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
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck %s
; Make sure the branch targets are correct after lowering llvm.amdgcn.if
define i32 @divergent_if_swap_brtarget_order0(i32 %value) {
; CHECK-LABEL: divergent_if_swap_brtarget_order0:
; CHECK: ; %bb.0: ; %entry
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT: v_cmp_ne_u32_e32 vcc, 0, v0
; CHECK-NEXT: ; implicit-def: $vgpr0
; CHECK-NEXT: s_and_saveexec_b64 s[4:5], vcc
; CHECK-NEXT: s_cbranch_execz .LBB0_2
; CHECK-NEXT: ; %bb.1: ; %if.true
; CHECK-NEXT: global_load_dword v0, v[0:1], off glc
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: .LBB0_2: ; %endif
; CHECK-NEXT: s_or_b64 exec, exec, s[4:5]
; CHECK-NEXT: s_setpc_b64 s[30:31]
entry:
%c = icmp ne i32 %value, 0
br i1 %c, label %if.true, label %endif
if.true:
%val = load volatile i32, ptr addrspace(1) undef
br label %endif
endif:
%v = phi i32 [ %val, %if.true ], [ undef, %entry ]
ret i32 %v
}
define i32 @divergent_if_swap_brtarget_order1(i32 %value) {
; CHECK-LABEL: divergent_if_swap_brtarget_order1:
; CHECK: ; %bb.0: ; %entry
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT: v_cmp_ne_u32_e32 vcc, 0, v0
; CHECK-NEXT: ; implicit-def: $vgpr0
; CHECK-NEXT: s_and_saveexec_b64 s[4:5], vcc
; CHECK-NEXT: s_cbranch_execz .LBB1_2
; CHECK-NEXT: ; %bb.1: ; %if.true
; CHECK-NEXT: global_load_dword v0, v[0:1], off glc
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: .LBB1_2: ; %endif
; CHECK-NEXT: s_or_b64 exec, exec, s[4:5]
; CHECK-NEXT: s_setpc_b64 s[30:31]
entry:
%c = icmp ne i32 %value, 0
br i1 %c, label %if.true, label %endif
endif:
%v = phi i32 [ %val, %if.true ], [ undef, %entry ]
ret i32 %v
if.true:
%val = load volatile i32, ptr addrspace(1) undef
br label %endif
}
; Make sure and 1 is inserted on llvm.amdgcn.if
define i32 @divergent_if_nonboolean_condition0(i32 %value) {
; CHECK-LABEL: divergent_if_nonboolean_condition0:
; CHECK: ; %bb.0: ; %entry
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT: v_and_b32_e32 v0, 1, v0
; CHECK-NEXT: v_cmp_ne_u32_e32 vcc, 0, v0
; CHECK-NEXT: ; implicit-def: $vgpr0
; CHECK-NEXT: s_and_saveexec_b64 s[4:5], vcc
; CHECK-NEXT: s_cbranch_execz .LBB2_2
; CHECK-NEXT: ; %bb.1: ; %if.true
; CHECK-NEXT: global_load_dword v0, v[0:1], off glc
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: .LBB2_2: ; %endif
; CHECK-NEXT: s_or_b64 exec, exec, s[4:5]
; CHECK-NEXT: s_setpc_b64 s[30:31]
entry:
%c = trunc i32 %value to i1
br i1 %c, label %if.true, label %endif
if.true:
%val = load volatile i32, ptr addrspace(1) undef
br label %endif
endif:
%v = phi i32 [ %val, %if.true ], [ undef, %entry ]
ret i32 %v
}
; Make sure and 1 is inserted on llvm.amdgcn.if
define i32 @divergent_if_nonboolean_condition1(ptr addrspace(1) %ptr) {
; CHECK-LABEL: divergent_if_nonboolean_condition1:
; CHECK: ; %bb.0: ; %entry
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT: global_load_dword v0, v[0:1], off
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: v_and_b32_e32 v0, 1, v0
; CHECK-NEXT: v_cmp_ne_u32_e32 vcc, 0, v0
; CHECK-NEXT: ; implicit-def: $vgpr0
; CHECK-NEXT: s_and_saveexec_b64 s[4:5], vcc
; CHECK-NEXT: s_cbranch_execz .LBB3_2
; CHECK-NEXT: ; %bb.1: ; %if.true
; CHECK-NEXT: global_load_dword v0, v[0:1], off glc
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: .LBB3_2: ; %endif
; CHECK-NEXT: s_or_b64 exec, exec, s[4:5]
; CHECK-NEXT: s_setpc_b64 s[30:31]
entry:
%value = load i32, ptr addrspace(1) %ptr
%c = trunc i32 %value to i1
br i1 %c, label %if.true, label %endif
if.true:
%val = load volatile i32, ptr addrspace(1) undef
br label %endif
endif:
%v = phi i32 [ %val, %if.true ], [ undef, %entry ]
ret i32 %v
}
@external_constant = external addrspace(4) constant i32, align 4
@const.ptr = external addrspace(4) constant ptr, align 4
; Make sure this case compiles. G_ICMP was mis-mapped due to having
; the result register class constrained by llvm.amdgcn.if lowering.
