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
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,WAVE64 %s
; RUN: llc -global-isel -march=amdgcn -mcpu=gfx1031 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,WAVE32 %s
; End to end tests for scalar vs. vector boolean legalization strategies.
define amdgpu_ps float @select_vgpr_sgpr_trunc_cond(i32 inreg %a, i32 %b, i32 %c) {
; WAVE64-LABEL: select_vgpr_sgpr_trunc_cond:
; WAVE64: ; %bb.0:
; WAVE64-NEXT: s_and_b32 s0, 1, s0
; WAVE64-NEXT: v_cmp_ne_u32_e64 vcc, 0, s0
; WAVE64-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc
; WAVE64-NEXT: ; return to shader part epilog
;
; WAVE32-LABEL: select_vgpr_sgpr_trunc_cond:
; WAVE32: ; %bb.0:
; WAVE32-NEXT: s_and_b32 s0, 1, s0
; WAVE32-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0
; WAVE32-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo
; WAVE32-NEXT: ; return to shader part epilog
%cc = trunc i32 %a to i1
%r = select i1 %cc, i32 %b, i32 %c
%r.f = bitcast i32 %r to float
ret float %r.f
}
define amdgpu_ps float @select_vgpr_sgpr_trunc_and_cond(i32 inreg %a.0, i32 inreg %a.1, i32 %b, i32 %c) {
; WAVE64-LABEL: select_vgpr_sgpr_trunc_and_cond:
; WAVE64: ; %bb.0:
; WAVE64-NEXT: s_and_b32 s0, s0, s1
; WAVE64-NEXT: s_and_b32 s0, 1, s0
; WAVE64-NEXT: v_cmp_ne_u32_e64 vcc, 0, s0
; WAVE64-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc
; WAVE64-NEXT: ; return to shader part epilog
;
; WAVE32-LABEL: select_vgpr_sgpr_trunc_and_cond:
; WAVE32: ; %bb.0:
; WAVE32-NEXT: s_and_b32 s0, s0, s1
; WAVE32-NEXT: s_and_b32 s0, 1, s0
; WAVE32-NEXT: v_cmp_ne_u32_e64 vcc_lo, 0, s0
; WAVE32-NEXT: v_cndmask_b32_e32 v0, v1, v0, vcc_lo
; WAVE32-NEXT: ; return to shader part epilog
%cc.0 = trunc i32 %a.0 to i1
%cc.1 = trunc i32 %a.1 to i1
%and = and i1 %cc.0, %cc.1
%r = select i1 %and, i32 %b, i32 %c
%r.f = bitcast i32 %r to float
ret float %r.f
}
define amdgpu_ps i32 @select_sgpr_trunc_and_cond(i32 inreg %a.0, i32 inreg %a.1, i32 inreg %b, i32 inreg %c) {
; GCN-LABEL: select_sgpr_trunc_and_cond:
; GCN: ; %bb.0:
; GCN-NEXT: s_and_b32 s0, s0, s1
; GCN-NEXT: s_and_b32 s0, s0, 1
; GCN-NEXT: s_cmp_lg_u32 s0, 0
; GCN-NEXT: s_cselect_b32 s0, s2, s3
; GCN-NEXT: ; return to shader part epilog
%cc.0 = trunc i32 %a.0 to i1
%cc.1 = trunc i32 %a.1 to i1
%and = and i1 %cc.0, %cc.1
%r = select i1 %and, i32 %b, i32 %c
ret i32 %r
}
define amdgpu_kernel void @sgpr_trunc_brcond(i32 %cond) {
; WAVE64-LABEL: sgpr_trunc_brcond:
; WAVE64: ; %bb.0: ; %entry
; WAVE64-NEXT: s_load_dword s0, s[0:1], 0x24
; WAVE64-NEXT: s_waitcnt lgkmcnt(0)
; WAVE64-NEXT: s_xor_b32 s0, s0, 1
; WAVE64-NEXT: s_and_b32 s0, s0, 1
; WAVE64-NEXT: s_cmp_lg_u32 s0, 0
; WAVE64-NEXT: s_cbranch_scc1 .LBB3_2
; WAVE64-NEXT: ; %bb.1: ; %bb0
