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
| ; RUN: opt -S -mtriple=amdgcn-- -structurizecfg -si-annotate-control-flow < %s | FileCheck -check-prefix=OPT %s
; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; OPT-LABEL: {{^}}define amdgpu_vs void @multi_else_break(
; OPT: main_body:
; OPT: LOOP.outer:
; OPT: LOOP:
; OPT: [[if:%[0-9]+]] = call { i1, i64 } @llvm.amdgcn.if.i64(
; OPT: [[if_exec:%[0-9]+]] = extractvalue { i1, i64 } [[if]], 1
;
; OPT: Flow:
;
; Ensure two if.break calls, for both the inner and outer loops
; OPT: call void @llvm.amdgcn.end.cf
; OPT-NEXT: call i64 @llvm.amdgcn.if.break.i64.i64(i1
; OPT-NEXT: call i1 @llvm.amdgcn.loop.i64(i64
; OPT-NEXT: call i64 @llvm.amdgcn.if.break.i64.i64(i1
;
; OPT: Flow1:
; GCN-LABEL: {{^}}multi_else_break:
; GCN: ; %main_body
; GCN: s_mov_b64 [[LEFT_OUTER:s\[[0-9]+:[0-9]+\]]], 0{{$}}
; GCN: [[FLOW2:BB[0-9]+_[0-9]+]]: ; %Flow2
; GCN: s_or_b64 exec, exec, [[TMP0:s\[[0-9]+:[0-9]+\]]]
; GCN: s_and_b64 [[TMP1:s\[[0-9]+:[0-9]+\]]], exec, [[BREAK_OUTER:s\[[0-9]+:[0-9]+\]]]
; GCN: s_or_b64 [[TMP1]], [[TMP1]], [[LEFT_OUTER]]
; GCN: s_mov_b64 [[LEFT_OUTER]], [[TMP1]]
; GCN: s_andn2_b64 exec, exec, [[TMP1]]
; GCN: s_cbranch_execz [[IF_BLOCK:BB[0-9]+_[0-9]+]]
; GCN: [[OUTER_LOOP:BB[0-9]+_[0-9]+]]: ; %LOOP.outer{{$}}
; GCN: s_mov_b64 [[LEFT_INNER:s\[[0-9]+:[0-9]+\]]], 0{{$}}
; GCN: ; %Flow
; GCN: s_or_b64 exec, exec, [[SAVE_EXEC:s\[[0-9]+:[0-9]+\]]]
; GCN: s_and_b64 [[TMP0]], exec, [[BREAK_INNER:s\[[0-9]+:[0-9]+\]]]
; GCN: s_or_b64 [[TMP0]], [[TMP0]], [[LEFT_INNER]]
; GCN: s_mov_b64 [[LEFT_INNER]], [[TMP0]]
; GCN: s_andn2_b64 exec, exec, [[TMP0]]
; GCN: s_cbranch_execz [[FLOW2]]
; GCN: [[INNER_LOOP:BB[0-9]+_[0-9]+]]: ; %LOOP{{$}}
; GCN: s_and_saveexec_b64 [[SAVE_EXEC]], vcc
; FIXME: duplicate comparison
; GCN: ; %ENDIF
; GCN-DAG: v_cmp_eq_u32_e32 vcc,
; GCN-DAG: v_cmp_ne_u32_e64 [[TMP51NEG:s\[[0-9]+:[0-9]+\]]],
; GCN: [[IF_BLOCK]]: ; %IF
; GCN-NEXT: s_endpgm
define amdgpu_vs void @multi_else_break(<4 x float> %vec, i32 %ub, i32 %cont) {
main_body:
br label %LOOP.outer
LOOP.outer: ; preds = %ENDIF, %main_body
%tmp43 = phi i32 [ 0, %main_body ], [ %tmp47, %ENDIF ]
br label %LOOP
LOOP: ; preds = %ENDIF, %LOOP.outer
%tmp45 = phi i32 [ %tmp43, %LOOP.outer ], [ %tmp47, %ENDIF ]
%tmp47 = add i32 %tmp45, 1
%tmp48 = icmp slt i32 %tmp45, %ub
br i1 %tmp48, label %ENDIF, label %IF
IF: ; preds = %LOOP
ret void
ENDIF: ; preds = %LOOP
%tmp51 = icmp eq i32 %tmp47, %cont
br i1 %tmp51, label %LOOP, label %LOOP.outer
