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
297
298
299
300
301
302
303
304
305
306
307
|
; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx908 -O0 -global-isel -stop-after=irtranslator -verify-machineinstrs --amdgpu-lower-module-lds-strategy=module -o - %s | FileCheck %s
define amdgpu_kernel void @asm_convergent() convergent{
; CHECK-LABEL: name: asm_convergent
; CHECK: bb.1 (%ir-block.0):
; CHECK-NEXT: INLINEASM &s_barrier, 33 /* sideeffect isconvergent attdialect */, !0
; CHECK-NEXT: S_ENDPGM 0
call void asm sideeffect "s_barrier", ""() convergent, !srcloc !0
ret void
}
define amdgpu_kernel void @asm_simple_memory_clobber() {
; CHECK-LABEL: name: asm_simple_memory_clobber
; CHECK: bb.1 (%ir-block.0):
; CHECK-NEXT: INLINEASM &"", 25 /* sideeffect mayload maystore attdialect */, !0
; CHECK-NEXT: INLINEASM &"", 1 /* sideeffect attdialect */, !0
; CHECK-NEXT: S_ENDPGM 0
call void asm sideeffect "", "~{memory}"(), !srcloc !0
call void asm sideeffect "", ""(), !srcloc !0
ret void
}
define amdgpu_kernel void @asm_simple_vgpr_clobber() {
; CHECK-LABEL: name: asm_simple_vgpr_clobber
; CHECK: bb.1 (%ir-block.0):
; CHECK-NEXT: INLINEASM &"v_mov_b32 v0, 7", 1 /* sideeffect attdialect */, 12 /* clobber */, implicit-def early-clobber $vgpr0, !0
; CHECK-NEXT: S_ENDPGM 0
call void asm sideeffect "v_mov_b32 v0, 7", "~{v0}"(), !srcloc !0
ret void
}
define amdgpu_kernel void @asm_simple_sgpr_clobber() {
; CHECK-LABEL: name: asm_simple_sgpr_clobber
; CHECK: bb.1 (%ir-block.0):
; CHECK-NEXT: INLINEASM &"s_mov_b32 s0, 7", 1 /* sideeffect attdialect */, 12 /* clobber */, implicit-def early-clobber $sgpr0, !0
; CHECK-NEXT: S_ENDPGM 0
call void asm sideeffect "s_mov_b32 s0, 7", "~{s0}"(), !srcloc !0
ret void
}
define amdgpu_kernel void @asm_simple_agpr_clobber() {
; CHECK-LABEL: name: asm_simple_agpr_clobber
; CHECK: bb.1 (%ir-block.0):
; CHECK-NEXT: INLINEASM &"; def a0", 1 /* sideeffect attdialect */, 12 /* clobber */, implicit-def early-clobber $agpr0, !0
; CHECK-NEXT: S_ENDPGM 0
call void asm sideeffect "; def a0", "~{a0}"(), !srcloc !0
ret void
}
define i32 @asm_vgpr_early_clobber() {
; CHECK-LABEL: name: asm_vgpr_early_clobber
; CHECK: bb.1 (%ir-block.0):
; CHECK-NEXT: INLINEASM &"v_mov_b32 $0, 7; v_mov_b32 $1, 7", 1 /* sideeffect attdialect */, 1769483 /* regdef-ec:VGPR_32 */, def early-clobber %0, 1769483 /* regdef-ec:VGPR_32 */, def early-clobber %1, !0
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY %0
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY %1
; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s32) = G_ADD [[COPY]], [[COPY1]]
; CHECK-NEXT: $vgpr0 = COPY [[ADD]](s32)
; CHECK-NEXT: SI_RETURN implicit $vgpr0
call { i32, i32 } asm sideeffect "v_mov_b32 $0, 7; v_mov_b32 $1, 7", "=&v,=&v"(), !srcloc !0
%asmresult = extractvalue { i32, i32 } %1, 0
%asmresult1 = extractvalue { i32, i32 } %1, 1
%add = add i32 %asmresult, %asmresult1
ret i32 %add
}
define i32 @test_specific_vgpr_output() nounwind {
; CHECK-LABEL: name: test_specific_vgpr_output
; CHECK: bb.1.entry:
; CHECK-NEXT: INLINEASM &"v_mov_b32 v1, 7", 0 /* attdialect */, 10 /* regdef */, implicit-def $vgpr1
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr1
; CHECK-NEXT: $vgpr0 = COPY [[COPY]](s32)
; CHECK-NEXT: SI_RETURN implicit $vgpr0
entry:
%0 = tail call i32 asm "v_mov_b32 v1, 7", "={v1}"() nounwind
ret i32 %0
}
define i32 @test_single_vgpr_output() nounwind {
; CHECK-LABEL: name: test_single_vgpr_output
; CHECK: bb.1.entry:
