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
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=aarch64 -aarch64-neon-syntax=generic | FileCheck %s -check-prefixes=CHECK,SDAG
; RUN: llc < %s -global-isel=1 -global-isel-abort=2 -mtriple=aarch64 -aarch64-neon-syntax=generic 2>&1 | FileCheck %s --check-prefixes=CHECK,GISEL
; Function Attrs: nounwind readnone
declare i8 @llvm.vector.reduce.add.v2i8(<2 x i8>)
declare i8 @llvm.vector.reduce.add.v3i8(<3 x i8>)
declare i8 @llvm.vector.reduce.add.v4i8(<4 x i8>)
declare i8 @llvm.vector.reduce.add.v8i8(<8 x i8>)
declare i8 @llvm.vector.reduce.add.v16i8(<16 x i8>)
declare i8 @llvm.vector.reduce.add.v32i8(<32 x i8>)
declare i16 @llvm.vector.reduce.add.v2i16(<2 x i16>)
declare i16 @llvm.vector.reduce.add.v3i16(<3 x i16>)
declare i16 @llvm.vector.reduce.add.v4i16(<4 x i16>)
declare i16 @llvm.vector.reduce.add.v8i16(<8 x i16>)
declare i16 @llvm.vector.reduce.add.v16i16(<16 x i16>)
declare i32 @llvm.vector.reduce.add.v2i32(<2 x i32>)
declare i32 @llvm.vector.reduce.add.v3i32(<3 x i32>)
declare i32 @llvm.vector.reduce.add.v4i32(<4 x i32>)
declare i32 @llvm.vector.reduce.add.v8i32(<8 x i32>)
declare i64 @llvm.vector.reduce.add.v2i64(<2 x i64>)
declare i64 @llvm.vector.reduce.add.v3i64(<3 x i64>)
declare i64 @llvm.vector.reduce.add.v4i64(<4 x i64>)
declare i128 @llvm.vector.reduce.add.v2i128(<2 x i128>)
; GISEL: warning: Instruction selection used fallback path for addv_v2i8
; GISEL-NEXT: warning: Instruction selection used fallback path for addv_v3i8
; GISEL-NEXT: warning: Instruction selection used fallback path for addv_v4i8
; GISEL-NEXT: warning: Instruction selection used fallback path for addv_v2i16
; GISEL-NEXT: warning: Instruction selection used fallback path for addv_v3i16
; GISEL-NEXT: warning: Instruction selection used fallback path for addv_v3i32
; GISEL-NEXT: warning: Instruction selection used fallback path for addv_v3i64
; GISEL-NEXT: warning: Instruction selection used fallback path for addv_v2i128
define i8 @add_B(ptr %arr) {
; CHECK-LABEL: add_B:
; CHECK: // %bb.0:
; CHECK-NEXT: ldr q0, [x0]
; CHECK-NEXT: addv b0, v0.16b
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
%bin.rdx = load <16 x i8>, ptr %arr
%r = call i8 @llvm.vector.reduce.add.v16i8(<16 x i8> %bin.rdx)
ret i8 %r
}
define i16 @add_H(ptr %arr) {
; CHECK-LABEL: add_H:
; CHECK: // %bb.0:
; CHECK-NEXT: ldr q0, [x0]
; CHECK-NEXT: addv h0, v0.8h
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
%bin.rdx = load <8 x i16>, ptr %arr
%r = call i16 @llvm.vector.reduce.add.v8i16(<8 x i16> %bin.rdx)
ret i16 %r
}
define i32 @add_S( ptr %arr) {
; CHECK-LABEL: add_S:
; CHECK: // %bb.0:
; CHECK-NEXT: ldr q0, [x0]
; CHECK-NEXT: addv s0, v0.4s
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
%bin.rdx = load <4 x i32>, ptr %arr
%r = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> %bin.rdx)
ret i32 %r
}
define i64 @add_D(ptr %arr) {
; CHECK-LABEL: add_D:
; CHECK: // %bb.0:
; CHECK-NEXT: ldr q0, [x0]
; CHECK-NEXT: addp d0, v0.2d
; CHECK-NEXT: fmov x0, d0
; CHECK-NEXT: ret
%bin.rdx = load <2 x i64>, ptr %arr
%r = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %bin.rdx)
ret i64 %r
}
define i32 @oversized_ADDV_256(ptr noalias nocapture readonly %arg1, ptr noalias nocapture readonly %arg2) {
; SDAG-LABEL: oversized_ADDV_256:
; SDAG: // %bb.0: // %entry
; SDAG-NEXT: ldr d0, [x0]
; SDAG-NEXT: ldr d1, [x1]
; SDAG-NEXT: uabdl v0.8h, v0.8b, v1.8b
