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
|
; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; Make sure to test with f32 and i32 compares. If we have to use float
; compares, we always have multiple condition registers. If we can do
; scalar compares, we don't want to use multiple condition registers.
; GCN-LABEL: {{^}}opt_select_i32_and_cmp_i32:
; GCN-DAG: s_cmp_lg_u32
; GCN: s_cselect_b64 [[CMP1:s\[[0-9]+:[0-9]+\]]], -1, 0
; GCN-DAG: s_cmp_lg_u32
; GCN: s_cselect_b64 [[CMP2:s\[[0-9]+:[0-9]+\]]], -1, 0
; GCN: s_and_b64 vcc, [[CMP1]], [[CMP2]]
; GCN: v_cndmask_b32_e32 [[RESULT:v[0-9]+]], {{v[0-9]+}}, {{v[0-9]+}}, vcc
; GCN-NOT: [[RESULT]]
; GCN: buffer_store_dword [[RESULT]]
define amdgpu_kernel void @opt_select_i32_and_cmp_i32(i32 addrspace(1)* %out, i32 %a, i32 %b, i32 %c, i32 %x, i32 %y) #0 {
%icmp0 = icmp ne i32 %a, %b
%icmp1 = icmp ne i32 %a, %c
%and = and i1 %icmp0, %icmp1
%select = select i1 %and, i32 %x, i32 %y
store i32 %select, i32 addrspace(1)* %out
ret void
}
; GCN-LABEL: {{^}}opt_select_i32_and_cmp_f32:
; GCN-DAG: v_cmp_lg_f32_e32 vcc
; GCN-DAG: v_cmp_lg_f32_e64 [[CMP1:s\[[0-9]+:[0-9]+\]]]
; GCN: s_and_b64 vcc, vcc, [[CMP1]]
; GCN: v_cndmask_b32_e32 [[RESULT:v[0-9]+]], {{v[0-9]+}}, {{v[0-9]+}}, vcc
; GCN-NOT: [[RESULT]]
; GCN: buffer_store_dword [[RESULT]]
define amdgpu_kernel void @opt_select_i32_and_cmp_f32(i32 addrspace(1)* %out, float %a, float %b, float %c, i32 %x, i32 %y) #0 {
%fcmp0 = fcmp one float %a, %b
%fcmp1 = fcmp one float %a, %c
%and = and i1 %fcmp0, %fcmp1
%select = select i1 %and, i32 %x, i32 %y
store i32 %select, i32 addrspace(1)* %out
ret void
}
; GCN-LABEL: {{^}}opt_select_i64_and_cmp_i32:
; GCN-DAG: s_cmp_lg_u32
; GCN: s_cselect_b64 [[CMP1:s\[[0-9]+:[0-9]+\]]], -1, 0
; GCN-DAG: s_cmp_lg_u32
; GCN: s_cselect_b64 [[CMP2:s\[[0-9]+:[0-9]+\]]], -1, 0
; GCN: s_and_b64 vcc, [[CMP1]], [[CMP2]]
; GCN: v_cndmask_b32_e32 v[[RESULT1:[0-9]+]], {{v[0-9]+}}, {{v[0-9]+}}, vcc
; GCN: v_cndmask_b32_e32 v[[RESULT0:[0-9]+]], {{v[0-9]+}}, {{v[0-9]+}}, vcc
; GCN: buffer_store_dwordx2 v{{\[}}[[RESULT0]]:[[RESULT1]]{{\]}}
define amdgpu_kernel void @opt_select_i64_and_cmp_i32(i64 addrspace(1)* %out, i32 %a, i32 %b, i32 %c, i64 %x, i64 %y) #0 {
%icmp0 = icmp ne i32 %a, %b
%icmp1 = icmp ne i32 %a, %c
%and = and i1 %icmp0, %icmp1
%select = select i1 %and, i64 %x, i64 %y
store i64 %select, i64 addrspace(1)* %out
ret void
}
; GCN-LABEL: {{^}}opt_select_i64_and_cmp_f32:
; GCN-DAG: v_cmp_lg_f32_e32 vcc,
; GCN-DAG: v_cmp_lg_f32_e64 [[CMP1:s\[[0-9]+:[0-9]+\]]]
; GCN: s_and_b64 vcc, vcc, [[CMP1]]
; GCN: v_cndmask_b32_e32 v[[RESULT1:[0-9]+]], {{v[0-9]+}}, {{v[0-9]+}}, vcc
; GCN: v_cndmask_b32_e32 v[[RESULT0:[0-9]+]], {{v[0-9]+}}, {{v[0-9]+}}, vcc
; GCN: buffer_store_dwordx2 v{{\[}}[[RESULT0]]:[[RESULT1]]{{\]}}
define amdgpu_kernel void @opt_select_i64_and_cmp_f32(i64 addrspace(1)* %out, float %a, float %b, float %c, i64 %x, i64 %y) #0 {
%fcmp0 = fcmp one float %a, %b
%fcmp1 = fcmp one float %a, %c
%and = and i1 %fcmp0, %fcmp1
%select = select i1 %and, i64 %x, i64 %y
store i64 %select, i64 addrspace(1)* %out
ret void
}
; GCN-LABEL: {{^}}opt_select_i32_or_cmp_i32:
; GCN-DAG: s_cmp_lg_u32
; GCN: s_cselect_b64 [[CMP1:s\[[0-9]+:[0-9]+\]]], -1, 0
; GCN-DAG: s_cmp_lg_u32
; GCN: s_cselect_b64 [[CMP2:s\[[0-9]+:[0-9]+\]]], -1, 0