; CHECK-NEXT: INLINEASM &"v_mov_b32 $0, 7", 0 /* attdialect */, 1769482 /* regdef:VGPR_32 */, def %0
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY %0
; CHECK-NEXT: $vgpr0 = COPY [[COPY]](s32)
; CHECK-NEXT: SI_RETURN implicit $vgpr0
entry:
%0 = tail call i32 asm "v_mov_b32 $0, 7", "=v"() nounwind
ret i32 %0
}
define i32 @test_single_sgpr_output_s32() nounwind {
; CHECK-LABEL: name: test_single_sgpr_output_s32
; CHECK: bb.1.entry:
; CHECK-NEXT: INLINEASM &"s_mov_b32 $0, 7", 0 /* attdialect */, 1900554 /* regdef:SReg_32 */, def %0
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY %0
; CHECK-NEXT: $vgpr0 = COPY [[COPY]](s32)
; CHECK-NEXT: SI_RETURN implicit $vgpr0
entry:
%0 = tail call i32 asm "s_mov_b32 $0, 7", "=s"() nounwind
ret i32 %0
}
; Check support for returning several floats
define float @test_multiple_register_outputs_same() #0 {
; CHECK-LABEL: name: test_multiple_register_outputs_same
; CHECK: bb.1 (%ir-block.0):
; CHECK-NEXT: INLINEASM &"v_mov_b32 $0, 0; v_mov_b32 $1, 1", 0 /* attdialect */, 1769482 /* regdef:VGPR_32 */, def %0, 1769482 /* regdef:VGPR_32 */, def %1
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY %0
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY %1
; CHECK-NEXT: [[FADD:%[0-9]+]]:_(s32) = G_FADD [[COPY]], [[COPY1]]
; CHECK-NEXT: $vgpr0 = COPY [[FADD]](s32)
; CHECK-NEXT: SI_RETURN implicit $vgpr0
%1 = call { float, float } asm "v_mov_b32 $0, 0; v_mov_b32 $1, 1", "=v,=v"()
%asmresult = extractvalue { float, float } %1, 0
%asmresult1 = extractvalue { float, float } %1, 1
%add = fadd float %asmresult, %asmresult1
ret float %add
}
; Check support for returning several floats
define double @test_multiple_register_outputs_mixed() #0 {
; CHECK-LABEL: name: test_multiple_register_outputs_mixed
; CHECK: bb.1 (%ir-block.0):
; CHECK-NEXT: INLINEASM &"v_mov_b32 $0, 0; v_add_f64 $1, 0, 0", 0 /* attdialect */, 1769482 /* regdef:VGPR_32 */, def %0, 3080202 /* regdef:VReg_64 */, def %1
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY %0
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY %1
; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](s64)
; CHECK-NEXT: $vgpr0 = COPY [[UV]](s32)
; CHECK-NEXT: $vgpr1 = COPY [[UV1]](s32)
; CHECK-NEXT: SI_RETURN implicit $vgpr0, implicit $vgpr1
%1 = call { float, double } asm "v_mov_b32 $0, 0; v_add_f64 $1, 0, 0", "=v,=v"()
%asmresult = extractvalue { float, double } %1, 1
ret double %asmresult
}
define float @test_vector_output() nounwind {
; CHECK-LABEL: name: test_vector_output
; CHECK: bb.1 (%ir-block.0):
; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; CHECK-NEXT: INLINEASM &"v_add_f64 $0, 0, 0", 1 /* sideeffect attdialect */, 10 /* regdef */, implicit-def $vgpr14_vgpr15
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr14_vgpr15
; CHECK-NEXT: [[EVEC:%[0-9]+]]:_(s32) = G_EXTRACT_VECTOR_ELT [[COPY]](<2 x s32>), [[C]](s32)
; CHECK-NEXT: $vgpr0 = COPY [[EVEC]](s32)
; CHECK-NEXT: SI_RETURN implicit $vgpr0
%1 = tail call <2 x float> asm sideeffect "v_add_f64 $0, 0, 0", "={v[14:15]}"() nounwind
%2 = extractelement <2 x float> %1, i32 0
ret float %2
}
define amdgpu_kernel void @test_input_vgpr_imm() {
; CHECK-LABEL: name: test_input_vgpr_imm
; CHECK: bb.1 (%ir-block.0):
; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY [[C]](s32)
; CHECK-NEXT: INLINEASM &"v_mov_b32 v0, $0", 1 /* sideeffect attdialect */, 1769481 /* reguse:VGPR_32 */, [[COPY]]
; CHECK-NEXT: S_ENDPGM 0
call void asm sideeffect "v_mov_b32 v0, $0", "v"(i32 42)
ret void
}
