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
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=CHECK %s
; Check that WQM is not triggered by the softwqm intrinsic alone.
;
define amdgpu_ps float @test1(i32 inreg %idx0, i32 inreg %idx1) {
; CHECK-LABEL: test1:
; CHECK: ; %bb.0: ; %main_body
; CHECK-NEXT: v_mov_b32_e32 v0, s0
; CHECK-NEXT: v_mov_b32_e32 v1, s1
; CHECK-NEXT: buffer_load_dword v0, v0, s[0:3], 0 idxen
; CHECK-NEXT: buffer_load_dword v1, v1, s[0:3], 0 idxen
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: v_add_f32_e32 v0, v0, v1
; CHECK-NEXT: ; kill: def $vgpr0 killed $vgpr0 killed $exec
; CHECK-NEXT: ; return to shader part epilog
main_body:
%src0 = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> undef, i32 %idx0, i32 0, i32 0, i32 0)
%src1 = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> undef, i32 %idx1, i32 0, i32 0, i32 0)
%out = fadd float %src0, %src1
%out.0 = call float @llvm.amdgcn.softwqm.f32(float %out)
ret float %out.0
}
; Check that the softwqm intrinsic works correctly for integers.
;
define amdgpu_ps float @test2(i32 inreg %idx0, i32 inreg %idx1) {
; CHECK-LABEL: test2:
; CHECK: ; %bb.0: ; %main_body
; CHECK-NEXT: v_mov_b32_e32 v0, s0
; CHECK-NEXT: v_mov_b32_e32 v1, s1
; CHECK-NEXT: buffer_load_dword v0, v0, s[0:3], 0 idxen
; CHECK-NEXT: buffer_load_dword v1, v1, s[0:3], 0 idxen
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: v_add_f32_e32 v0, v0, v1
; CHECK-NEXT: ; kill: def $vgpr0 killed $vgpr0 killed $exec
; CHECK-NEXT: ; return to shader part epilog
main_body:
%src0 = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> undef, i32 %idx0, i32 0, i32 0, i32 0)
%src1 = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> undef, i32 %idx1, i32 0, i32 0, i32 0)
%out = fadd float %src0, %src1
%out.0 = bitcast float %out to i32
%out.1 = call i32 @llvm.amdgcn.softwqm.i32(i32 %out.0)
%out.2 = bitcast i32 %out.1 to float
ret float %out.2
}
; Make sure the transition from WQM to Exact to softwqm does not trigger WQM.
;
define amdgpu_ps float @test_softwqm1(i32 inreg %idx0, i32 inreg %idx1) {
; CHECK-LABEL: test_softwqm1:
; CHECK: ; %bb.0: ; %main_body
; CHECK-NEXT: v_mov_b32_e32 v0, s0
; CHECK-NEXT: v_mov_b32_e32 v2, s1
; CHECK-NEXT: buffer_load_dword v1, v0, s[0:3], 0 idxen
; CHECK-NEXT: buffer_load_dword v2, v2, s[0:3], 0 idxen
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: v_add_f32_e32 v1, v1, v2
; CHECK-NEXT: buffer_store_dword v1, v0, s[0:3], 0 idxen
; CHECK-NEXT: v_add_f32_e32 v0, v1, v1
; CHECK-NEXT: ; kill: def $vgpr0 killed $vgpr0 killed $exec
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: ; return to shader part epilog
main_body:
%src0 = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> undef, i32 %idx0, i32 0, i32 0, i32 0)
%src1 = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> undef, i32 %idx1, i32 0, i32 0, i32 0)
%temp = fadd float %src0, %src1
call void @llvm.amdgcn.struct.buffer.store.f32(float %temp, <4 x i32> undef, i32 %idx0, i32 0, i32 0, i32 0)
%out = fadd float %temp, %temp
%out.0 = call float @llvm.amdgcn.softwqm.f32(float %out)
ret float %out.0
}
; Make sure the transition from WQM to Exact to softwqm does trigger WQM.
