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
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=aarch64 | FileCheck %s
define <4 x i16> @mlai16_trunc(<4 x i16> %vec0, <4 x i16> %vec1, <4 x i16> %vec2) {
; CHECK-LABEL: mlai16_trunc:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: smull v0.4s, v1.4h, v0.4h
; CHECK-NEXT: uaddw v0.4s, v0.4s, v2.4h
; CHECK-NEXT: xtn v0.4h, v0.4s
; CHECK-NEXT: ret
entry:
%v0 = sext <4 x i16> %vec0 to <4 x i32>
%v1 = sext <4 x i16> %vec1 to <4 x i32>
%v2 = sext <4 x i16> %vec2 to <4 x i32>
%v3 = mul <4 x i32> %v1, %v0
%v4 = add <4 x i32> %v3, %v2
%v5 = trunc <4 x i32> %v4 to <4 x i16>
ret <4 x i16> %v5
}
define <4 x i32> @mlai16_and(<4 x i16> %vec0, <4 x i16> %vec1, <4 x i16> %vec2) {
; CHECK-LABEL: mlai16_and:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: smull v0.4s, v1.4h, v0.4h
; CHECK-NEXT: movi v1.2d, #0x00ffff0000ffff
; CHECK-NEXT: uaddw v0.4s, v0.4s, v2.4h
; CHECK-NEXT: and v0.16b, v0.16b, v1.16b
; CHECK-NEXT: ret
entry:
%v0 = sext <4 x i16> %vec0 to <4 x i32>
%v1 = sext <4 x i16> %vec1 to <4 x i32>
%v2 = sext <4 x i16> %vec2 to <4 x i32>
%v3 = mul <4 x i32> %v1, %v0
%v4 = add <4 x i32> %v3, %v2
%v5 = and <4 x i32> %v4, <i32 65535, i32 65535, i32 65535, i32 65535>
ret <4 x i32> %v5
}
define void @mlai16_loadstore(ptr %a, ptr %b, ptr %c) {
; CHECK-LABEL: mlai16_loadstore:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr d0, [x0, #16]
; CHECK-NEXT: ldr d1, [x1, #16]
; CHECK-NEXT: smull v0.4s, v1.4h, v0.4h
; CHECK-NEXT: ldr d1, [x2, #16]
; CHECK-NEXT: uaddw v0.4s, v0.4s, v1.4h
; CHECK-NEXT: xtn v0.4h, v0.4s
; CHECK-NEXT: str d0, [x0, #16]
; CHECK-NEXT: ret
entry:
%scevgep0 = getelementptr i16, ptr %a, i32 8
%vec0 = load <4 x i16>, ptr %scevgep0, align 8
%v0 = sext <4 x i16> %vec0 to <4 x i32>
%scevgep1 = getelementptr i16, ptr %b, i32 8
%vec1 = load <4 x i16>, ptr %scevgep1, align 8
%v1 = sext <4 x i16> %vec1 to <4 x i32>
%scevgep2 = getelementptr i16, ptr %c, i32 8
%vec2 = load <4 x i16>, ptr %scevgep2, align 8
%v2 = sext <4 x i16> %vec2 to <4 x i32>
%v3 = mul <4 x i32> %v1, %v0
%v4 = add <4 x i32> %v3, %v2
%v5 = trunc <4 x i32> %v4 to <4 x i16>
%scevgep3 = getelementptr i16, ptr %a, i32 8
store <4 x i16> %v5, ptr %scevgep3, align 8
ret void
}
define <4 x i16> @addmuli16_trunc(<4 x i16> %vec0, <4 x i16> %vec1, <4 x i16> %vec2) {
; CHECK-LABEL: addmuli16_trunc:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: smull v1.4s, v1.4h, v2.4h