define amdgpu_kernel void @test_input_sgpr_imm() {
; CHECK-LABEL: name: test_input_sgpr_imm
; CHECK: bb.1 (%ir-block.0):
; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
; CHECK-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY [[C]](s32)
; CHECK-NEXT: INLINEASM &"s_mov_b32 s0, $0", 1 /* sideeffect attdialect */, 1900553 /* reguse:SReg_32 */, [[COPY]]
; CHECK-NEXT: S_ENDPGM 0
call void asm sideeffect "s_mov_b32 s0, $0", "s"(i32 42)
ret void
}
define amdgpu_kernel void @test_input_imm() {
; CHECK-LABEL: name: test_input_imm
; CHECK: bb.1 (%ir-block.0):
; CHECK-NEXT: INLINEASM &"s_mov_b32 s0, $0", 9 /* sideeffect mayload attdialect */, 13 /* imm */, 42
; CHECK-NEXT: INLINEASM &"s_mov_b64 s[0:1], $0", 9 /* sideeffect mayload attdialect */, 13 /* imm */, 42
; CHECK-NEXT: S_ENDPGM 0
call void asm sideeffect "s_mov_b32 s0, $0", "i"(i32 42)
call void asm sideeffect "s_mov_b64 s[0:1], $0", "i"(i64 42)
ret void
}
define float @test_input_vgpr(i32 %src) nounwind {
; CHECK-LABEL: name: test_input_vgpr
; CHECK: bb.1.entry:
; CHECK-NEXT: liveins: $vgpr0
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]](s32)
; CHECK-NEXT: INLINEASM &"v_add_f32 $0, 1.0, $1", 0 /* attdialect */, 1769482 /* regdef:VGPR_32 */, def %1, 1769481 /* reguse:VGPR_32 */, [[COPY1]]
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY %1
; CHECK-NEXT: $vgpr0 = COPY [[COPY2]](s32)
; CHECK-NEXT: SI_RETURN implicit $vgpr0
entry:
%0 = tail call float asm "v_add_f32 $0, 1.0, $1", "=v,v"(i32 %src) nounwind
ret float %0
}
define i32 @test_memory_constraint(ptr addrspace(3) %a) nounwind {
; CHECK-LABEL: name: test_memory_constraint
; CHECK: bb.1 (%ir-block.0):
; CHECK-NEXT: liveins: $vgpr0
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
; CHECK-NEXT: INLINEASM &"ds_read_b32 $0, $1", 8 /* mayload attdialect */, 1769482 /* regdef:VGPR_32 */, def %1, 262158 /* mem:m */, [[COPY]](p3)
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY %1
; CHECK-NEXT: $vgpr0 = COPY [[COPY1]](s32)
; CHECK-NEXT: SI_RETURN implicit $vgpr0
%1 = tail call i32 asm "ds_read_b32 $0, $1", "=v,*m"(ptr addrspace(3) elementtype(i32) %a)
ret i32 %1
}
define i32 @test_vgpr_matching_constraint(i32 %a) nounwind {
; CHECK-LABEL: name: test_vgpr_matching_constraint
; CHECK: bb.1 (%ir-block.0):
; CHECK-NEXT: liveins: $vgpr0
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY]], [[C]]
; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[AND]](s32)
; CHECK-NEXT: INLINEASM &";", 1 /* sideeffect attdialect */, 1769482 /* regdef:VGPR_32 */, def %3, 2147483657 /* reguse tiedto:$0 */, [[COPY1]](tied-def 3)
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY %3
; CHECK-NEXT: $vgpr0 = COPY [[COPY2]](s32)
; CHECK-NEXT: SI_RETURN implicit $vgpr0
%and = and i32 %a, 1
%asm = call i32 asm sideeffect ";", "=v,0"(i32 %and)
ret i32 %asm
}
define i32 @test_sgpr_matching_constraint() nounwind {
; CHECK-LABEL: name: test_sgpr_matching_constraint
; CHECK: bb.1.entry:
; CHECK-NEXT: INLINEASM &"s_mov_b32 $0, 7", 0 /* attdialect */, 1900554 /* regdef:SReg_32 */, def %0
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY %0
; CHECK-NEXT: INLINEASM &"s_mov_b32 $0, 8", 0 /* attdialect */, 1900554 /* regdef:SReg_32 */, def %2
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY %2
; CHECK-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY [[COPY]](s32)
; CHECK-NEXT: [[COPY3:%[0-9]+]]:sreg_32 = COPY [[COPY1]](s32)