;
define amdgpu_ps float @test_softwqm2(i32 inreg %idx0, i32 inreg %idx1) {
; CHECK-LABEL: test_softwqm2:
; CHECK: ; %bb.0: ; %main_body
; CHECK-NEXT: s_mov_b64 s[2:3], exec
; CHECK-NEXT: s_wqm_b64 exec, exec
; CHECK-NEXT: v_mov_b32_e32 v0, s0
; CHECK-NEXT: v_mov_b32_e32 v2, s1
; CHECK-NEXT: buffer_load_dword v1, v0, s[0:3], 0 idxen
; CHECK-NEXT: buffer_load_dword v2, v2, s[0:3], 0 idxen
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: v_add_f32_e32 v1, v1, v2
; CHECK-NEXT: v_mov_b32_e32 v2, v1
; CHECK-NEXT: v_add_f32_e32 v1, v1, v1
; CHECK-NEXT: s_and_b64 exec, exec, s[2:3]
; CHECK-NEXT: buffer_store_dword v2, v0, s[0:3], 0 idxen
; CHECK-NEXT: s_wqm_b64 exec, exec
; CHECK-NEXT: v_mov_b32_e32 v0, v1
; CHECK-NEXT: s_and_b64 exec, exec, s[2:3]
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: ; return to shader part epilog
main_body:
%src0 = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> undef, i32 %idx0, i32 0, i32 0, i32 0)
%src1 = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> undef, i32 %idx1, i32 0, i32 0, i32 0)
%temp = fadd float %src0, %src1
%temp.0 = call float @llvm.amdgcn.wqm.f32(float %temp)
call void @llvm.amdgcn.struct.buffer.store.f32(float %temp.0, <4 x i32> undef, i32 %idx0, i32 0, i32 0, i32 0)
%out = fadd float %temp, %temp
%out.0 = call float @llvm.amdgcn.softwqm.f32(float %out)
ret float %out.0
}
; NOTE: llvm.amdgcn.wwm is deprecated, use llvm.amdgcn.strict.wwm instead.
; Make sure the transition from Exact to STRICT_WWM then softwqm does not trigger WQM.
;
define amdgpu_ps float @test_wwm1(i32 inreg %idx0, i32 inreg %idx1) {
; CHECK-LABEL: test_wwm1:
; CHECK: ; %bb.0: ; %main_body
; CHECK-NEXT: s_or_saveexec_b64 s[2:3], -1
; CHECK-NEXT: v_mov_b32_e32 v1, s0
; CHECK-NEXT: buffer_load_dword v2, v1, s[0:3], 0 idxen
; CHECK-NEXT: s_mov_b64 exec, s[2:3]
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: buffer_store_dword v2, v1, s[0:3], 0 idxen
; CHECK-NEXT: s_or_saveexec_b64 s[2:3], -1
; CHECK-NEXT: v_mov_b32_e32 v1, s1
; CHECK-NEXT: buffer_load_dword v1, v1, s[0:3], 0 idxen
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: v_add_f32_e32 v1, v2, v1
; CHECK-NEXT: s_mov_b64 exec, s[2:3]
; CHECK-NEXT: v_mov_b32_e32 v0, v1
; CHECK-NEXT: v_add_f32_e32 v0, v0, v0
; CHECK-NEXT: ; kill: def $vgpr0 killed $vgpr0 killed $exec
; CHECK-NEXT: ; return to shader part epilog
main_body:
%src0 = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> undef, i32 %idx0, i32 0, i32 0, i32 0)
call void @llvm.amdgcn.struct.buffer.store.f32(float %src0, <4 x i32> undef, i32 %idx0, i32 0, i32 0, i32 0)
%src1 = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> undef, i32 %idx1, i32 0, i32 0, i32 0)
%temp = fadd float %src0, %src1
%temp.0 = call float @llvm.amdgcn.wwm.f32(float %temp)
%out = fadd float %temp.0, %temp.0
%out.0 = call float @llvm.amdgcn.softwqm.f32(float %out)
ret float %out.0
}
; Make sure the transition from Exact to STRICT_WWM then softwqm does not trigger WQM.