; CHECK-NEXT: smlal v1.4s, v0.4h, v2.4h
; CHECK-NEXT: xtn v0.4h, v1.4s
; CHECK-NEXT: ret
entry:
%v0 = sext <4 x i16> %vec0 to <4 x i32>
%v1 = sext <4 x i16> %vec1 to <4 x i32>
%v2 = sext <4 x i16> %vec2 to <4 x i32>
%v3 = add <4 x i32> %v1, %v0
%v4 = mul <4 x i32> %v3, %v2
%v5 = trunc <4 x i32> %v4 to <4 x i16>
ret <4 x i16> %v5
}
define <4 x i32> @addmuli16_and(<4 x i16> %vec0, <4 x i16> %vec1, <4 x i16> %vec2) {
; CHECK-LABEL: addmuli16_and:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: smull v1.4s, v1.4h, v2.4h
; CHECK-NEXT: movi v3.2d, #0x00ffff0000ffff
; CHECK-NEXT: smlal v1.4s, v0.4h, v2.4h
; CHECK-NEXT: and v0.16b, v1.16b, v3.16b
; CHECK-NEXT: ret
entry:
%v0 = sext <4 x i16> %vec0 to <4 x i32>
%v1 = sext <4 x i16> %vec1 to <4 x i32>
%v2 = sext <4 x i16> %vec2 to <4 x i32>
%v3 = add <4 x i32> %v1, %v0
%v4 = mul <4 x i32> %v3, %v2
%v5 = and <4 x i32> %v4, <i32 65535, i32 65535, i32 65535, i32 65535>
ret <4 x i32> %v5
}
define void @addmuli16_loadstore(ptr %a, ptr %b, ptr %c) {
; CHECK-LABEL: addmuli16_loadstore:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr d0, [x1, #16]
; CHECK-NEXT: ldr d1, [x2, #16]
; CHECK-NEXT: ldr d2, [x0, #16]
; CHECK-NEXT: smull v0.4s, v0.4h, v1.4h
; CHECK-NEXT: smlal v0.4s, v2.4h, v1.4h
; CHECK-NEXT: xtn v0.4h, v0.4s
; CHECK-NEXT: str d0, [x0, #16]
; CHECK-NEXT: ret
entry:
%scevgep0 = getelementptr i16, ptr %a, i32 8
%vec0 = load <4 x i16>, ptr %scevgep0, align 8
%v0 = sext <4 x i16> %vec0 to <4 x i32>
%scevgep1 = getelementptr i16, ptr %b, i32 8
%vec1 = load <4 x i16>, ptr %scevgep1, align 8
%v1 = sext <4 x i16> %vec1 to <4 x i32>
%scevgep2 = getelementptr i16, ptr %c, i32 8
%vec2 = load <4 x i16>, ptr %scevgep2, align 8
%v2 = sext <4 x i16> %vec2 to <4 x i32>
%v3 = add <4 x i32> %v1, %v0
%v4 = mul <4 x i32> %v3, %v2
%v5 = trunc <4 x i32> %v4 to <4 x i16>
%scevgep3 = getelementptr i16, ptr %a, i32 8
store <4 x i16> %v5, ptr %scevgep3, align 8
ret void
}
define <2 x i32> @mlai32_trunc(<2 x i32> %vec0, <2 x i32> %vec1, <2 x i32> %vec2) {
; CHECK-LABEL: mlai32_trunc:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: smull v0.2d, v1.2s, v0.2s
; CHECK-NEXT: uaddw v0.2d, v0.2d, v2.2s
; CHECK-NEXT: xtn v0.2s, v0.2d
; CHECK-NEXT: ret
entry:
%v0 = sext <2 x i32> %vec0 to <2 x i64>
%v1 = sext <2 x i32> %vec1 to <2 x i64>
%v2 = sext <2 x i32> %vec2 to <2 x i64>
%v3 = mul <2 x i64> %v1, %v0
%v4 = add <2 x i64> %v3, %v2
%v5 = trunc <2 x i64> %v4 to <2 x i32>
ret <2 x i32> %v5
}