}
; OPT-LABEL: define amdgpu_kernel void @multi_if_break_loop(
; OPT: llvm.amdgcn.if.break
; OPT: llvm.amdgcn.loop
; OPT: llvm.amdgcn.if.break
; OPT: llvm.amdgcn.end.cf
; GCN-LABEL: {{^}}multi_if_break_loop:
; GCN: s_mov_b64 [[BROKEN_THREADS_MASK:s\[[0-9]+:[0-9]+\]]], 0{{$}}
; GCN: [[LOOP:BB[0-9]+_[0-9]+]]: ; %Flow4
; GCN: s_and_b64 [[BROKEN_THREADS_MASK]], exec, [[BROKEN_THREADS_MASK]]
; GCN: s_or_b64 [[BROKEN_THREADS_MASK]], [[BROKEN_THREADS_MASK]], [[SAVED:s\[[0-9]+:[0-9]+\]]]
; GCN: s_andn2_b64 exec, exec, [[BROKEN_THREADS_MASK]]
; GCN-NEXT: s_cbranch_execz [[LOOP_EXIT:BB[0-9]+_[0-9]+]]
; GCN: ; %bb1{{$}}
; GCN: buffer_load_dword [[LOAD0:v[0-9]+]],
; GCN: s_mov_b64 [[SAVED]], [[BROKEN_THREADS_MASK]]
; GCN: ; %LeafBlock1
; GCN: v_cmp_eq_u32_e32 vcc, 1, [[LOAD0]]
; GCN: s_and_b64 vcc, exec, vcc
; GCN: s_cbranch_vccz [[FLOW:BB[0-9]+_[0-9]+]]
; GCN: ; %case1
; GCN: buffer_load_dword [[LOAD2:v[0-9]+]],
; GCN: v_cmp_ge_i32_e32 vcc, {{v[0-9]+}}, [[LOAD2]]
; GCN: s_orn2_b64 [[BROKEN_THREADS_MASK]], vcc, exec
; GCN: BB1_{{[0-9]+}}:
; GCN: s_mov_b64 [[FALSE_MASK:s\[[0-9]+:[0-9]+\]]], 0
; GCN: s_and_b64 vcc, exec, [[FALSE_MASK]]
; GCN: s_cbranch_vccz [[LOOP]]
; GCN: ; %LeafBlock
; GCN: v_cmp_eq_u32_e32 vcc, 0, [[LOAD0]]
; GCN: s_and_b64 vcc, exec, vcc
; GCN: s_cbranch_vccz [[LOOP]]
; GCN: ; %case0
; GCN: buffer_load_dword [[LOAD1:v[0-9]+]],
; GCN-DAG: v_cmp_ge_i32_e32 vcc, {{v[0-9]+}}, [[LOAD1]]
; GCN: s_andn2_b64 [[BROKEN_THREADS_MASK]], [[BROKEN_THREADS_MASK]], exec
; GCN: s_and_b64 [[TMP_MASK:s\[[0-9]+:[0-9]+\]]], vcc, exec
; GCN: s_or_b64 [[BROKEN_THREADS_MASK]], [[BROKEN_THREADS_MASK]], [[TMP_MASK]]
; GCN: s_branch [[LOOP]]
; GCN: [[LOOP_EXIT]]: ; %Flow6
; GCN: s_or_b64 exec, exec, [[BROKEN_THREADS_MASK]]
define amdgpu_kernel void @multi_if_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, %case0 ], [ %lsr.iv.next, %case1 ]
%lsr.iv.next = add i32 %lsr.iv, 1
%cmp0 = icmp slt i32 %lsr.iv.next, 0
%load0 = load volatile i32, i32 addrspace(1)* undef, align 4
switch i32 %load0, label %bb9 [
i32 0, label %case0
i32 1, label %case1
]
case0:
%load1 = load volatile i32, i32 addrspace(1)* undef, align 4
%cmp1 = icmp slt i32 %tmp, %load1
br i1 %cmp1, label %bb1, label %bb9
case1:
%load2 = load volatile i32, i32 addrspace(1)* undef, align 4
%cmp2 = icmp slt i32 %tmp, %load2
br i1 %cmp2, label %bb1, label %bb9
bb9:
ret void
}
declare i32 @llvm.amdgcn.workitem.id.x() #1
attributes #0 = { nounwind }
attributes #1 = { nounwind readnone }
|