*]] = icmp eq i32 [[TMP6]], 0
; IR-DPP-NEXT: br i1 [[TMP22]], label [[TMP23:%.*]], label [[TMP25:%.*]]
; IR-DPP: 23:
; IR-DPP-NEXT: [[TMP24:%.*]] = atomicrmw fadd ptr addrspace(1) [[PTR:%.*]], float [[TMP21]] seq_cst, align 4
; IR-DPP-NEXT: br label [[TMP25]]
; IR-DPP: 25:
; IR-DPP-NEXT: ret void
;
%id.x = call i32 @llvm.amdgcn.workitem.id.x()
%divValue = bitcast i32 %id.x to float
%result = atomicrmw fadd ptr addrspace(1) %ptr, float %divValue seq_cst
ret void
}
define amdgpu_kernel void @global_atomic_fsub_uni_value(ptr addrspace(1) %ptr) #0 {
; IR-ITERATIVE-LABEL: @global_atomic_fsub_uni_value(
; IR-ITERATIVE-NEXT: [[TMP1:%.*]] = call i64 @llvm.amdgcn.ballot.i64(i1 true)
; IR-ITERATIVE-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
; IR-ITERATIVE-NEXT: [[TMP3:%.*]] = lshr i64 [[TMP1]], 32
; IR-ITERATIVE-NEXT: [[TMP4:%.*]] = trunc i64 [[TMP3]] to i32
; IR-ITERATIVE-NEXT: [[TMP5:%.*]] = call i32 @llvm.amdgcn.mbcnt.lo(i32 [[TMP2]], i32 0)
; IR-ITERATIVE-NEXT: [[TMP6:%.*]] = call i32 @llvm.amdgcn.mbcnt.hi(i32 [[TMP4]], i32 [[TMP5]])
; IR-ITERATIVE-NEXT: [[TMP7:%.*]] = call i64 @llvm.ctpop.i64(i64 [[TMP1]])
; IR-ITERATIVE-NEXT: [[TMP8:%.*]] = trunc i64 [[TMP7]] to i32
; IR-ITERATIVE-NEXT: [[TMP9:%.*]] = uitofp i32 [[TMP8]] to float
; IR-ITERATIVE-NEXT: [[TMP10:%.*]] = fmul float 4.000000e+00, [[TMP9]]
; IR-ITERATIVE-NEXT: [[TMP11:%.*]] = icmp eq i32 [[TMP6]], 0
; IR-ITERATIVE-NEXT: br i1 [[TMP11]], label [[TMP12:%.*]], label [[TMP14:%.*]]
; IR-ITERATIVE: 12:
; IR-ITERATIVE-NEXT: [[TMP13:%.*]] = atomicrmw fadd ptr addrspace(1) [[PTR:%.*]], float [[TMP10]] seq_cst, align 4
; IR-ITERATIVE-NEXT: br label [[TMP14]]
; IR-ITERATIVE: 14:
; IR-ITERATIVE-NEXT: ret void
;
; IR-DPP-LABEL: @global_atomic_fsub_uni_value(
; IR-DPP-NEXT: [[TMP1:%.*]] = call i64 @llvm.amdgcn.ballot.i64(i1 true)
; IR-DPP-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
; IR-DPP-NEXT: [[TMP3:%.*]] = lshr i64 [[TMP1]], 32
; IR-DPP-NEXT: [[TMP4:%.*]] = trunc i64 [[TMP3]] to i32
; IR-DPP-NEXT: [[TMP5:%.*]] = call i32 @llvm.amdgcn.mbcnt.lo(i32 [[TMP2]], i32 0)
; IR-DPP-NEXT: [[TMP6:%.*]] = call i32 @llvm.amdgcn.mbcnt.hi(i32 [[TMP4]], i32 [[TMP5]])
; IR-DPP-NEXT: [[TMP7:%.*]] = call i64 @llvm.ctpop.i64(i64 [[TMP1]])
; IR-DPP-NEXT: [[TMP8:%.*]] = trunc i64 [[TMP7]] to i32
; IR-DPP-NEXT: [[TMP9:%.*]] = uitofp i32 [[TMP8]] to float
; IR-DPP-NEXT: [[TMP10:%.*]] = fmul float 4.000000e+00, [[TMP9]]
; IR-DPP-NEXT: [[TMP11:%.*]] = icmp eq i32 [[TMP6]], 0
; IR-DPP-NEXT: br i1 [[TMP11]], label [[TMP12:%.*]], label [[TMP14:%.*]]
