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
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
| ; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefix=GCN %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefixes=GCN,VI %s
; RUN: llc -march=amdgcn -mcpu=kabini -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefixes=GCN,16BANK %s
; RUN: llc -march=amdgcn -mcpu=stoney -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefixes=GCN,16BANK %s
; GCN-LABEL: {{^}}v_interp:
; GCN-NOT: s_wqm
; GCN: s_mov_b32 m0, s{{[0-9]+}}
; GCN-DAG: v_interp_p1_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr0.x{{$}}
; GCN-DAG: v_interp_p1_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr0.y{{$}}
; GCN-DAG: v_interp_p2_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr0.y{{$}}
; GCN-DAG: v_interp_mov_f32{{(_e32)*}} v{{[0-9]+}}, p0, attr0.x{{$}}
define amdgpu_ps void @v_interp(<16 x i8> addrspace(4)* inreg %arg, <16 x i8> addrspace(4)* inreg %arg1, <32 x i8> addrspace(4)* inreg %arg2, i32 inreg %arg3, <2 x float> %arg4) #0 {
main_body:
%i = extractelement <2 x float> %arg4, i32 0
%j = extractelement <2 x float> %arg4, i32 1
%p0_0 = call float @llvm.amdgcn.interp.p1(float %i, i32 0, i32 0, i32 %arg3)
%p1_0 = call float @llvm.amdgcn.interp.p2(float %p0_0, float %j, i32 0, i32 0, i32 %arg3)
%p0_1 = call float @llvm.amdgcn.interp.p1(float %i, i32 1, i32 0, i32 %arg3)
%p1_1 = call float @llvm.amdgcn.interp.p2(float %p0_1, float %j, i32 1, i32 0, i32 %arg3)
%const = call float @llvm.amdgcn.interp.mov(i32 2, i32 0, i32 0, i32 %arg3)
%w = fadd float %p1_1, %const
call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %p0_0, float %p0_0, float %p1_1, float %w, i1 true, i1 true) #0
ret void
}
; GCN-LABEL: {{^}}v_interp_p1:
; GCN: s_movk_i32 m0, 0x100
; GCN-DAG: v_interp_p1_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr0.x{{$}}
; GCN-DAG: v_interp_p1_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr0.y{{$}}
; GCN-DAG: v_interp_p1_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr0.z{{$}}
; GCN-DAG: v_interp_p1_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr0.w{{$}}
; GCN-DAG: v_interp_p1_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr0.x{{$}}
; GCN-DAG: v_interp_p1_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr1.x{{$}}
; GCN-DAG: v_interp_p1_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr2.y{{$}}
; GCN-DAG: v_interp_p1_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr3.z{{$}}
; GCN-DAG: v_interp_p1_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr4.w{{$}}
; GCN-DAG: v_interp_p1_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr63.w{{$}}
; GCN-DAG: v_interp_p1_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr64.w{{$}}
; GCN-DAG: v_interp_p1_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr64.x{{$}}
define amdgpu_ps void @v_interp_p1(float %i) #0 {
bb:
%p0_0 = call float @llvm.amdgcn.interp.p1(float %i, i32 0, i32 0, i32 256)
%p0_1 = call float @llvm.amdgcn.interp.p1(float %i, i32 1, i32 0, i32 256)
%p0_2 = call float @llvm.amdgcn.interp.p1(float %i, i32 2, i32 0, i32 256)
%p0_3 = call float @llvm.amdgcn.interp.p1(float %i, i32 3, i32 0, i32 256)
%p0_4 = call float @llvm.amdgcn.interp.p1(float %i, i32 4, i32 0, i32 256)
%p0_5 = call float @llvm.amdgcn.interp.p1(float %i, i32 0, i32 1, i32 256)
%p0_6 = call float @llvm.amdgcn.interp.p1(float %i, i32 1, i32 2, i32 256)
%p0_7 = call float @llvm.amdgcn.interp.p1(float %i, i32 2, i32 3, i32 256)