; SDAG-NEXT: uaddlv s0, v0.8h
; SDAG-NEXT: fmov w0, s0
; SDAG-NEXT: ret
;
; GISEL-LABEL: oversized_ADDV_256:
; GISEL: // %bb.0: // %entry
; GISEL-NEXT: ldr d1, [x0]
; GISEL-NEXT: ldr d2, [x1]
; GISEL-NEXT: movi v0.2d, #0000000000000000
; GISEL-NEXT: usubl v1.8h, v1.8b, v2.8b
; GISEL-NEXT: sshll v2.4s, v1.4h, #0
; GISEL-NEXT: sshll2 v3.4s, v1.8h, #0
; GISEL-NEXT: ssubw2 v0.4s, v0.4s, v1.8h
; GISEL-NEXT: cmlt v4.4s, v2.4s, #0
; GISEL-NEXT: cmlt v5.4s, v3.4s, #0
; GISEL-NEXT: neg v6.4s, v2.4s
; GISEL-NEXT: mov v1.16b, v4.16b
; GISEL-NEXT: bif v0.16b, v3.16b, v5.16b
; GISEL-NEXT: bsl v1.16b, v6.16b, v2.16b
; GISEL-NEXT: add v0.4s, v1.4s, v0.4s
; GISEL-NEXT: addv s0, v0.4s
; GISEL-NEXT: fmov w0, s0
; GISEL-NEXT: ret
entry:
%0 = load <8 x i8>, ptr %arg1, align 1
%1 = zext <8 x i8> %0 to <8 x i32>
%2 = load <8 x i8>, ptr %arg2, align 1
%3 = zext <8 x i8> %2 to <8 x i32>
%4 = sub nsw <8 x i32> %1, %3
%5 = icmp slt <8 x i32> %4, zeroinitializer
%6 = sub nsw <8 x i32> zeroinitializer, %4
%7 = select <8 x i1> %5, <8 x i32> %6, <8 x i32> %4
%r = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> %7)
ret i32 %r
}
declare i32 @llvm.vector.reduce.add.v16i32(<16 x i32>)
define i32 @oversized_ADDV_512(ptr %arr) {
; SDAG-LABEL: oversized_ADDV_512:
; SDAG: // %bb.0:
; SDAG-NEXT: ldp q0, q1, [x0, #32]
; SDAG-NEXT: ldp q2, q3, [x0]
; SDAG-NEXT: add v1.4s, v3.4s, v1.4s
; SDAG-NEXT: add v0.4s, v2.4s, v0.4s
; SDAG-NEXT: add v0.4s, v0.4s, v1.4s
; SDAG-NEXT: addv s0, v0.4s
; SDAG-NEXT: fmov w0, s0
; SDAG-NEXT: ret
;
; GISEL-LABEL: oversized_ADDV_512:
; GISEL: // %bb.0:
; GISEL-NEXT: ldp q0, q1, [x0]
; GISEL-NEXT: ldp q2, q3, [x0, #32]
; GISEL-NEXT: add v0.4s, v0.4s, v1.4s
; GISEL-NEXT: add v1.4s, v2.4s, v3.4s
; GISEL-NEXT: add v0.4s, v0.4s, v1.4s
; GISEL-NEXT: addv s0, v0.4s
; GISEL-NEXT: fmov w0, s0
; GISEL-NEXT: ret
%bin.rdx = load <16 x i32>, ptr %arr
%r = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> %bin.rdx)
ret i32 %r
}
define i8 @addv_combine_i8(<8 x i8> %a1, <8 x i8> %a2) {
; SDAG-LABEL: addv_combine_i8:
; SDAG: // %bb.0: // %entry
; SDAG-NEXT: add v0.8b, v0.8b, v1.8b
; SDAG-NEXT: addv b0, v0.8b
; SDAG-NEXT: fmov w0, s0
; SDAG-NEXT: ret
;
; GISEL-LABEL: addv_combine_i8:
; GISEL: // %bb.0: // %entry
; GISEL-NEXT: addv b0, v0.8b
; GISEL-NEXT: addv b1, v1.8b
; GISEL-NEXT: fmov w8, s0
; GISEL-NEXT: fmov w9, s1
; GISEL-NEXT: add w0, w9, w8, uxtb
; GISEL-NEXT: ret
entry:
%rdx.1 = call i8 @llvm.vector.reduce.add.v8i8(<8 x i8> %a1)
%rdx.2 = call i8 @llvm.vector.reduce.add.v8i8(<8 x i8> %a2)
%r = add i8 %rdx.1, %rdx.2
ret i8 %r
}
define i16 @addv_combine_i16(<4 x i16> %a1, <4 x i16> %a2) {
; SDAG-LABEL: addv_combine_i16:
; SDAG: // %bb.0: // %entry
; SDAG-NEXT: add v0.4h, v0.4h, v1.4h
; SDAG-NEXT: addv h0, v0.4h
; SDAG-NEXT: fmov w0, s0
; SDAG-NEXT: ret
;
; GISEL-LABEL: addv_combine_i16:
; GISEL: // %bb.0: // %entry
; GISEL-NEXT: addv h0, v0.4h
; GISEL-NEXT: addv h1, v1.4h
; GISEL-NEXT: fmov w8, s0
; GISEL-NEXT: fmov w9, s1
; GISEL-NEXT: add w0, w9, w8, uxth
; GISEL-NEXT: ret
entry:
%rdx.1 = call i16 @llvm.vector.reduce.add.v4i16(<4 x i16> %a1)
%rdx.2 = call i16 @llvm.vector.reduce.add.v4i16(<4 x i16> %a2)
%r = add i16 %rdx.1, %rdx.2
ret i16 %r
}
define i32 @addv_combine_i32(<4 x i32> %a1, <4 x i32> %a2) {
; SDAG-LABEL: addv_combine_i32:
; SDAG: // %bb.0: // %entry
; SDAG-NEXT: add v0.4s, v0.4s, v1.4s
; SDAG-NEXT: addv s0, v0.4s
; SDAG-NEXT: fmov w0, s0