; IR-DPP: 12:
; IR-DPP-NEXT: [[TMP13:%.*]] = atomicrmw fadd ptr addrspace(1) [[PTR:%.*]], float [[TMP10]] seq_cst, align 4
; IR-DPP-NEXT: br label [[TMP14]]
; IR-DPP: 14:
; IR-DPP-NEXT: ret void
;
%result = atomicrmw fadd ptr addrspace(1) %ptr, float 4.0 seq_cst
ret void
}
define amdgpu_kernel void @global_atomic_fsub_div_value(ptr addrspace(1) %ptr) #0 {
; IR-ITERATIVE-LABEL: @global_atomic_fsub_div_value(
; IR-ITERATIVE-NEXT: [[ID_X:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()
; IR-ITERATIVE-NEXT: [[DIVVALUE:%.*]] = bitcast i32 [[ID_X]] to float
; IR-ITERATIVE-NEXT: [[TMP1:%.*]] = call i64 @llvm.amdgcn.ballot.i64(i1 true)
; IR-ITERATIVE-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
; IR-ITERATIVE-NEXT: [[TMP3:%.*]] = lshr i64 [[TMP1]], 32
; IR-ITERATIVE-NEXT: [[TMP4:%.*]] = trunc i64 [[TMP3]] to i32
; IR-ITERATIVE-NEXT: [[TMP5:%.*]] = call i32 @llvm.amdgcn.mbcnt.lo(i32 [[TMP2]], i32 0)
; IR-ITERATIVE-NEXT: [[TMP6:%.*]] = call i32 @llvm.amdgcn.mbcnt.hi(i32 [[TMP4]], i32 [[TMP5]])
; IR-ITERATIVE-NEXT: [[TMP7:%.*]] = call i64 @llvm.amdgcn.ballot.i64(i1 true)
; IR-ITERATIVE-NEXT: br label [[COMPUTELOOP:%.*]]
; IR-ITERATIVE: 8:
; IR-ITERATIVE-NEXT: [[TMP9:%.*]] = atomicrmw fsub ptr addrspace(1) [[PTR:%.*]], float [[TMP14:%.*]] seq_cst, align 4
; IR-ITERATIVE-NEXT: br label [[TMP10:%.*]]
; IR-ITERATIVE: 10:
; IR-ITERATIVE-NEXT: ret void
; IR-ITERATIVE: ComputeLoop:
; IR-ITERATIVE-NEXT: [[ACCUMULATOR:%.*]] = phi float [ -0.000000e+00, [[TMP0:%.*]] ], [ [[TMP14]], [[COMPUTELOOP]] ]
; IR-ITERATIVE-NEXT: [[ACTIVEBITS:%.*]] = phi i64 [ [[TMP7]], [[TMP0]] ], [ [[TMP17:%.*]], [[COMPUTELOOP]] ]
; IR-ITERATIVE-NEXT: [[TMP11:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true)
; IR-ITERATIVE-NEXT: [[TMP12:%.*]] = trunc i64 [[TMP11]] to i32
; IR-ITERATIVE-NEXT: [[TMP13:%.*]] = call float @llvm.amdgcn.readlane.f32(float [[DIVVALUE]], i32 [[TMP12]])
; IR-ITERATIVE-NEXT: [[TMP14]] = fadd float [[ACCUMULATOR]], [[TMP13]]
; IR-ITERATIVE-NEXT: [[TMP15:%.*]] = shl i64 1, [[TMP11]]
; IR-ITERATIVE-NEXT: [[TMP16:%.*]] = xor i64 [[TMP15]], -1
; IR-ITERATIVE-NEXT: [[TMP17]] = and i64 [[ACTIVEBITS]], [[TMP16]]
; IR-ITERATIVE-NEXT: [[TMP18:%.*]] = icmp eq i64 [[TMP17]], 0
; IR-ITERATIVE-NEXT: br i1 [[TMP18]], label [[COMPUTEEND:%.*]], label [[COMPUTELOOP]]
; IR-ITERATIVE: ComputeEnd:
; IR-ITERATIVE-NEXT: [[TMP19:%.*]] = icmp eq i32 [[TMP6]], 0
; IR-ITERATIVE-NEXT: br i1 [[TMP19]], label [[TMP8:%.*]], label [[TMP10]]
;