define <2 x i64> @mlai32_and(<2 x i32> %vec0, <2 x i32> %vec1, <2 x i32> %vec2) {
; CHECK-LABEL: mlai32_and:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: smull v0.2d, v1.2s, v0.2s
; CHECK-NEXT: movi v1.2d, #0x000000ffffffff
; CHECK-NEXT: uaddw v0.2d, v0.2d, v2.2s
; CHECK-NEXT: and v0.16b, v0.16b, v1.16b
; CHECK-NEXT: ret
entry:
%v0 = sext <2 x i32> %vec0 to <2 x i64>
%v1 = sext <2 x i32> %vec1 to <2 x i64>
%v2 = sext <2 x i32> %vec2 to <2 x i64>
%v3 = mul <2 x i64> %v1, %v0
%v4 = add <2 x i64> %v3, %v2
%v5 = and <2 x i64> %v4, <i64 4294967295, i64 4294967295>
ret <2 x i64> %v5
}
define void @mlai32_loadstore(ptr %a, ptr %b, ptr %c) {
; CHECK-LABEL: mlai32_loadstore:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr d0, [x0, #32]
; CHECK-NEXT: ldr d1, [x1, #32]
; CHECK-NEXT: smull v0.2d, v1.2s, v0.2s
; CHECK-NEXT: ldr d1, [x2, #32]
; CHECK-NEXT: uaddw v0.2d, v0.2d, v1.2s
; CHECK-NEXT: xtn v0.2s, v0.2d
; CHECK-NEXT: str d0, [x0, #32]
; CHECK-NEXT: ret
entry:
%scevgep0 = getelementptr i32, ptr %a, i32 8
%vec0 = load <2 x i32>, ptr %scevgep0, align 8
%v0 = sext <2 x i32> %vec0 to <2 x i64>
%scevgep1 = getelementptr i32, ptr %b, i32 8
%vec1 = load <2 x i32>, ptr %scevgep1, align 8
%v1 = sext <2 x i32> %vec1 to <2 x i64>
%scevgep2 = getelementptr i32, ptr %c, i32 8
%vec2 = load <2 x i32>, ptr %scevgep2, align 8
%v2 = sext <2 x i32> %vec2 to <2 x i64>
%v3 = mul <2 x i64> %v1, %v0
%v4 = add <2 x i64> %v3, %v2
%v5 = trunc <2 x i64> %v4 to <2 x i32>
%scevgep3 = getelementptr i32, ptr %a, i32 8
store <2 x i32> %v5, ptr %scevgep3, align 8
ret void
}
define <2 x i32> @addmuli32_trunc(<2 x i32> %vec0, <2 x i32> %vec1, <2 x i32> %vec2) {
; CHECK-LABEL: addmuli32_trunc:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: smull v1.2d, v1.2s, v2.2s
; CHECK-NEXT: smlal v1.2d, v0.2s, v2.2s
; CHECK-NEXT: xtn v0.2s, v1.2d
; CHECK-NEXT: ret
entry:
%v0 = sext <2 x i32> %vec0 to <2 x i64>
%v1 = sext <2 x i32> %vec1 to <2 x i64>
%v2 = sext <2 x i32> %vec2 to <2 x i64>
%v3 = add <2 x i64> %v1, %v0
%v4 = mul <2 x i64> %v3, %v2
%v5 = trunc <2 x i64> %v4 to <2 x i32>
ret <2 x i32> %v5
}
define <2 x i64> @addmuli32_and(<2 x i32> %vec0, <2 x i32> %vec1, <2 x i32> %vec2) {
; CHECK-LABEL: addmuli32_and:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: smull v1.2d, v1.2s, v2.2s
; CHECK-NEXT: movi v3.2d, #0x000000ffffffff
; CHECK-NEXT: smlal v1.2d, v0.2s, v2.2s
; CHECK-NEXT: and v0.16b, v1.16b, v3.16b
; CHECK-NEXT: ret
entry:
%v0 = sext <2 x i32> %vec0 to <2 x i64>
%v1 = sext <2 x i32> %vec1 to <2 x i64>
%v2 = sext <2 x i32> %vec2 to <2 x i64>
%v3 = add <2 x i64> %v1, %v0
%v4 = mul <2 x i64> %v3, %v2
%v5 = and <2 x i64> %v4, <i64 4294967295, i64 4294967295>
ret <2 x i64> %v5
}
define void @addmuli32_loadstore(ptr %a, ptr %b, ptr %c) {
; CHECK-LABEL: addmuli32_loadstore:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr d0, [x1, #32]
; CHECK-NEXT: ldr d1, [x2, #32]
; CHECK-NEXT: ldr d2, [x0, #32]
; CHECK-NEXT: smull v0.2d, v0.2s, v1.2s
; CHECK-NEXT: smlal v0.2d, v2.2s, v1.2s
; CHECK-NEXT: xtn v0.2s, v0.2d
; CHECK-NEXT: str d0, [x0, #32]
; CHECK-NEXT: ret
entry:
%scevgep0 = getelementptr i32, ptr %a, i32 8
%vec0 = load <2 x i32>, ptr %scevgep0, align 8
%v0 = sext <2 x i32> %vec0 to <2 x i64>
%scevgep1 = getelementptr i32, ptr %b, i32 8
%vec1 = load <2 x i32>, ptr %scevgep1, align 8
%v1 = sext <2 x i32> %vec1 to <2 x i64>
%scevgep2 = getelementptr i32, ptr %c, i32 8
%vec2 = load <2 x i32>, ptr %scevgep2, align 8
%v2 = sext <2 x i32> %vec2 to <2 x i64>
%v3 = add <2 x i64> %v1, %v0
%v4 = mul <2 x i64> %v3, %v2
%v5 = trunc <2 x i64> %v4 to <2 x i32>
%scevgep3 = getelementptr i32, ptr %a, i32 8
store <2 x i32> %v5, ptr %scevgep3, align 8
ret void
}
define void @func1(ptr %a, ptr %b, ptr %c) {
; CHECK-LABEL: func1:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr d0, [x2, #16]
; CHECK-NEXT: ldr d1, [x1, #16]
; CHECK-NEXT: ldr d2, [x0, #16]
; CHECK-NEXT: sshll v0.4s, v0.4h, #0
; CHECK-NEXT: uaddw v0.4s, v0.4s, v1.4h
; CHECK-NEXT: xtn v1.4h, v0.4s
; CHECK-NEXT: str d1, [x0, #16]
; CHECK-NEXT: ldr d1, [x2, #16]
; CHECK-NEXT: sshll v1.4s, v1.4h, #0
; CHECK-NEXT: mul v0.4s, v1.4s, v0.4s
; CHECK-NEXT: xtn v1.4h, v0.4s
; CHECK-NEXT: str d1, [x1, #16]
; CHECK-NEXT: ldr d1, [x2, #16]
; CHECK-NEXT: smlal v0.4s, v1.4h, v2.4h
; CHECK-NEXT: xtn v0.4h, v0.4s
; CHECK-NEXT: str d0, [x0, #16]
; CHECK-NEXT: ret
entry:
; The test case trying to vectorize the pseudo code below.
; a[i] = b[i] + c[i];
; b[i] = aptr c[i];
; a[i] = b[i] + aptr c[i];
; Checking that vector load a[i] for "a[i] = b[i] + aptr c[i]" is
; scheduled before the first vector store to "a[i] = b[i] + c[i]".
; Checking that there is no vector load a[i] scheduled between the vector
; stores to a[i], otherwise the load of a[i] will be polluted by the first
; vector store to a[i].
; This test case check that the chain information is updated during
; lowerMUL for the new created Load SDNode.
