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
|
; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
declare float @llvm.fabs.f32(float) #1
declare double @llvm.fabs.f64(double) #1
; SI-LABEL: {{^}}test_isinf_pattern:
; SI: v_mov_b32_e32 [[MASK:v[0-9]+]], 0x204{{$}}
; SI: v_cmp_class_f32_e32 vcc, s{{[0-9]+}}, [[MASK]]
; SI-NOT: v_cmp
; SI: s_endpgm
define amdgpu_kernel void @test_isinf_pattern(i32 addrspace(1)* nocapture %out, float %x) #0 {
%fabs = tail call float @llvm.fabs.f32(float %x) #1
%cmp = fcmp oeq float %fabs, 0x7FF0000000000000
%ext = zext i1 %cmp to i32
store i32 %ext, i32 addrspace(1)* %out, align 4
ret void
}
; SI-LABEL: {{^}}test_not_isinf_pattern_0:
; SI-NOT: v_cmp_class
; SI: s_endpgm
define amdgpu_kernel void @test_not_isinf_pattern_0(i32 addrspace(1)* nocapture %out, float %x) #0 {
%fabs = tail call float @llvm.fabs.f32(float %x) #1
%cmp = fcmp ueq float %fabs, 0x7FF0000000000000
%ext = zext i1 %cmp to i32
store i32 %ext, i32 addrspace(1)* %out, align 4
ret void
}
; SI-LABEL: {{^}}test_not_isinf_pattern_1:
; SI-NOT: v_cmp_class
; SI: s_endpgm
define amdgpu_kernel void @test_not_isinf_pattern_1(i32 addrspace(1)* nocapture %out, float %x) #0 {
%fabs = tail call float @llvm.fabs.f32(float %x) #1
%cmp = fcmp oeq float %fabs, 0xFFF0000000000000
%ext = zext i1 %cmp to i32
store i32 %ext, i32 addrspace(1)* %out, align 4
ret void
}
; SI-LABEL: {{^}}test_isfinite_pattern_0:
; SI-NOT: v_cmp
; SI: v_mov_b32_e32 [[MASK:v[0-9]+]], 0x1f8{{$}}
; SI: v_cmp_class_f32_e32 vcc, s{{[0-9]+}}, [[MASK]]
; SI-NOT: v_cmp
; SI: s_endpgm
define amdgpu_kernel void @test_isfinite_pattern_0(i32 addrspace(1)* nocapture %out, float %x) #0 {
%ord = fcmp ord float %x, 0.000000e+00
%x.fabs = tail call float @llvm.fabs.f32(float %x) #1
%ninf = fcmp une float %x.fabs, 0x7FF0000000000000
%and = and i1 %ord, %ninf
%ext = zext i1 %and to i32
store i32 %ext, i32 addrspace(1)* %out, align 4
ret void
}
; Use negative infinity
; SI-LABEL: {{^}}test_isfinite_not_pattern_0:
; SI-NOT: v_cmp_class_f32
; SI: s_endpgm
define amdgpu_kernel void @test_isfinite_not_pattern_0(i32 addrspace(1)* nocapture %out, float %x) #0 {
%ord = fcmp ord float %x, 0.000000e+00
%x.fabs = tail call float @llvm.fabs.f32(float %x) #1
%ninf = fcmp une float %x.fabs, 0xFFF0000000000000
%and = and i1 %ord, %ninf
%ext = zext i1 %and to i32
store i32 %ext, i32 addrspace(1)* %out, align 4
ret void
}
; No fabs
; SI-LABEL: {{^}}test_isfinite_not_pattern_1:
; SI-NOT: v_cmp_class_f32
; SI: s_endpgm
define amdgpu_kernel void @test_isfinite_not_pattern_1(i32 addrspace(1)* nocapture %out, float %x) #0 {
%ord = fcmp ord float %x, 0.000000e+00
%ninf = fcmp une float %x, 0x7FF0000000000000
%and = and i1 %ord, %ninf
%ext = zext i1 %and to i32
store i32 %ext, i32 addrspace(1)* %out, align 4
ret void
}
; fabs of different value
; SI-LABEL: {{^}}test_isfinite_not_pattern_2:
; SI-NOT: v_cmp_class_f32
; SI: s_endpgm
define amdgpu_kernel void @test_isfinite_not_pattern_2(i32 addrspace(1)* nocapture %out, float %x, float %y) #0 {
%ord = fcmp ord float %x, 0.000000e+00
%x.fabs = tail call float @llvm.fabs.f32(float %y) #1
%ninf = fcmp une float %x.fabs, 0x7FF0000000000000
%and = and i1 %ord, %ninf
%ext = zext i1 %and to i32
store i32 %ext, i32 addrspace(1)* %out, align 4
ret void
}
; Wrong ordered compare type
; SI-LABEL: {{^}}test_isfinite_not_pattern_3:
; SI-NOT: v_cmp_class_f32
; SI: s_endpgm
define amdgpu_kernel void @test_isfinite_not_pattern_3(i32 addrspace(1)* nocapture %out, float %x) #0 {
%ord = fcmp uno float %x, 0.000000e+00
%x.fabs = tail call float @llvm.fabs.f32(float %x) #1
%ninf = fcmp une float %x.fabs, 0x7FF0000000000000
%and = and i1 %ord, %ninf
%ext = zext i1 %and to i32
store i32 %ext, i32 addrspace(1)* %out, align 4
ret void
}
; Wrong unordered compare
; SI-LABEL: {{^}}test_isfinite_not_pattern_4:
; SI-NOT: v_cmp_class_f32
; SI: s_endpgm
define amdgpu_kernel void @test_isfinite_not_pattern_4(i32 addrspace(1)* nocapture %out, float %x) #0 {
%ord = fcmp ord float %x, 0.000000e+00
%x.fabs = tail call float @llvm.fabs.f32(float %x) #1
%ninf = fcmp one float %x.fabs, 0x7FF0000000000000
%and = and i1 %ord, %ninf
%ext = zext i1 %and to i32
store i32 %ext, i32 addrspace(1)* %out, align 4
ret void
}
attributes #0 = { nounwind }
attributes #1 = { nounwind readnone }
|