%p0_8 = call float @llvm.amdgcn.interp.p1(float %i, i32 3, i32 4, i32 256)
%p0_9 = call float @llvm.amdgcn.interp.p1(float %i, i32 3, i32 63, i32 256)
%p0_10 = call float @llvm.amdgcn.interp.p1(float %i, i32 3, i32 64, i32 256)
%p0_11 = call float @llvm.amdgcn.interp.p1(float %i, i32 4, i32 64, i32 256)
store volatile float %p0_0, float addrspace(1)* undef
store volatile float %p0_1, float addrspace(1)* undef
store volatile float %p0_2, float addrspace(1)* undef
store volatile float %p0_3, float addrspace(1)* undef
store volatile float %p0_4, float addrspace(1)* undef
store volatile float %p0_5, float addrspace(1)* undef
store volatile float %p0_6, float addrspace(1)* undef
store volatile float %p0_7, float addrspace(1)* undef
store volatile float %p0_8, float addrspace(1)* undef
store volatile float %p0_9, float addrspace(1)* undef
store volatile float %p0_10, float addrspace(1)* undef
store volatile float %p0_11, float addrspace(1)* undef
ret void
}
; GCN-LABEL: {{^}}v_interp_p2:
; GCN: s_movk_i32 m0, 0x100
; GCN-DAG: v_interp_p2_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr0.x{{$}}
; GCN-DAG: v_interp_p2_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr0.y{{$}}
; GCN-DAG: v_interp_p2_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr0.z{{$}}
; GCN-DAG: v_interp_p2_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr0.w{{$}}
; GCN-DAG: v_interp_p2_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr0.x{{$}}
; GCN-DAG: v_interp_p2_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr0.x{{$}}
; GCN-DAG: v_interp_p2_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr63.x{{$}}
; GCN-DAG: v_interp_p2_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr64.x{{$}}
; GCN-DAG: v_interp_p2_f32{{(_e32)*}} v{{[0-9]+}}, v{{[0-9]+}}, attr64.x{{$}}
define amdgpu_ps void @v_interp_p2(float %x, float %j) #0 {
bb:
%p2_0 = call float @llvm.amdgcn.interp.p2(float %x, float %j, i32 0, i32 0, i32 256)
%p2_1 = call float @llvm.amdgcn.interp.p2(float %x, float %j, i32 1, i32 0, i32 256)
%p2_2 = call float @llvm.amdgcn.interp.p2(float %x, float %j, i32 2, i32 0, i32 256)
%p2_3 = call float @llvm.amdgcn.interp.p2(float %x, float %j, i32 3, i32 0, i32 256)
%p2_4 = call float @llvm.amdgcn.interp.p2(float %x, float %j, i32 4, i32 0, i32 256)
%p2_5 = call float @llvm.amdgcn.interp.p2(float %x, float %j, i32 0, i32 1, i32 256)
%p2_6 = call float @llvm.amdgcn.interp.p2(float %x, float %j, i32 0, i32 63, i32 256)
%p2_7 = call float @llvm.amdgcn.interp.p2(float %x, float %j, i32 0, i32 64, i32 256)
%p2_8 = call float @llvm.amdgcn.interp.p2(float %x, float %j, i32 4, i32 64, i32 256)
store volatile float %p2_0, float addrspace(1)* undef
store volatile float %p2_1, float addrspace(1)* undef
store volatile float %p2_2, float addrspace(1)* undef
store volatile float %p2_3, float addrspace(1)* undef
store volatile float %p2_4, float addrspace(1)* undef
store volatile float %p2_5, float addrspace(1)* undef
store volatile float %p2_6, float addrspace(1)* undef
store volatile float %p2_7, float addrspace(1)* undef
store volatile float %p2_8, float addrspace(1)* undef
ret void
}
; GCN-LABEL: {{^}}v_interp_mov:
; GCN: s_movk_i32 m0, 0x100
; GCN-DAG: v_interp_mov_f32{{(_e32)*}} v{{[0-9]+}}, p10, attr0.x{{$}}
; GCN-DAG: v_interp_mov_f32{{(_e32)*}} v{{[0-9]+}}, p20, attr0.x{{$}}
; GCN-DAG: v_interp_mov_f32{{(_e32)*}} v{{[0-9]+}}, p0, attr0.x{{$}}
; GCN-DAG: v_interp_mov_f32{{(_e32)*}} v{{[0-9]+}}, invalid_param_3, attr0.x{{$}}