; SDAG-NEXT: ret
;
; GISEL-LABEL: addv_combine_i32:
; GISEL: // %bb.0: // %entry
; GISEL-NEXT: addv s0, v0.4s
; GISEL-NEXT: addv s1, v1.4s
; GISEL-NEXT: fmov w8, s0
; GISEL-NEXT: fmov w9, s1
; GISEL-NEXT: add w0, w8, w9
; GISEL-NEXT: ret
entry:
%rdx.1 = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> %a1)
%rdx.2 = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> %a2)
%r = add i32 %rdx.1, %rdx.2
ret i32 %r
}
define i64 @addv_combine_i64(<2 x i64> %a1, <2 x i64> %a2) {
; SDAG-LABEL: addv_combine_i64:
; SDAG: // %bb.0: // %entry
; SDAG-NEXT: add v0.2d, v0.2d, v1.2d
; SDAG-NEXT: addp d0, v0.2d
; SDAG-NEXT: fmov x0, d0
; SDAG-NEXT: ret
;
; GISEL-LABEL: addv_combine_i64:
; GISEL: // %bb.0: // %entry
; GISEL-NEXT: addp d0, v0.2d
; GISEL-NEXT: addp d1, v1.2d
; GISEL-NEXT: fmov x8, d0
; GISEL-NEXT: fmov x9, d1
; GISEL-NEXT: add x0, x8, x9
; GISEL-NEXT: ret
entry:
%rdx.1 = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %a1)
%rdx.2 = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %a2)
%r = add i64 %rdx.1, %rdx.2
ret i64 %r
}
define i8 @addv_v2i8(<2 x i8> %a) {
; CHECK-LABEL: addv_v2i8:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: addp v0.2s, v0.2s, v0.2s
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
entry:
%arg1 = call i8 @llvm.vector.reduce.add.v2i8(<2 x i8> %a)
ret i8 %arg1
}
define i8 @addv_v3i8(<3 x i8> %a) {
; CHECK-LABEL: addv_v3i8:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: movi v0.2d, #0000000000000000
; CHECK-NEXT: mov v0.h[0], w0
; CHECK-NEXT: mov v0.h[1], w1
; CHECK-NEXT: mov v0.h[2], w2
; CHECK-NEXT: addv h0, v0.4h
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
entry:
%arg1 = call i8 @llvm.vector.reduce.add.v3i8(<3 x i8> %a)
ret i8 %arg1
}
define i8 @addv_v4i8(<4 x i8> %a) {
; CHECK-LABEL: addv_v4i8:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: addv h0, v0.4h
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
entry:
%arg1 = call i8 @llvm.vector.reduce.add.v4i8(<4 x i8> %a)
ret i8 %arg1
}
define i8 @addv_v8i8(<8 x i8> %a) {
; CHECK-LABEL: addv_v8i8:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: addv b0, v0.8b
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
entry:
%arg1 = call i8 @llvm.vector.reduce.add.v8i8(<8 x i8> %a)
ret i8 %arg1
}
define i8 @addv_v16i8(<16 x i8> %a) {
; CHECK-LABEL: addv_v16i8:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: addv b0, v0.16b
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
entry:
%arg1 = call i8 @llvm.vector.reduce.add.v16i8(<16 x i8> %a)
ret i8 %arg1
}
define i8 @addv_v32i8(<32 x i8> %a) {
; CHECK-LABEL: addv_v32i8:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: add v0.16b, v0.16b, v1.16b
; CHECK-NEXT: addv b0, v0.16b
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
entry:
%arg1 = call i8 @llvm.vector.reduce.add.v32i8(<32 x i8> %a)
ret i8 %arg1
}
define i16 @addv_v2i16(<2 x i16> %a) {
; CHECK-LABEL: addv_v2i16:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: addp v0.2s, v0.2s, v0.2s
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
entry:
%arg1 = call i16 @llvm.vector.reduce.add.v2i16(<2 x i16> %a)
ret i16 %arg1
}
define i16 @addv_v3i16(<3 x i16> %a) {
; CHECK-LABEL: addv_v3i16:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: mov v0.h[3], wzr
; CHECK-NEXT: addv h0, v0.4h
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
entry:
%arg1 = call i16 @llvm.vector.reduce.add.v3i16(<3 x i16> %a)
ret i16 %arg1