; WAVE64-NEXT: v_mov_b32_e32 v0, 0
; WAVE64-NEXT: global_store_dword v[0:1], v0, off
; WAVE64-NEXT: s_waitcnt vmcnt(0)
; WAVE64-NEXT: .LBB3_2: ; %bb1
; WAVE64-NEXT: v_mov_b32_e32 v0, 1
; WAVE64-NEXT: global_store_dword v[0:1], v0, off
; WAVE64-NEXT: s_waitcnt vmcnt(0)
;
; WAVE32-LABEL: sgpr_trunc_brcond:
; WAVE32: ; %bb.0: ; %entry
; WAVE32-NEXT: s_load_dword s0, s[0:1], 0x24
; WAVE32-NEXT: s_waitcnt lgkmcnt(0)
; WAVE32-NEXT: s_xor_b32 s0, s0, 1
; WAVE32-NEXT: s_and_b32 s0, s0, 1
; WAVE32-NEXT: s_cmp_lg_u32 s0, 0
; WAVE32-NEXT: s_cbranch_scc1 .LBB3_2
; WAVE32-NEXT: ; %bb.1: ; %bb0
; WAVE32-NEXT: v_mov_b32_e32 v0, 0
; WAVE32-NEXT: global_store_dword v[0:1], v0, off
; WAVE32-NEXT: s_waitcnt_vscnt null, 0x0
; WAVE32-NEXT: .LBB3_2: ; %bb1
; WAVE32-NEXT: v_mov_b32_e32 v0, 1
; WAVE32-NEXT: global_store_dword v[0:1], v0, off
; WAVE32-NEXT: s_waitcnt_vscnt null, 0x0
entry:
%trunc = trunc i32 %cond to i1
br i1 %trunc, label %bb0, label %bb1
bb0:
store volatile i32 0, ptr addrspace(1) undef
unreachable
bb1:
store volatile i32 1, ptr addrspace(1) undef
unreachable
}
define amdgpu_kernel void @brcond_sgpr_trunc_and(i32 %cond0, i32 %cond1) {
; WAVE64-LABEL: brcond_sgpr_trunc_and:
; WAVE64: ; %bb.0: ; %entry
; WAVE64-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
; WAVE64-NEXT: s_waitcnt lgkmcnt(0)
; WAVE64-NEXT: s_and_b32 s0, s0, s1
; WAVE64-NEXT: s_xor_b32 s0, s0, 1
; WAVE64-NEXT: s_and_b32 s0, s0, 1
; WAVE64-NEXT: s_cmp_lg_u32 s0, 0
; WAVE64-NEXT: s_cbranch_scc1 .LBB4_2
; WAVE64-NEXT: ; %bb.1: ; %bb0
; WAVE64-NEXT: v_mov_b32_e32 v0, 0
; WAVE64-NEXT: global_store_dword v[0:1], v0, off
; WAVE64-NEXT: s_waitcnt vmcnt(0)
; WAVE64-NEXT: .LBB4_2: ; %bb1
; WAVE64-NEXT: v_mov_b32_e32 v0, 1
; WAVE64-NEXT: global_store_dword v[0:1], v0, off
; WAVE64-NEXT: s_waitcnt vmcnt(0)
;
; WAVE32-LABEL: brcond_sgpr_trunc_and:
; WAVE32: ; %bb.0: ; %entry
; WAVE32-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
; WAVE32-NEXT: s_waitcnt lgkmcnt(0)
; WAVE32-NEXT: s_and_b32 s0, s0, s1
; WAVE32-NEXT: s_xor_b32 s0, s0, 1
; WAVE32-NEXT: s_and_b32 s0, s0, 1
; WAVE32-NEXT: s_cmp_lg_u32 s0, 0
; WAVE32-NEXT: s_cbranch_scc1 .LBB4_2
; WAVE32-NEXT: ; %bb.1: ; %bb0
; WAVE32-NEXT: v_mov_b32_e32 v0, 0
; WAVE32-NEXT: global_store_dword v[0:1], v0, off
; WAVE32-NEXT: s_waitcnt_vscnt null, 0x0
; WAVE32-NEXT: .LBB4_2: ; %bb1
; WAVE32-NEXT: v_mov_b32_e32 v0, 1
; WAVE32-NEXT: global_store_dword v[0:1], v0, off
; WAVE32-NEXT: s_waitcnt_vscnt null, 0x0
entry:
%trunc0 = trunc i32 %cond0 to i1
%trunc1 = trunc i32 %cond1 to i1
%and = and i1 %trunc0, %trunc1
br i1 %and, label %bb0, label %bb1
bb0:
store volatile i32 0, ptr addrspace(1) undef
unreachable
bb1:
store volatile i32 1, ptr addrspace(1) undef
unreachable
}
|