; GCN: s_or_b64 vcc, [[CMP1]], [[CMP2]]
; GCN: v_cndmask_b32_e32 [[RESULT:v[0-9]+]], {{v[0-9]+}}, {{v[0-9]+}}, vcc
; GCN-NOT: [[RESULT]]
; GCN: buffer_store_dword [[RESULT]]
; GCN: s_endpgm
define amdgpu_kernel void @opt_select_i32_or_cmp_i32(i32 addrspace(1)* %out, i32 %a, i32 %b, i32 %c, i32 %x, i32 %y) #0 {
%icmp0 = icmp ne i32 %a, %b
%icmp1 = icmp ne i32 %a, %c
%or = or i1 %icmp0, %icmp1
%select = select i1 %or, i32 %x, i32 %y
store i32 %select, i32 addrspace(1)* %out
ret void
}
; GCN-LABEL: {{^}}opt_select_i32_or_cmp_f32:
; GCN-DAG: v_cmp_lg_f32_e32 vcc
; GCN-DAG: v_cmp_lg_f32_e64 [[CMP1:s\[[0-9]+:[0-9]+\]]]
; GCN: s_or_b64 vcc, vcc, [[CMP1]]
; GCN: v_cndmask_b32_e32 [[RESULT:v[0-9]+]], {{v[0-9]+}}, {{v[0-9]+}}, vcc
; GCN-NOT: [[RESULT]]
; GCN: buffer_store_dword [[RESULT]]
define amdgpu_kernel void @opt_select_i32_or_cmp_f32(i32 addrspace(1)* %out, float %a, float %b, float %c, i32 %x, i32 %y) #0 {
%fcmp0 = fcmp one float %a, %b
%fcmp1 = fcmp one float %a, %c
%or = or i1 %fcmp0, %fcmp1
%select = select i1 %or, i32 %x, i32 %y
store i32 %select, i32 addrspace(1)* %out
ret void
}
; GCN-LABEL: {{^}}opt_select_i64_or_cmp_i32:
; GCN-DAG: s_cmp_lg_u32
; GCN: s_cselect_b64 [[CMP1:s\[[0-9]+:[0-9]+\]]], -1, 0
; GCN-DAG: s_cmp_lg_u32
; GCN: s_cselect_b64 [[CMP2:s\[[0-9]+:[0-9]+\]]], -1, 0
; GCN: s_or_b64 vcc, [[CMP1]], [[CMP2]]
; GCN: v_cndmask_b32_e32 v[[RESULT1:[0-9]+]], {{v[0-9]+}}, {{v[0-9]+}}, vcc
; GCN: v_cndmask_b32_e32 v[[RESULT0:[0-9]+]], {{v[0-9]+}}, {{v[0-9]+}}, vcc
; GCN: buffer_store_dwordx2 v{{\[}}[[RESULT0]]:[[RESULT1]]{{\]}}
define amdgpu_kernel void @opt_select_i64_or_cmp_i32(i64 addrspace(1)* %out, i32 %a, i32 %b, i32 %c, i64 %x, i64 %y) #0 {
%icmp0 = icmp ne i32 %a, %b
%icmp1 = icmp ne i32 %a, %c
%or = or i1 %icmp0, %icmp1
%select = select i1 %or, i64 %x, i64 %y
store i64 %select, i64 addrspace(1)* %out
ret void
}
; GCN-LABEL: {{^}}opt_select_i64_or_cmp_f32:
; GCN-DAG: v_cmp_lg_f32_e32 vcc,
; GCN-DAG: v_cmp_lg_f32_e64 [[CMP1:s\[[0-9]+:[0-9]+\]]]
; GCN: s_or_b64 vcc, vcc, [[CMP1]]
; GCN: v_cndmask_b32_e32 v[[RESULT1:[0-9]+]], {{v[0-9]+}}, {{v[0-9]+}}, vcc
; GCN: v_cndmask_b32_e32 v[[RESULT0:[0-9]+]], {{v[0-9]+}}, {{v[0-9]+}}, vcc
; GCN: buffer_store_dwordx2 v{{\[}}[[RESULT0]]:[[RESULT1]]{{\]}}
define amdgpu_kernel void @opt_select_i64_or_cmp_f32(i64 addrspace(1)* %out, float %a, float %b, float %c, i64 %x, i64 %y) #0 {
%fcmp0 = fcmp one float %a, %b
%fcmp1 = fcmp one float %a, %c
%or = or i1 %fcmp0, %fcmp1
%select = select i1 %or, i64 %x, i64 %y
store i64 %select, i64 addrspace(1)* %out
ret void
}
; GCN-LABEL: {{^}}regression:
; GCN: v_cmp_neq_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}}, 1.0
; GCN: v_cmp_neq_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}}, 0
; GCN: v_cmp_eq_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}}, 0
define amdgpu_kernel void @regression(float addrspace(1)* %out, float %c0, float %c1) #0 {
entry:
%cmp0 = fcmp oeq float %c0, 1.0
br i1 %cmp0, label %if0, label %endif
if0:
%cmp1 = fcmp oeq float %c1, 0.0
br i1 %cmp1, label %if1, label %endif
if1:
%cmp2 = xor i1 %cmp1, true
br label %endif
endif:
%tmp0 = phi i1 [ true, %entry ], [ %cmp2, %if1 ], [ false, %if0 ]
%tmp2 = select i1 %tmp0, float 4.0, float 0.0
store float %tmp2, float addrspace(1)* %out
ret void
}
attributes #0 = { nounwind }
|
