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
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck -check-prefix=GFX90A %s
define amdgpu_ps void @atomic_swap_1d_agpr(<8 x i32> inreg %rsrc, i32 %s) {
; GFX90A-LABEL: atomic_swap_1d_agpr:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def a0
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: image_atomic_swap a0, v0, s[0:7] dmask:0x1 unorm glc
; GFX90A-NEXT: s_endpgm
%data = call i32 asm "; def $0", "=a"()
%v = call i32 @llvm.amdgcn.image.atomic.swap.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
call void asm "; use $0", "a"(i32 %v)
ret void
}
define amdgpu_ps void @atomic_add_2d_agpr(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
; GFX90A-LABEL: atomic_add_2d_agpr:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def a0
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: image_atomic_add a0, v[0:1], s[0:7] dmask:0x1 unorm glc
; GFX90A-NEXT: s_endpgm
%data = call i32 asm "; def $0", "=a"()
%v = call i32 @llvm.amdgcn.image.atomic.add.2d.i32.i32(i32 %data, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
call void asm "; use $0", "a"(i32 %v)
ret void
}
; FIXME: This should directly use the AGPRs
define amdgpu_ps void @atomic_cmpswap_1d_agpr(<8 x i32> inreg %rsrc, i32 %s) {
; GFX90A-LABEL: atomic_cmpswap_1d_agpr:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def a0
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def a1
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: v_accvgpr_read_b32 v2, a0
; GFX90A-NEXT: v_accvgpr_read_b32 v3, a1
; GFX90A-NEXT: image_atomic_cmpswap v[2:3], v0, s[0:7] dmask:0x3 unorm glc
; GFX90A-NEXT: s_endpgm
%cmp = call i32 asm "; def $0", "=a"()
%swap = call i32 asm "; def $0", "=a"()
%v = call i32 @llvm.amdgcn.image.atomic.cmpswap.1d.i32.i32(i32 %cmp, i32 %swap, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
call void asm "; use $0", "a"(i32 %v)
ret void
}
define amdgpu_ps void @atomic_swap_1d_i64_agpr(<8 x i32> inreg %rsrc, i32 %s) {
; GFX90A-LABEL: atomic_swap_1d_i64_agpr:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def a[0:1]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: image_atomic_swap a[0:1], v0, s[0:7] dmask:0x3 unorm glc
; GFX90A-NEXT: s_endpgm
%data = call i64 asm "; def $0", "=a"()
%v = call i64 @llvm.amdgcn.image.atomic.swap.1d.i64.i32(i64 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
call void asm "; use $0", "a"(i64 %v)
ret void
}
define amdgpu_ps void @atomic_cmpswap_1d_64_agpr(<8 x i32> inreg %rsrc, i32 %s) {
; GFX90A-LABEL: atomic_cmpswap_1d_64_agpr:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def a[0:1]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: v_accvgpr_read_b32 v3, a1
; GFX90A-NEXT: v_accvgpr_read_b32 v2, a0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def a[0:1]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: v_accvgpr_read_b32 v5, a1
; GFX90A-NEXT: v_accvgpr_read_b32 v4, a0
; GFX90A-NEXT: image_atomic_cmpswap v[2:5], v0, s[0:7] dmask:0xf unorm glc
; GFX90A-NEXT: s_endpgm
%cmp = call i64 asm "; def $0", "=a"()
%swap = call i64 asm "; def $0", "=a"()
%v = call i64 @llvm.amdgcn.image.atomic.cmpswap.1d.i64.i32(i64 %cmp, i64 %swap, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
call void asm "; use $0", "a"(i64 %v)
ret void
}
define amdgpu_ps void @atomic_swap_1d_agpr_noret(<8 x i32> inreg %rsrc, i32 %s) {
; GFX90A-LABEL: atomic_swap_1d_agpr_noret:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def a0
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: image_atomic_swap a0, v0, s[0:7] dmask:0x1 unorm
; GFX90A-NEXT: s_endpgm
%data = call i32 asm "; def $0", "=a"()
%unused = call i32 @llvm.amdgcn.image.atomic.swap.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
ret void
}
define amdgpu_ps void @atomic_add_2d_agpr_noret(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
; GFX90A-LABEL: atomic_add_2d_agpr_noret:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def a0
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: image_atomic_add a0, v[0:1], s[0:7] dmask:0x1 unorm
; GFX90A-NEXT: s_endpgm
%data = call i32 asm "; def $0", "=a"()
%unused = call i32 @llvm.amdgcn.image.atomic.add.2d.i32.i32(i32 %data, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
ret void
}
define amdgpu_ps void @atomic_cmpswap_1d_agpr_noret(<8 x i32> inreg %rsrc, i32 %s) {
; GFX90A-LABEL: atomic_cmpswap_1d_agpr_noret:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def a0
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def a1
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: image_atomic_cmpswap a[0:1], v0, s[0:7] dmask:0x3 unorm
; GFX90A-NEXT: s_endpgm
%cmp = call i32 asm "; def $0", "=a"()
%swap = call i32 asm "; def $0", "=a"()
%unused = call i32 @llvm.amdgcn.image.atomic.cmpswap.1d.i32.i32(i32 %cmp, i32 %swap, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
ret void
}
define amdgpu_ps void @atomic_swap_1d_i64_agpr_noret(<8 x i32> inreg %rsrc, i32 %s) {
; GFX90A-LABEL: atomic_swap_1d_i64_agpr_noret:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def a[0:1]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: image_atomic_swap a[0:1], v0, s[0:7] dmask:0x3 unorm
; GFX90A-NEXT: s_endpgm
%data = call i64 asm "; def $0", "=a"()
%unused = call i64 @llvm.amdgcn.image.atomic.swap.1d.i64.i32(i64 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
ret void
}
define amdgpu_ps void @atomic_cmpswap_1d_64_agpr_noret(<8 x i32> inreg %rsrc, i32 %s) {
; GFX90A-LABEL: atomic_cmpswap_1d_64_agpr_noret:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def a[0:1]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def a[2:3]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: image_atomic_cmpswap a[0:3], v0, s[0:7] dmask:0xf unorm
; GFX90A-NEXT: s_endpgm
%cmp = call i64 asm "; def $0", "=a"()
%swap = call i64 asm "; def $0", "=a"()
%unused = call i64 @llvm.amdgcn.image.atomic.cmpswap.1d.i64.i32(i64 %cmp, i64 %swap, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
ret void
}
|
