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
170
171
172
173
174
175
176
177
178
179
180
181
|
; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
; XUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare i32 @llvm.r600.read.tidig.x() nounwind readnone
declare float @llvm.fabs.f32(float) nounwind readnone
; GCN-LABEL: {{^}}madmk_f32:
; GCN-DAG: buffer_load_dword [[VA:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
; GCN-DAG: buffer_load_dword [[VB:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
; GCN: v_madmk_f32 {{v[0-9]+}}, [[VA]], [[VB]], 0x41200000
define void @madmk_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) nounwind {
%tid = tail call i32 @llvm.r600.read.tidig.x() nounwind readnone
%gep.0 = getelementptr float addrspace(1)* %in, i32 %tid
%gep.1 = getelementptr float addrspace(1)* %gep.0, i32 1
%out.gep = getelementptr float addrspace(1)* %out, i32 %tid
%a = load float addrspace(1)* %gep.0, align 4
%b = load float addrspace(1)* %gep.1, align 4
%mul = fmul float %a, 10.0
%madmk = fadd float %mul, %b
store float %madmk, float addrspace(1)* %out.gep, align 4
ret void
}
; GCN-LABEL: {{^}}madmk_2_use_f32:
; GCN-DAG: buffer_load_dword [[VA:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
; GCN-DAG: buffer_load_dword [[VB:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
; GCN-DAG: buffer_load_dword [[VC:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:8
; GCN-DAG: v_mov_b32_e32 [[VK:v[0-9]+]], 0x41200000
; GCN-DAG: v_mad_f32 {{v[0-9]+}}, [[VA]], [[VK]], [[VB]]
; GCN-DAG: v_mad_f32 {{v[0-9]+}}, [[VA]], [[VK]], [[VC]]
; GCN: s_endpgm
define void @madmk_2_use_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) nounwind {
%tid = tail call i32 @llvm.r600.read.tidig.x() nounwind readnone
%in.gep.0 = getelementptr float addrspace(1)* %in, i32 %tid
%in.gep.1 = getelementptr float addrspace(1)* %in.gep.0, i32 1
%in.gep.2 = getelementptr float addrspace(1)* %in.gep.0, i32 2
%out.gep.0 = getelementptr float addrspace(1)* %out, i32 %tid
%out.gep.1 = getelementptr float addrspace(1)* %in.gep.0, i32 1
%a = load float addrspace(1)* %in.gep.0, align 4
%b = load float addrspace(1)* %in.gep.1, align 4
%c = load float addrspace(1)* %in.gep.2, align 4
%mul0 = fmul float %a, 10.0
%mul1 = fmul float %a, 10.0
%madmk0 = fadd float %mul0, %b
%madmk1 = fadd float %mul1, %c
store float %madmk0, float addrspace(1)* %out.gep.0, align 4
store float %madmk1, float addrspace(1)* %out.gep.1, align 4
ret void
}
; We don't get any benefit if the constant is an inline immediate.
; GCN-LABEL: {{^}}madmk_inline_imm_f32:
; GCN-DAG: buffer_load_dword [[VA:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
; GCN-DAG: buffer_load_dword [[VB:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
; GCN: v_mad_f32 {{v[0-9]+}}, 4.0, [[VA]], [[VB]]
define void @madmk_inline_imm_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) nounwind {
%tid = tail call i32 @llvm.r600.read.tidig.x() nounwind readnone
%gep.0 = getelementptr float addrspace(1)* %in, i32 %tid
%gep.1 = getelementptr float addrspace(1)* %gep.0, i32 1
%out.gep = getelementptr float addrspace(1)* %out, i32 %tid
%a = load float addrspace(1)* %gep.0, align 4
%b = load float addrspace(1)* %gep.1, align 4
%mul = fmul float %a, 4.0
%madmk = fadd float %mul, %b
store float %madmk, float addrspace(1)* %out.gep, align 4
ret void
}
; GCN-LABEL: {{^}}s_s_madmk_f32:
; GCN-NOT: v_madmk_f32
; GCN: v_mad_f32
; GCN: s_endpgm
define void @s_s_madmk_f32(float addrspace(1)* noalias %out, float %a, float %b) nounwind {
%tid = tail call i32 @llvm.r600.read.tidig.x() nounwind readnone
%out.gep = getelementptr float addrspace(1)* %out, i32 %tid
%mul = fmul float %a, 10.0