; IR-DPP-LABEL: @global_atomic_fsub_div_value(
; IR-DPP-NEXT: [[ID_X:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()
; IR-DPP-NEXT: [[DIVVALUE:%.*]] = bitcast i32 [[ID_X]] to float
; IR-DPP-NEXT: [[TMP1:%.*]] = call i64 @llvm.amdgcn.ballot.i64(i1 true)
; IR-DPP-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
; IR-DPP-NEXT: [[TMP3:%.*]] = lshr i64 [[TMP1]], 32
; IR-DPP-NEXT: [[TMP4:%.*]] = trunc i64 [[TMP3]] to i32
; IR-DPP-NEXT: [[TMP5:%.*]] = call i32 @llvm.amdgcn.mbcnt.lo(i32 [[TMP2]], i32 0)
; IR-DPP-NEXT: [[TMP6:%.*]] = call i32 @llvm.amdgcn.mbcnt.hi(i32 [[TMP4]], i32 [[TMP5]])
; IR-DPP-NEXT: [[TMP7:%.*]] = call float @llvm.amdgcn.set.inactive.f32(float [[DIVVALUE]], float -0.000000e+00)
; IR-DPP-NEXT: [[TMP8:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP7]], i32 273, i32 15, i32 15, i1 false)
; IR-DPP-NEXT: [[TMP9:%.*]] = fadd float [[TMP7]], [[TMP8]]
; IR-DPP-NEXT: [[TMP10:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP9]], i32 274, i32 15, i32 15, i1 false)
; IR-DPP-NEXT: [[TMP11:%.*]] = fadd float [[TMP9]], [[TMP10]]
; IR-DPP-NEXT: [[TMP12:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP11]], i32 276, i32 15, i32 15, i1 false)
; IR-DPP-NEXT: [[TMP13:%.*]] = fadd float [[TMP11]], [[TMP12]]
; IR-DPP-NEXT: [[TMP14:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP13]], i32 280, i32 15, i32 15, i1 false)
; IR-DPP-NEXT: [[TMP15:%.*]] = fadd float [[TMP13]], [[TMP14]]
; IR-DPP-NEXT: [[TMP16:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP15]], i32 322, i32 10, i32 15, i1 false)
; IR-DPP-NEXT: [[TMP17:%.*]] = fadd float [[TMP15]], [[TMP16]]
; IR-DPP-NEXT: [[TMP18:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP17]], i32 323, i32 12, i32 15, i1 false)
; IR-DPP-NEXT: [[TMP19:%.*]] = fadd float [[TMP17]], [[TMP18]]
; IR-DPP-NEXT: [[TMP20:%.*]] = call float @llvm.amdgcn.readlane.f32(float [[TMP19]], i32 63)
; IR-DPP-NEXT: [[TMP21:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP20]])
; IR-DPP-NEXT: [[TMP22:%.*]] = icmp eq i32 [[TMP6]], 0
; IR-DPP-NEXT: br i1 [[TMP22]], label [[TMP23:%.*]], label [[TMP25:%.*]]
; IR-DPP: 23:
; IR-DPP-NEXT: [[TMP24:%.*]] = atomicrmw fsub ptr addrspace(1) [[PTR:%.*]], float [[TMP21]] seq_cst, align 4
; IR-DPP-NEXT: br label [[TMP25]]
; IR-DPP: 25:
; IR-DPP-NEXT: ret void
;
%id.x = call i32 @llvm.amdgcn.workitem.id.x()
%divValue = bitcast i32 %id.x to float
%result = atomicrmw fsub ptr addrspace(1) %ptr, float %divValue seq_cst
ret void
}
attributes #0 = {"target-cpu"="gfx906"}
|