}
define i16 @addv_v4i16(<4 x i16> %a) {
; CHECK-LABEL: addv_v4i16:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: addv h0, v0.4h
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
entry:
%arg1 = call i16 @llvm.vector.reduce.add.v4i16(<4 x i16> %a)
ret i16 %arg1
}
define i16 @addv_v8i16(<8 x i16> %a) {
; CHECK-LABEL: addv_v8i16:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: addv h0, v0.8h
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
entry:
%arg1 = call i16 @llvm.vector.reduce.add.v8i16(<8 x i16> %a)
ret i16 %arg1
}
define i16 @addv_v16i16(<16 x i16> %a) {
; CHECK-LABEL: addv_v16i16:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: add v0.8h, v0.8h, v1.8h
; CHECK-NEXT: addv h0, v0.8h
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
entry:
%arg1 = call i16 @llvm.vector.reduce.add.v16i16(<16 x i16> %a)
ret i16 %arg1
}
define i32 @addv_v2i32(<2 x i32> %a) {
; CHECK-LABEL: addv_v2i32:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: addp v0.2s, v0.2s, v0.2s
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
entry:
%arg1 = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> %a)
ret i32 %arg1
}
define i32 @addv_v3i32(<3 x i32> %a) {
; CHECK-LABEL: addv_v3i32:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: mov v0.s[3], wzr
; CHECK-NEXT: addv s0, v0.4s
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
entry:
%arg1 = call i32 @llvm.vector.reduce.add.v3i32(<3 x i32> %a)
ret i32 %arg1
}
define i32 @addv_v4i32(<4 x i32> %a) {
; CHECK-LABEL: addv_v4i32:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: addv s0, v0.4s
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
entry:
%arg1 = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> %a)
ret i32 %arg1
}
define i32 @addv_v8i32(<8 x i32> %a) {
; CHECK-LABEL: addv_v8i32:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: add v0.4s, v0.4s, v1.4s
; CHECK-NEXT: addv s0, v0.4s
; CHECK-NEXT: fmov w0, s0
; CHECK-NEXT: ret
entry:
%arg1 = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> %a)
ret i32 %arg1
}
define i64 @addv_v2i64(<2 x i64> %a) {
; CHECK-LABEL: addv_v2i64:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: addp d0, v0.2d
; CHECK-NEXT: fmov x0, d0
; CHECK-NEXT: ret
entry:
%arg1 = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %a)
ret i64 %arg1
}
define i64 @addv_v3i64(<3 x i64> %a) {
; CHECK-LABEL: addv_v3i64:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: // kill: def $d2 killed $d2 def $q2
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
; CHECK-NEXT: // kill: def $d1 killed $d1 def $q1
; CHECK-NEXT: mov v0.d[1], v1.d[0]
; CHECK-NEXT: mov v2.d[1], xzr
; CHECK-NEXT: add v0.2d, v0.2d, v2.2d
; CHECK-NEXT: addp d0, v0.2d
; CHECK-NEXT: fmov x0, d0
; CHECK-NEXT: ret
entry:
%arg1 = call i64 @llvm.vector.reduce.add.v3i64(<3 x i64> %a)
ret i64 %arg1
}
define i64 @addv_v4i64(<4 x i64> %a) {
; CHECK-LABEL: addv_v4i64:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: add v0.2d, v0.2d, v1.2d
; CHECK-NEXT: addp d0, v0.2d
; CHECK-NEXT: fmov x0, d0
; CHECK-NEXT: ret
entry:
%arg1 = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> %a)
ret i64 %arg1
}
define i128 @addv_v2i128(<2 x i128> %a) {
; CHECK-LABEL: addv_v2i128:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: adds x0, x0, x2
; CHECK-NEXT: adc x1, x1, x3
; CHECK-NEXT: ret
entry:
%arg1 = call i128 @llvm.vector.reduce.add.v2i128(<2 x i128> %a)
ret i128 %arg1
}
|