define void @constrained_if_register_class() {
; CHECK-LABEL: constrained_if_register_class:
; CHECK: ; %bb.0: ; %bb
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT: s_getpc_b64 s[4:5]
; CHECK-NEXT: s_add_u32 s4, s4, external_constant@gotpcrel32@lo+4
; CHECK-NEXT: s_addc_u32 s5, s5, external_constant@gotpcrel32@hi+12
; CHECK-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x0
; CHECK-NEXT: s_waitcnt lgkmcnt(0)
; CHECK-NEXT: s_load_dword s4, s[4:5], 0x0
; CHECK-NEXT: s_waitcnt lgkmcnt(0)
; CHECK-NEXT: s_cmp_lg_u32 s4, 0
; CHECK-NEXT: s_cbranch_scc0 .LBB4_2
; CHECK-NEXT: .LBB4_1: ; %bb12
; CHECK-NEXT: s_setpc_b64 s[30:31]
; CHECK-NEXT: .LBB4_2: ; %bb2
; CHECK-NEXT: s_getpc_b64 s[4:5]
; CHECK-NEXT: s_add_u32 s4, s4, const.ptr@gotpcrel32@lo+4
; CHECK-NEXT: s_addc_u32 s5, s5, const.ptr@gotpcrel32@hi+12
; CHECK-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x0
; CHECK-NEXT: v_mov_b32_e32 v0, 0
; CHECK-NEXT: s_waitcnt lgkmcnt(0)
; CHECK-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x0
; CHECK-NEXT: s_waitcnt lgkmcnt(0)
; CHECK-NEXT: global_load_dword v0, v0, s[4:5]
; CHECK-NEXT: s_mov_b32 s4, -1
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: v_cmp_gt_f32_e32 vcc, 1.0, v0
; CHECK-NEXT: s_cbranch_vccnz .LBB4_4
; CHECK-NEXT: ; %bb.3: ; %bb7
; CHECK-NEXT: s_mov_b32 s4, 0
; CHECK-NEXT: .LBB4_4: ; %bb8
; CHECK-NEXT: s_cmp_lg_u32 s4, 0
; CHECK-NEXT: s_cbranch_scc1 .LBB4_1
; CHECK-NEXT: ; %bb.5: ; %bb11
; CHECK-NEXT: v_mov_b32_e32 v0, 4.0
; CHECK-NEXT: buffer_store_dword v0, v0, s[0:3], 0 offen
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: s_setpc_b64 s[30:31]
bb:
%tmp = load i32, ptr addrspace(4) @external_constant
%tmp1 = icmp ne i32 %tmp, 0
br i1 %tmp1, label %bb12, label %bb2
bb2:
%ptr = load ptr, ptr addrspace(4) @const.ptr
%tmp4 = load float, ptr %ptr, align 4
%tmp5 = fcmp olt float %tmp4, 1.0
%tmp6 = or i1 %tmp5, false
br i1 %tmp6, label %bb8, label %bb7
bb7:
br label %bb8
bb8:
%tmp9 = phi i32 [ 0, %bb7 ], [ -1, %bb2 ]
%tmp10 = icmp eq i32 %tmp9, 0
br i1 %tmp10, label %bb11, label %bb12
bb11:
store float 4.0, ptr addrspace(5) undef, align 4
br label %bb12
bb12:
ret void
}
define amdgpu_kernel void @break_loop(i32 %arg) {
; CHECK-LABEL: break_loop:
; CHECK: ; %bb.0: ; %bb
; CHECK-NEXT: s_load_dword s2, s[4:5], 0x0
; CHECK-NEXT: s_mov_b64 s[0:1], 0
; CHECK-NEXT: ; implicit-def: $vgpr1
; CHECK-NEXT: s_waitcnt lgkmcnt(0)
; CHECK-NEXT: v_subrev_u32_e32 v0, s2, v0
; CHECK-NEXT: ; implicit-def: $sgpr2_sgpr3
; CHECK-NEXT: s_branch .LBB5_3
; CHECK-NEXT: .LBB5_1: ; %bb4
; CHECK-NEXT: ; in Loop: Header=BB5_3 Depth=1
; CHECK-NEXT: global_load_dword v2, v[0:1], off glc
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: s_andn2_b64 s[2:3], s[2:3], exec
; CHECK-NEXT: v_cmp_ge_i32_e32 vcc, v0, v2
; CHECK-NEXT: s_and_b64 s[4:5], exec, vcc
; CHECK-NEXT: s_or_b64 s[2:3], s[2:3], s[4:5]
; CHECK-NEXT: .LBB5_2: ; %Flow
; CHECK-NEXT: ; in Loop: Header=BB5_3 Depth=1
; CHECK-NEXT: s_and_b64 s[4:5], exec, s[2:3]
; CHECK-NEXT: s_or_b64 s[0:1], s[4:5], s[0:1]
; CHECK-NEXT: s_andn2_b64 exec, exec, s[0:1]
; CHECK-NEXT: s_cbranch_execz .LBB5_5
; CHECK-NEXT: .LBB5_3: ; %bb1
; CHECK-NEXT: ; =>This Inner Loop Header: Depth=1
; CHECK-NEXT: v_add_u32_e32 v1, 1, v1
; CHECK-NEXT: s_andn2_b64 s[2:3], s[2:3], exec
; CHECK-NEXT: s_and_b64 s[4:5], exec, -1
; CHECK-NEXT: v_cmp_le_i32_e32 vcc, 0, v1
; CHECK-NEXT: s_or_b64 s[2:3], s[2:3], s[4:5]
; CHECK-NEXT: s_cbranch_vccz .LBB5_1
; CHECK-NEXT: ; %bb.4: ; in Loop: Header=BB5_3 Depth=1
; CHECK-NEXT: ; implicit-def: $vgpr1
; CHECK-NEXT: s_branch .LBB5_2
; CHECK-NEXT: .LBB5_5: ; %bb9
; CHECK-NEXT: s_endpgm
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, ptr addrspace(1) undef, align 4
%cmp1 = icmp slt i32 %tmp, %load
br i1 %cmp1, label %bb1, label %bb9
bb9:
ret void
}
declare i32 @llvm.amdgcn.workitem.id.x()
|
