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
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -march=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck --check-prefix=SI %s
; RUN: llc < %s -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs | FileCheck --check-prefix=FLAT %s
define amdgpu_kernel void @break_inserted_outside_of_loop(ptr addrspace(1) %out, i32 %a) {
; SI-LABEL: break_inserted_outside_of_loop:
; SI: ; %bb.0: ; %main_body
; SI-NEXT: s_load_dword s2, s[0:1], 0xb
; SI-NEXT: v_mbcnt_lo_u32_b32_e64 v0, -1, 0
; SI-NEXT: s_waitcnt lgkmcnt(0)
; SI-NEXT: v_and_b32_e32 v0, s2, v0
; SI-NEXT: v_and_b32_e32 v0, 1, v0
; SI-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; SI-NEXT: s_mov_b64 s[2:3], 0
; SI-NEXT: .LBB0_1: ; %ENDIF
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_and_b64 s[4:5], exec, vcc
; SI-NEXT: s_or_b64 s[2:3], s[4:5], s[2:3]
; SI-NEXT: s_andn2_b64 exec, exec, s[2:3]
; SI-NEXT: s_cbranch_execnz .LBB0_1
; SI-NEXT: ; %bb.2: ; %ENDLOOP
; SI-NEXT: s_or_b64 exec, exec, s[2:3]
; SI-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x9
; SI-NEXT: s_mov_b32 s3, 0xf000
; SI-NEXT: s_mov_b32 s2, -1
; SI-NEXT: v_mov_b32_e32 v0, 0
; SI-NEXT: s_waitcnt lgkmcnt(0)
; SI-NEXT: buffer_store_dword v0, off, s[0:3], 0
; SI-NEXT: s_endpgm
;
; FLAT-LABEL: break_inserted_outside_of_loop:
; FLAT: ; %bb.0: ; %main_body
; FLAT-NEXT: s_load_dword s2, s[0:1], 0x2c
; FLAT-NEXT: v_mbcnt_lo_u32_b32 v0, -1, 0
; FLAT-NEXT: s_waitcnt lgkmcnt(0)
; FLAT-NEXT: v_and_b32_e32 v0, s2, v0
; FLAT-NEXT: v_and_b32_e32 v0, 1, v0
; FLAT-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; FLAT-NEXT: s_mov_b64 s[2:3], 0
; FLAT-NEXT: .LBB0_1: ; %ENDIF
; FLAT-NEXT: ; =>This Inner Loop Header: Depth=1
; FLAT-NEXT: s_and_b64 s[4:5], exec, vcc
; FLAT-NEXT: s_or_b64 s[2:3], s[4:5], s[2:3]
; FLAT-NEXT: s_andn2_b64 exec, exec, s[2:3]
; FLAT-NEXT: s_cbranch_execnz .LBB0_1
; FLAT-NEXT: ; %bb.2: ; %ENDLOOP
; FLAT-NEXT: s_or_b64 exec, exec, s[2:3]
; FLAT-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
; FLAT-NEXT: s_mov_b32 s3, 0xf000
; FLAT-NEXT: s_mov_b32 s2, -1
; FLAT-NEXT: v_mov_b32_e32 v0, 0
; FLAT-NEXT: s_waitcnt lgkmcnt(0)
; FLAT-NEXT: buffer_store_dword v0, off, s[0:3], 0
; FLAT-NEXT: s_endpgm
main_body:
%tid = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) #0
%0 = and i32 %a, %tid
%1 = trunc i32 %0 to i1
br label %ENDIF
ENDLOOP:
store i32 0, ptr addrspace(1) %out
ret void
ENDIF:
br i1 %1, label %ENDLOOP, label %ENDIF