; GCN-DAG: v_interp_mov_f32{{(_e32)*}} v{{[0-9]+}}, p10, attr0.x{{$}}
; GCN-DAG: v_interp_mov_f32{{(_e32)*}} v{{[0-9]+}}, p10, attr0.z{{$}}
; GCN-DAG: v_interp_mov_f32{{(_e32)*}} v{{[0-9]+}}, p10, attr0.w{{$}}
; GCN-DAG: v_interp_mov_f32{{(_e32)*}} v{{[0-9]+}}, p10, attr0.x{{$}}
; GCN-DAG: v_interp_mov_f32{{(_e32)*}} v{{[0-9]+}}, invalid_param_8, attr0.x{{$}}
; GCN-DAG: v_interp_mov_f32{{(_e32)*}} v{{[0-9]+}}, p10, attr63.y{{$}}
; GCN-DAG: v_interp_mov_f32{{(_e32)*}} v{{[0-9]+}}, p10, attr64.y{{$}}
; GCN-DAG: v_interp_mov_f32{{(_e32)*}} v{{[0-9]+}}, invalid_param_3, attr64.y{{$}}
; GCN-DAG: v_interp_mov_f32{{(_e32)*}} v{{[0-9]+}}, invalid_param_10, attr64.x{{$}}
define amdgpu_ps void @v_interp_mov(float %x, float %j) #0 {
bb:
%mov_0 = call float @llvm.amdgcn.interp.mov(i32 0, i32 0, i32 0, i32 256)
%mov_1 = call float @llvm.amdgcn.interp.mov(i32 1, i32 0, i32 0, i32 256)
%mov_2 = call float @llvm.amdgcn.interp.mov(i32 2, i32 0, i32 0, i32 256)
%mov_3 = call float @llvm.amdgcn.interp.mov(i32 3, i32 0, i32 0, i32 256)
%mov_4 = call float @llvm.amdgcn.interp.mov(i32 0, i32 1, i32 0, i32 256)
%mov_5 = call float @llvm.amdgcn.interp.mov(i32 0, i32 2, i32 0, i32 256)
%mov_6 = call float @llvm.amdgcn.interp.mov(i32 0, i32 3, i32 0, i32 256)
%mov_7 = call float @llvm.amdgcn.interp.mov(i32 0, i32 4, i32 0, i32 256)
%mov_8 = call float @llvm.amdgcn.interp.mov(i32 8, i32 4, i32 0, i32 256)
%mov_9 = call float @llvm.amdgcn.interp.mov(i32 0, i32 1, i32 63, i32 256)
%mov_10 = call float @llvm.amdgcn.interp.mov(i32 0, i32 1, i32 64, i32 256)
%mov_11 = call float @llvm.amdgcn.interp.mov(i32 3, i32 1, i32 64, i32 256)
%mov_12 = call float @llvm.amdgcn.interp.mov(i32 10, i32 4, i32 64, i32 256)
store volatile float %mov_0, float addrspace(1)* undef
store volatile float %mov_1, float addrspace(1)* undef
store volatile float %mov_2, float addrspace(1)* undef
store volatile float %mov_3, float addrspace(1)* undef
store volatile float %mov_4, float addrspace(1)* undef
store volatile float %mov_5, float addrspace(1)* undef
store volatile float %mov_6, float addrspace(1)* undef
store volatile float %mov_7, float addrspace(1)* undef
store volatile float %mov_8, float addrspace(1)* undef
store volatile float %mov_9, float addrspace(1)* undef
store volatile float %mov_10, float addrspace(1)* undef
store volatile float %mov_11, float addrspace(1)* undef
store volatile float %mov_12, float addrspace(1)* undef
ret void
}
; SI won't merge ds memory operations, because of the signed offset bug, so
; we only have check lines for VI.
;
; TODO: VI won't merge them either, because we are conservative about moving
; instructions past changes to physregs.
;
; TODO-VI-LABEL: v_interp_readnone:
; TODO-VI: s_mov_b32 m0, 0
; TODO-VI-DAG: v_mov_b32_e32 [[ZERO:v[0-9]+]], 0
; TODO-VI-DAG: v_interp_mov_f32_e32 v{{[0-9]+}}, p0, attr0.x{{$}}
; TODO-VI: s_mov_b32 m0, -1{{$}}
; TODO-VI: ds_write2_b32 v{{[0-9]+}}, [[ZERO]], [[ZERO]] offset1:4
;define amdgpu_ps void @v_interp_readnone(float addrspace(3)* %lds) #0 {
;bb:
; store float 0.000000e+00, float addrspace(3)* %lds
; %tmp1 = call float @llvm.amdgcn.interp.mov(i32 2, i32 0, i32 0, i32 0)
; %tmp2 = getelementptr float, float addrspace(3)* %lds, i32 4
; store float 0.000000e+00, float addrspace(3)* %tmp2
; call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %tmp1, float %tmp1, float %tmp1, float %tmp1, i1 true, i1 true) #0
; ret void
;}
; Thest that v_interp_p1 uses different source and destination registers
; on 16 bank LDS chips.