%madmk = fadd float %mul, %b
store float %madmk, float addrspace(1)* %out.gep, align 4
ret void
}
; GCN-LABEL: {{^}}v_s_madmk_f32:
; GCN-NOT: v_madmk_f32
; GCN: v_mad_f32
; GCN: s_endpgm
define void @v_s_madmk_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in, float %b) nounwind {
%tid = tail call i32 @llvm.r600.read.tidig.x() nounwind readnone
%gep.0 = getelementptr float addrspace(1)* %in, i32 %tid
%out.gep = getelementptr float addrspace(1)* %out, i32 %tid
%a = load float addrspace(1)* %gep.0, align 4
%mul = fmul float %a, 10.0
%madmk = fadd float %mul, %b
store float %madmk, float addrspace(1)* %out.gep, align 4
ret void
}
; GCN-LABEL: {{^}}scalar_vector_madmk_f32:
; GCN-NOT: v_madmk_f32
; GCN: v_mad_f32
; GCN: s_endpgm
define void @scalar_vector_madmk_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in, float %a) nounwind {
%tid = tail call i32 @llvm.r600.read.tidig.x() nounwind readnone
%gep.0 = getelementptr float addrspace(1)* %in, i32 %tid
%out.gep = getelementptr float addrspace(1)* %out, i32 %tid
%b = load float addrspace(1)* %gep.0, align 4
%mul = fmul float %a, 10.0
%madmk = fadd float %mul, %b
store float %madmk, float addrspace(1)* %out.gep, align 4
ret void
}
; GCN-LABEL: {{^}}no_madmk_src0_modifier_f32:
; GCN-DAG: buffer_load_dword [[VA:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
; GCN-DAG: buffer_load_dword [[VB:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
; GCN: v_mad_f32 {{v[0-9]+}}, |{{v[0-9]+}}|, {{v[0-9]+}}, {{[sv][0-9]+}}
define void @no_madmk_src0_modifier_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) nounwind {
%tid = tail call i32 @llvm.r600.read.tidig.x() nounwind readnone
%gep.0 = getelementptr float addrspace(1)* %in, i32 %tid
%gep.1 = getelementptr float addrspace(1)* %gep.0, i32 1
%out.gep = getelementptr float addrspace(1)* %out, i32 %tid
%a = load float addrspace(1)* %gep.0, align 4
%b = load float addrspace(1)* %gep.1, align 4
%a.fabs = call float @llvm.fabs.f32(float %a) nounwind readnone
%mul = fmul float %a.fabs, 10.0
%madmk = fadd float %mul, %b
store float %madmk, float addrspace(1)* %out.gep, align 4
ret void
}
; GCN-LABEL: {{^}}no_madmk_src2_modifier_f32:
; GCN-DAG: buffer_load_dword [[VA:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
; GCN-DAG: buffer_load_dword [[VB:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
; GCN: v_mad_f32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}}, |{{[sv][0-9]+}}|
define void @no_madmk_src2_modifier_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) nounwind {
%tid = tail call i32 @llvm.r600.read.tidig.x() nounwind readnone
%gep.0 = getelementptr float addrspace(1)* %in, i32 %tid
%gep.1 = getelementptr float addrspace(1)* %gep.0, i32 1
%out.gep = getelementptr float addrspace(1)* %out, i32 %tid
%a = load float addrspace(1)* %gep.0, align 4
%b = load float addrspace(1)* %gep.1, align 4
%b.fabs = call float @llvm.fabs.f32(float %b) nounwind readnone
%mul = fmul float %a, 10.0
%madmk = fadd float %mul, %b.fabs
store float %madmk, float addrspace(1)* %out.gep, align 4
ret void
}
; GCN-LABEL: {{^}}madmk_add_inline_imm_f32:
; GCN: buffer_load_dword [[A:v[0-9]+]]
; GCN: v_mov_b32_e32 [[VK:v[0-9]+]], 0x41200000
; GCN: v_mad_f32 {{v[0-9]+}}, [[VK]], [[A]], 2.0
define void @madmk_add_inline_imm_f32(float addrspace(1)* noalias %out, float addrspace(1)* noalias %in) nounwind {
%tid = tail call i32 @llvm.r600.read.tidig.x() nounwind readnone
%gep.0 = getelementptr float addrspace(1)* %in, i32 %tid
%out.gep = getelementptr float addrspace(1)* %out, i32 %tid
%a = load float addrspace(1)* %gep.0, align 4
%mul = fmul float %a, 10.0
%madmk = fadd float %mul, 2.0
store float %madmk, float addrspace(1)* %out.gep, align 4
ret void
}
|