%scevgep0 = getelementptr i16, ptr %a, i32 8
%vec0 = load <4 x i16>, ptr %scevgep0, align 8
%scevgep1 = getelementptr i16, ptr %b, i32 8
%vec1 = load <4 x i16>, ptr %scevgep1, align 8
%0 = zext <4 x i16> %vec1 to <4 x i32>
%scevgep2 = getelementptr i16, ptr %c, i32 8
%vec2 = load <4 x i16>, ptr %scevgep2, align 8
%1 = sext <4 x i16> %vec2 to <4 x i32>
%vec3 = add <4 x i32> %1, %0
%2 = trunc <4 x i32> %vec3 to <4 x i16>
%scevgep3 = getelementptr i16, ptr %a, i32 8
store <4 x i16> %2, ptr %scevgep3, align 8
%vec4 = load <4 x i16>, ptr %scevgep2, align 8
%3 = sext <4 x i16> %vec4 to <4 x i32>
%vec5 = mul <4 x i32> %3, %vec3
%4 = trunc <4 x i32> %vec5 to <4 x i16>
store <4 x i16> %4, ptr %scevgep1, align 8
%5 = sext <4 x i16> %vec0 to <4 x i32>
%vec6 = load <4 x i16>, ptr %scevgep2, align 8
%6 = sext <4 x i16> %vec6 to <4 x i32>
%vec7 = mul <4 x i32> %6, %5
%vec8 = add <4 x i32> %vec7, %vec5
%7 = trunc <4 x i32> %vec8 to <4 x i16>
store <4 x i16> %7, ptr %scevgep3, align 8
ret void
}
define void @func2(ptr %a, ptr %b, ptr %c) {
; CHECK-LABEL: func2:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: ldr d0, [x2, #16]
; CHECK-NEXT: ldr d1, [x1, #16]
; CHECK-NEXT: ldr d2, [x0, #16]
; CHECK-NEXT: sshll v0.4s, v0.4h, #0
; CHECK-NEXT: uaddw v0.4s, v0.4s, v1.4h
; CHECK-NEXT: xtn v1.4h, v0.4s
; CHECK-NEXT: str d1, [x0, #16]
; CHECK-NEXT: ldr d1, [x2, #16]
; CHECK-NEXT: sshll v1.4s, v1.4h, #0
; CHECK-NEXT: mul v0.4s, v1.4s, v0.4s
; CHECK-NEXT: xtn v1.4h, v0.4s
; CHECK-NEXT: str d1, [x1, #16]
; CHECK-NEXT: ldr d1, [x2, #16]
; CHECK-NEXT: smlal v0.4s, v1.4h, v2.4h
; CHECK-NEXT: uaddw v0.4s, v0.4s, v2.4h
; CHECK-NEXT: xtn v0.4h, v0.4s
; CHECK-NEXT: str d0, [x0, #16]
; CHECK-NEXT: ret
entry:
; The test case trying to vectorize the pseudo code below.
; a[i] = b[i] + c[i];
; b[i] = aptr c[i];
; a[i] = b[i] + aptr c[i] + a[i];
; Checking that vector load a[i] for "a[i] = b[i] + aptr c[i] + a[i]"
; is scheduled before the first vector store to "a[i] = b[i] + c[i]".
; Checking that there is no vector load a[i] scheduled between the first
; vector store to a[i] and the vector add of a[i], otherwise the load of
; a[i] will be polluted by the first vector store to a[i].
; This test case check that both the chain and value of the new created
; Load SDNode are updated during lowerMUL.
%scevgep0 = getelementptr i16, ptr %a, i32 8
%vec0 = load <4 x i16>, ptr %scevgep0, align 8
%scevgep1 = getelementptr i16, ptr %b, i32 8
%vec1 = load <4 x i16>, ptr %scevgep1, align 8
%0 = zext <4 x i16> %vec1 to <4 x i32>
%scevgep2 = getelementptr i16, ptr %c, i32 8
%vec2 = load <4 x i16>, ptr %scevgep2, align 8
%1 = sext <4 x i16> %vec2 to <4 x i32>
%vec3 = add <4 x i32> %1, %0
%2 = trunc <4 x i32> %vec3 to <4 x i16>
%scevgep3 = getelementptr i16, ptr %a, i32 8
store <4 x i16> %2, ptr %scevgep3, align 8
%vec4 = load <4 x i16>, ptr %scevgep2, align 8
%3 = sext <4 x i16> %vec4 to <4 x i32>
%vec5 = mul <4 x i32> %3, %vec3
%4 = trunc <4 x i32> %vec5 to <4 x i16>
store <4 x i16> %4, ptr %scevgep1, align 8
%5 = sext <4 x i16> %vec0 to <4 x i32>
%vec6 = load <4 x i16>, ptr %scevgep2, align 8
%6 = sext <4 x i16> %vec6 to <4 x i32>
%vec7 = mul <4 x i32> %6, %5
%vec8 = add <4 x i32> %vec7, %vec5
%vec9 = add <4 x i32> %vec8, %5
%7 = trunc <4 x i32> %vec9 to <4 x i16>
store <4 x i16> %7, ptr %scevgep3, align 8
ret void
}
|