; GCN-LABEL: {{^}}v_interp_p1_bank16_bug:
; 16BANK-NOT: v_interp_p1_f32{{(_e32)*}} [[DST:v[0-9]+]], [[DST]]
define amdgpu_ps void @v_interp_p1_bank16_bug([6 x <16 x i8>] addrspace(4)* inreg %arg, [17 x <16 x i8>] addrspace(4)* inreg %arg13, [17 x <4 x i32>] addrspace(4)* inreg %arg14, [34 x <8 x i32>] addrspace(4)* inreg %arg15, float inreg %arg16, i32 inreg %arg17, <2 x i32> %arg18, <2 x i32> %arg19, <2 x i32> %arg20, <3 x i32> %arg21, <2 x i32> %arg22, <2 x i32> %arg23, <2 x i32> %arg24, float %arg25, float %arg26, float %arg27, float %arg28, float %arg29, float %arg30, i32 %arg31, float %arg32, float %arg33) #0 {
main_body:
%i.i = extractelement <2 x i32> %arg19, i32 0
%j.i = extractelement <2 x i32> %arg19, i32 1
%i.f.i = bitcast i32 %i.i to float
%j.f.i = bitcast i32 %j.i to float
%p1.i = call float @llvm.amdgcn.interp.p1(float %i.f.i, i32 0, i32 0, i32 %arg17) #0
%p2.i = call float @llvm.amdgcn.interp.p2(float %p1.i, float %j.f.i, i32 0, i32 0, i32 %arg17) #0
%i.i7 = extractelement <2 x i32> %arg19, i32 0
%j.i8 = extractelement <2 x i32> %arg19, i32 1
%i.f.i9 = bitcast i32 %i.i7 to float
%j.f.i10 = bitcast i32 %j.i8 to float
%p1.i11 = call float @llvm.amdgcn.interp.p1(float %i.f.i9, i32 1, i32 0, i32 %arg17) #0
%p2.i12 = call float @llvm.amdgcn.interp.p2(float %p1.i11, float %j.f.i10, i32 1, i32 0, i32 %arg17) #0
%i.i1 = extractelement <2 x i32> %arg19, i32 0
%j.i2 = extractelement <2 x i32> %arg19, i32 1
%i.f.i3 = bitcast i32 %i.i1 to float
%j.f.i4 = bitcast i32 %j.i2 to float
%p1.i5 = call float @llvm.amdgcn.interp.p1(float %i.f.i3, i32 2, i32 0, i32 %arg17) #0
%p2.i6 = call float @llvm.amdgcn.interp.p2(float %p1.i5, float %j.f.i4, i32 2, i32 0, i32 %arg17) #0
%tmp = call float @llvm.fabs.f32(float %p2.i)
%tmp34 = call float @llvm.fabs.f32(float %p2.i12)
%tmp35 = call float @llvm.fabs.f32(float %p2.i6)
%tmp36 = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float %tmp, float %tmp34)
%tmp38 = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float %tmp35, float 1.000000e+00)
call void @llvm.amdgcn.exp.compr.v2f16(i32 0, i32 15, <2 x half> %tmp36, <2 x half> %tmp38, i1 true, i1 true) #0
ret void
}
declare float @llvm.fabs.f32(float) #1
declare float @llvm.amdgcn.interp.p1(float, i32, i32, i32) #1
declare float @llvm.amdgcn.interp.p2(float, float, i32, i32, i32) #1
declare float @llvm.amdgcn.interp.mov(i32, i32, i32, i32) #1
declare void @llvm.amdgcn.exp.f32(i32, i32, float, float, float, float, i1, i1) #0
declare void @llvm.amdgcn.exp.compr.v2f16(i32, i32, <2 x half>, <2 x half>, i1, i1) #0
declare <2 x half> @llvm.amdgcn.cvt.pkrtz(float, float) #1
attributes #0 = { nounwind }
attributes #1 = { nounwind readnone }
|