;
define amdgpu_ps float @test_strict_wwm1(i32 inreg %idx0, i32 inreg %idx1) {
; CHECK-LABEL: test_strict_wwm1:
; CHECK: ; %bb.0: ; %main_body
; CHECK-NEXT: s_or_saveexec_b64 s[2:3], -1
; CHECK-NEXT: v_mov_b32_e32 v1, s0
; CHECK-NEXT: buffer_load_dword v2, v1, s[0:3], 0 idxen
; CHECK-NEXT: s_mov_b64 exec, s[2:3]
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: buffer_store_dword v2, v1, s[0:3], 0 idxen
; CHECK-NEXT: s_or_saveexec_b64 s[2:3], -1
; CHECK-NEXT: v_mov_b32_e32 v1, s1
; CHECK-NEXT: buffer_load_dword v1, v1, s[0:3], 0 idxen
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: v_add_f32_e32 v1, v2, v1
; CHECK-NEXT: s_mov_b64 exec, s[2:3]
; CHECK-NEXT: v_mov_b32_e32 v0, v1
; CHECK-NEXT: v_add_f32_e32 v0, v0, v0
; CHECK-NEXT: ; kill: def $vgpr0 killed $vgpr0 killed $exec
; CHECK-NEXT: ; return to shader part epilog
main_body:
%src0 = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> undef, i32 %idx0, i32 0, i32 0, i32 0)
call void @llvm.amdgcn.struct.buffer.store.f32(float %src0, <4 x i32> undef, i32 %idx0, i32 0, i32 0, i32 0)
%src1 = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> undef, i32 %idx1, i32 0, i32 0, i32 0)
%temp = fadd float %src0, %src1
%temp.0 = call float @llvm.amdgcn.strict.wwm.f32(float %temp)
%out = fadd float %temp.0, %temp.0
%out.0 = call float @llvm.amdgcn.softwqm.f32(float %out)
ret float %out.0
}
; Check that softwqm on one case of branch does not trigger WQM for shader.
;
define amdgpu_ps float @test_control_flow_0(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, i32 inreg %idx0, i32 inreg %idx1, i32 %c, i32 %z, float %data) {
; CHECK-LABEL: test_control_flow_0:
; CHECK: ; %bb.0: ; %main_body
; CHECK-NEXT: v_cmp_ne_u32_e32 vcc, 0, v1
; CHECK-NEXT: s_and_saveexec_b64 s[0:1], vcc
; CHECK-NEXT: s_xor_b64 s[0:1], exec, s[0:1]
; CHECK-NEXT: s_cbranch_execz .LBB6_2
; CHECK-NEXT: ; %bb.1: ; %ELSE
; CHECK-NEXT: buffer_store_dword v2, v0, s[0:3], 0 idxen
; CHECK-NEXT: .LBB6_2: ; %Flow
; CHECK-NEXT: s_andn2_saveexec_b64 s[0:1], s[0:1]
; CHECK-NEXT: s_cbranch_execz .LBB6_4
; CHECK-NEXT: ; %bb.3: ; %IF
; CHECK-NEXT: v_mov_b32_e32 v0, s12
; CHECK-NEXT: v_mov_b32_e32 v1, s13
; CHECK-NEXT: buffer_load_dword v0, v0, s[0:3], 0 idxen
; CHECK-NEXT: buffer_load_dword v1, v1, s[0:3], 0 idxen
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: v_add_f32_e32 v2, v0, v1
; CHECK-NEXT: ; kill: def $vgpr2 killed $vgpr2 killed $exec
; CHECK-NEXT: .LBB6_4: ; %END
; CHECK-NEXT: s_or_b64 exec, exec, s[0:1]
; CHECK-NEXT: v_mov_b32_e32 v0, v2
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: ; return to shader part epilog
main_body:
%cmp = icmp eq i32 %z, 0
br i1 %cmp, label %IF, label %ELSE
IF:
%src0 = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> undef, i32 %idx0, i32 0, i32 0, i32 0)
%src1 = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> undef, i32 %idx1, i32 0, i32 0, i32 0)
%out = fadd float %src0, %src1
%data.if = call float @llvm.amdgcn.softwqm.f32(float %out)
br label %END
ELSE:
call void @llvm.amdgcn.struct.buffer.store.f32(float %data, <4 x i32> undef, i32 %c, i32 0, i32 0, i32 0)
br label %END
END:
%r = phi float [ %data.if, %IF ], [ %data, %ELSE ]
ret float %r
}
; Check that softwqm on one case of branch is treated as WQM in WQM shader.