; CHECK-NEXT: INLINEASM &"s_add_u32 $0, $1, $2", 0 /* attdialect */, 1900554 /* regdef:SReg_32 */, def %4, 1900553 /* reguse:SReg_32 */, [[COPY2]], 2147483657 /* reguse tiedto:$0 */, [[COPY3]](tied-def 3)
; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY %4
; CHECK-NEXT: $vgpr0 = COPY [[COPY4]](s32)
; CHECK-NEXT: SI_RETURN implicit $vgpr0
entry:
%asm0 = tail call i32 asm "s_mov_b32 $0, 7", "=s"() nounwind
%asm1 = tail call i32 asm "s_mov_b32 $0, 8", "=s"() nounwind
%asm2 = tail call i32 asm "s_add_u32 $0, $1, $2", "=s,s,0"(i32 %asm0, i32 %asm1) nounwind
ret i32 %asm2
}
define void @test_many_matching_constraints(i32 %a, i32 %b, i32 %c) nounwind {
; CHECK-LABEL: name: test_many_matching_constraints
; CHECK: bb.1 (%ir-block.0):
; CHECK-NEXT: liveins: $vgpr0, $vgpr1, $vgpr2
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr2
; CHECK-NEXT: [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
; CHECK-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY2]](s32)
; CHECK-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[COPY]](s32)
; CHECK-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[COPY1]](s32)
; CHECK-NEXT: INLINEASM &"; ", 1 /* sideeffect attdialect */, 1769482 /* regdef:VGPR_32 */, def %3, 1769482 /* regdef:VGPR_32 */, def %4, 1769482 /* regdef:VGPR_32 */, def %5, 2147483657 /* reguse tiedto:$0 */, [[COPY3]](tied-def 3), 2147614729 /* reguse tiedto:$2 */, [[COPY4]](tied-def 7), 2147549193 /* reguse tiedto:$1 */, [[COPY5]](tied-def 5)
; CHECK-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY %3
; CHECK-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY %4
; CHECK-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY %5
; CHECK-NEXT: G_STORE [[COPY6]](s32), [[DEF]](p1) :: (store (s32) into `ptr addrspace(1) undef`, addrspace 1)
; CHECK-NEXT: G_STORE [[COPY7]](s32), [[DEF]](p1) :: (store (s32) into `ptr addrspace(1) undef`, addrspace 1)
; CHECK-NEXT: G_STORE [[COPY8]](s32), [[DEF]](p1) :: (store (s32) into `ptr addrspace(1) undef`, addrspace 1)
; CHECK-NEXT: SI_RETURN
%asm = call {i32, i32, i32} asm sideeffect "; ", "=v,=v,=v,0,2,1"(i32 %c, i32 %a, i32 %b)
%asmresult0 = extractvalue {i32, i32, i32} %asm, 0
store i32 %asmresult0, ptr addrspace(1) undef
%asmresult1 = extractvalue {i32, i32, i32} %asm, 1
store i32 %asmresult1, ptr addrspace(1) undef
%asmresult2 = extractvalue {i32, i32, i32} %asm, 2
store i32 %asmresult2, ptr addrspace(1) undef
ret void
}
define i32 @test_sgpr_to_vgpr_move_matching_constraint() nounwind {
; CHECK-LABEL: name: test_sgpr_to_vgpr_move_matching_constraint
; CHECK: bb.1.entry:
; CHECK-NEXT: INLINEASM &"s_mov_b32 $0, 7", 0 /* attdialect */, 1900554 /* regdef:SReg_32 */, def %0
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY %0
; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]](s32)
; CHECK-NEXT: INLINEASM &"v_mov_b32 $0, $1", 0 /* attdialect */, 1769482 /* regdef:VGPR_32 */, def %2, 2147483657 /* reguse tiedto:$0 */, [[COPY1]](tied-def 3)
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY %2
; CHECK-NEXT: $vgpr0 = COPY [[COPY2]](s32)
; CHECK-NEXT: SI_RETURN implicit $vgpr0
entry:
%asm0 = tail call i32 asm "s_mov_b32 $0, 7", "=s"() nounwind
%asm1 = tail call i32 asm "v_mov_b32 $0, $1", "=v,0"(i32 %asm0) nounwind
ret i32 %asm1
}
define amdgpu_kernel void @asm_constraint_n_n() {
; CHECK-LABEL: name: asm_constraint_n_n
; CHECK: bb.1 (%ir-block.0):
; CHECK-NEXT: INLINEASM &"s_trap ${0:n}", 1 /* sideeffect attdialect */, 13 /* imm */, 10
; CHECK-NEXT: S_ENDPGM 0
tail call void asm sideeffect "s_trap ${0:n}", "n"(i32 10) #1
ret void
}
!0 = !{i32 70}
|