}
define amdgpu_kernel void @phi_cond_outside_loop(i32 %b) {
; SI-LABEL: phi_cond_outside_loop:
; SI: ; %bb.0: ; %entry
; SI-NEXT: v_mbcnt_lo_u32_b32_e64 v0, -1, 0
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v0
; SI-NEXT: s_mov_b64 s[2:3], 0
; SI-NEXT: s_mov_b64 s[4:5], 0
; SI-NEXT: s_and_saveexec_b64 s[6:7], vcc
; SI-NEXT: s_cbranch_execz .LBB1_2
; SI-NEXT: ; %bb.1: ; %else
; SI-NEXT: s_load_dword s0, s[0:1], 0x9
; SI-NEXT: s_waitcnt lgkmcnt(0)
; SI-NEXT: s_cmp_eq_u32 s0, 0
; SI-NEXT: s_cselect_b64 s[0:1], -1, 0
; SI-NEXT: s_and_b64 s[4:5], s[0:1], exec
; SI-NEXT: .LBB1_2: ; %endif
; SI-NEXT: s_or_b64 exec, exec, s[6:7]
; SI-NEXT: .LBB1_3: ; %loop
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_and_b64 s[0:1], exec, s[4:5]
; SI-NEXT: s_or_b64 s[2:3], s[0:1], s[2:3]
; SI-NEXT: s_andn2_b64 exec, exec, s[2:3]
; SI-NEXT: s_cbranch_execnz .LBB1_3
; SI-NEXT: ; %bb.4: ; %exit
; SI-NEXT: s_endpgm
;
; FLAT-LABEL: phi_cond_outside_loop:
; FLAT: ; %bb.0: ; %entry
; FLAT-NEXT: v_mbcnt_lo_u32_b32 v0, -1, 0
; FLAT-NEXT: v_cmp_ne_u32_e32 vcc, 0, v0
; FLAT-NEXT: s_mov_b64 s[2:3], 0
; FLAT-NEXT: s_mov_b64 s[4:5], 0
; FLAT-NEXT: s_and_saveexec_b64 s[6:7], vcc
; FLAT-NEXT: s_cbranch_execz .LBB1_2
; FLAT-NEXT: ; %bb.1: ; %else
; FLAT-NEXT: s_load_dword s0, s[0:1], 0x24
; FLAT-NEXT: s_waitcnt lgkmcnt(0)
; FLAT-NEXT: s_cmp_eq_u32 s0, 0
; FLAT-NEXT: s_cselect_b64 s[0:1], -1, 0
; FLAT-NEXT: s_and_b64 s[4:5], s[0:1], exec
; FLAT-NEXT: .LBB1_2: ; %endif
; FLAT-NEXT: s_or_b64 exec, exec, s[6:7]
; FLAT-NEXT: .LBB1_3: ; %loop
; FLAT-NEXT: ; =>This Inner Loop Header: Depth=1
; FLAT-NEXT: s_and_b64 s[0:1], exec, s[4:5]
; FLAT-NEXT: s_or_b64 s[2:3], s[0:1], s[2:3]
; FLAT-NEXT: s_andn2_b64 exec, exec, s[2:3]
; FLAT-NEXT: s_cbranch_execnz .LBB1_3
; FLAT-NEXT: ; %bb.4: ; %exit
; FLAT-NEXT: s_endpgm
entry:
%tid = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) #0
%0 = icmp eq i32 %tid , 0
br i1 %0, label %if, label %else
if:
br label %endif
else:
%1 = icmp eq i32 %b, 0
br label %endif
endif:
%2 = phi i1 [0, %if], [%1, %else]
br label %loop
loop:
br i1 %2, label %exit, label %loop
exit:
ret void
}
define amdgpu_kernel void @switch_unreachable(ptr addrspace(1) %g, ptr addrspace(3) %l, i32 %x) nounwind {
; SI-LABEL: switch_unreachable:
; SI: ; %bb.0: ; %centry
;
; FLAT-LABEL: switch_unreachable:
; FLAT: ; %bb.0: ; %centry
centry:
switch i32 %x, label %sw.default [
i32 0, label %sw.bb
i32 60, label %sw.bb
]
sw.bb:
unreachable
sw.default:
unreachable
sw.epilog:
ret void
}
declare float @llvm.fabs.f32(float) nounwind readnone
define amdgpu_kernel void @loop_land_info_assert(i32 %c0, i32 %c1, i32 %c2, i32 %c3, i32 %x, i32 %y, i1 %arg) nounwind {
; SI-LABEL: loop_land_info_assert:
; SI: ; %bb.0: ; %entry
; SI-NEXT: s_load_dword s2, s[0:1], 0xa
; SI-NEXT: s_waitcnt lgkmcnt(0)
; SI-NEXT: s_cmp_lt_i32 s2, 4
; SI-NEXT: s_cbranch_scc1 .LBB3_4
; SI-NEXT: ; %bb.1: ; %for.cond.preheader
; SI-NEXT: s_load_dword s0, s[0:1], 0xc
; SI-NEXT: s_waitcnt lgkmcnt(0)
; SI-NEXT: s_cmpk_lt_i32 s0, 0x3e8
; SI-NEXT: s_cbranch_scc0 .LBB3_4
; SI-NEXT: ; %bb.2: ; %for.body
; SI-NEXT: s_and_b64 vcc, exec, 0
; SI-NEXT: .LBB3_3: ; %self.loop
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_mov_b64 vcc, vcc
; SI-NEXT: s_cbranch_vccz .LBB3_3
; SI-NEXT: .LBB3_4: ; %DummyReturnBlock
; SI-NEXT: s_endpgm
;
; FLAT-LABEL: loop_land_info_assert:
; FLAT: ; %bb.0: ; %entry
; FLAT-NEXT: s_load_dword s2, s[0:1], 0x28
; FLAT-NEXT: s_waitcnt lgkmcnt(0)
; FLAT-NEXT: s_cmp_lt_i32 s2, 4
; FLAT-NEXT: s_cbranch_scc1 .LBB3_4
; FLAT-NEXT: ; %bb.1: ; %for.cond.preheader
; FLAT-NEXT: s_load_dword s0, s[0:1], 0x30
; FLAT-NEXT: s_waitcnt lgkmcnt(0)
; FLAT-NEXT: s_cmpk_lt_i32 s0, 0x3e8
; FLAT-NEXT: s_cbranch_scc0 .LBB3_4
; FLAT-NEXT: ; %bb.2: ; %for.body
; FLAT-NEXT: s_and_b64 vcc, exec, 0
; FLAT-NEXT: .LBB3_3: ; %self.loop
; FLAT-NEXT: ; =>This Inner Loop Header: Depth=1
; FLAT-NEXT: s_mov_b64 vcc, vcc
; FLAT-NEXT: s_cbranch_vccz .LBB3_3
; FLAT-NEXT: .LBB3_4: ; %DummyReturnBlock
; FLAT-NEXT: s_endpgm
entry:
%cmp = icmp sgt i32 %c0, 0
br label %while.cond.outer
while.cond.outer:
%tmp = load float, ptr addrspace(1) undef
br label %while.cond
while.cond:
%cmp1 = icmp slt i32 %c1, 4
br i1 %cmp1, label %convex.exit, label %for.cond
convex.exit:
%or = or i1 %cmp, %cmp1
br i1 %or, label %return, label %if.end
if.end:
%tmp3 = call float @llvm.fabs.f32(float %tmp) nounwind readnone
%cmp2 = fcmp olt float %tmp3, 0x3E80000000000000
br i1 %cmp2, label %if.else, label %while.cond.outer
if.else:
store volatile i32 3, ptr addrspace(1) undef, align 4
br label %while.cond
for.cond:
%cmp3 = icmp slt i32 %c3, 1000
br i1 %cmp3, label %for.body, label %return
for.body:
br i1 %cmp3, label %self.loop, label %if.end.2
if.end.2:
%or.cond2 = or i1 %cmp3, %arg
br i1 %or.cond2, label %return, label %for.cond
self.loop:
br label %self.loop
return:
ret void
}
declare i32 @llvm.amdgcn.mbcnt.lo(i32, i32) #0
attributes #0 = { nounwind readnone }
|