;
define amdgpu_ps float @test_control_flow_1(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, i32 inreg %idx0, i32 inreg %idx1, i32 %c, i32 %z, float %data) {
; CHECK-LABEL: test_control_flow_1:
; CHECK: ; %bb.0: ; %main_body
; CHECK-NEXT: s_mov_b64 s[14:15], exec
; CHECK-NEXT: s_wqm_b64 exec, exec
; CHECK-NEXT: v_cmp_ne_u32_e32 vcc, 0, v1
; CHECK-NEXT: s_and_saveexec_b64 s[16:17], vcc
; CHECK-NEXT: s_xor_b64 s[16:17], exec, s[16:17]
; CHECK-NEXT: s_cbranch_execz .LBB7_2
; CHECK-NEXT: ; %bb.1: ; %ELSE
; CHECK-NEXT: image_sample v1, v0, s[0:7], s[8:11] dmask:0x1
; CHECK-NEXT: s_and_saveexec_b64 s[18:19], s[14:15]
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: image_sample v1, v1, s[0:7], s[8:11] dmask:0x1
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: buffer_store_dword v1, v0, s[0:3], 0 idxen
; CHECK-NEXT: s_mov_b64 exec, s[18:19]
; CHECK-NEXT: .LBB7_2: ; %Flow
; CHECK-NEXT: s_andn2_saveexec_b64 s[0:1], s[16:17]
; CHECK-NEXT: s_cbranch_execz .LBB7_4
; CHECK-NEXT: ; %bb.3: ; %IF
; CHECK-NEXT: v_mov_b32_e32 v0, s12
; CHECK-NEXT: v_mov_b32_e32 v1, s13
; CHECK-NEXT: buffer_load_dword v0, v0, s[0:3], 0 idxen
; CHECK-NEXT: buffer_load_dword v1, v1, s[0:3], 0 idxen
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: v_add_f32_e32 v2, v0, v1
; CHECK-NEXT: ; kill: def $vgpr2 killed $vgpr2 killed $exec
; CHECK-NEXT: .LBB7_4: ; %END
; CHECK-NEXT: s_or_b64 exec, exec, s[0:1]
; CHECK-NEXT: s_and_b64 exec, exec, s[14:15]
; CHECK-NEXT: v_mov_b32_e32 v0, v2
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: ; return to shader part epilog
main_body:
%cmp = icmp eq i32 %z, 0
br i1 %cmp, label %IF, label %ELSE
IF:
%src0 = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> undef, i32 %idx0, i32 0, i32 0, i32 0)
%src1 = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> undef, i32 %idx1, i32 0, i32 0, i32 0)
%out = fadd float %src0, %src1
%data.if = call float @llvm.amdgcn.softwqm.f32(float %out)
br label %END
ELSE:
%c.bc = bitcast i32 %c to float
%tex = call <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f32(i32 15, float %c.bc, <8 x i32> %rsrc, <4 x i32> %sampler, i1 0, i32 0, i32 0) #0
%tex0 = extractelement <4 x float> %tex, i32 0
%dtex = call <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f32(i32 15, float %tex0, <8 x i32> %rsrc, <4 x i32> %sampler, i1 0, i32 0, i32 0) #0
%data.sample = extractelement <4 x float> %dtex, i32 0
call void @llvm.amdgcn.struct.buffer.store.f32(float %data.sample, <4 x i32> undef, i32 %c, i32 0, i32 0, i32 0)
br label %END
END:
%r = phi float [ %data.if, %IF ], [ %data, %ELSE ]
ret float %r
}
declare void @llvm.amdgcn.struct.buffer.store.f32(float, <4 x i32>, i32, i32, i32, i32 immarg) #2
declare void @llvm.amdgcn.struct.buffer.store.v4f32(<4 x float>, <4 x i32>, i32, i32, i32, i32 immarg) #2
declare float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32>, i32, i32, i32, i32 immarg) #3
declare <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f32(i32, float, <8 x i32>, <4 x i32>, i1, i32, i32) #3
declare <4 x float> @llvm.amdgcn.image.sample.2d.v4f32.f32(i32, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #3
declare float @llvm.amdgcn.wqm.f32(float) #3
declare float @llvm.amdgcn.softwqm.f32(float) #3
declare i32 @llvm.amdgcn.softwqm.i32(i32) #3
declare float @llvm.amdgcn.strict.wwm.f32(float) #3
declare float @llvm.amdgcn.wwm.f32(float) #3
declare void @llvm.amdgcn.wqm.demote(i1) #1
attributes #1 = { nounwind }
attributes #2 = { nounwind readonly }
attributes #3 = { nounwind readnone }
|
