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
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
|
dnl x86 mpn_sqr_basecase -- square an mpn number, optimised for atom.
dnl Contributed to the GNU project by Torbjorn Granlund and Marco Bodrato.
dnl Copyright 2011 Free Software Foundation, Inc.
dnl This file is part of the GNU MP Library.
dnl
dnl The GNU MP Library is free software; you can redistribute it and/or modify
dnl it under the terms of either:
dnl
dnl * the GNU Lesser General Public License as published by the Free
dnl Software Foundation; either version 3 of the License, or (at your
dnl option) any later version.
dnl
dnl or
dnl
dnl * the GNU General Public License as published by the Free Software
dnl Foundation; either version 2 of the License, or (at your option) any
dnl later version.
dnl
dnl or both in parallel, as here.
dnl
dnl The GNU MP Library is distributed in the hope that it will be useful, but
dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
dnl for more details.
dnl
dnl You should have received copies of the GNU General Public License and the
dnl GNU Lesser General Public License along with the GNU MP Library. If not,
dnl see https://www.gnu.org/licenses/.
include(`../config.m4')
C TODO
C * Check if 'jmp N(%esp)' is well-predicted enough to allow us to combine the
C 4 large loops into one; we could use it for the outer loop branch.
C * Optimise code outside of inner loops.
C * Write combined addmul_1 feed-in a wind-down code, and use when iterating
C outer each loop. ("Overlapping software pipelining")
C * Perhaps use caller-saves regs for inlined mul_1, allowing us to postpone
C all pushes.
C * Perhaps write special code for n < M, for some small M.
C * Replace inlined addmul_1 with smaller code from aorsmul_1.asm, or perhaps
C with even less pipelined code.
C * We run the outer loop until we have a 2-limb by 1-limb addmul_1 left.
C Consider breaking out earlier, saving high the cost of short loops.
C void mpn_sqr_basecase (mp_ptr wp,
C mp_srcptr xp, mp_size_t xn);
define(`rp', `%edi')
define(`up', `%esi')
define(`n', `%ecx')
define(`un', `%ebp')
TEXT
ALIGN(16)
PROLOGUE(mpn_sqr_basecase)
push %edi
push %esi
mov 12(%esp), rp
mov 16(%esp), up
mov 20(%esp), n
lea 4(rp), rp C write triangular product starting at rp[1]
dec n
movd (up), %mm7
jz L(one)
lea 4(up), up
push %ebx
push %ebp
mov n, %eax
movd (up), %mm0
neg n
pmuludq %mm7, %mm0
pxor %mm6, %mm6
mov n, un
and $3, %eax
jz L(of0)
cmp $2, %eax
jc L(of1)
jz L(of2)
C ================================================================
jmp L(m3)
ALIGN(16)
L(lm3): movd -4(up), %mm0
pmuludq %mm7, %mm0
psrlq $32, %mm6
lea 16(rp), rp
paddq %mm0, %mm6
movd (up), %mm0
pmuludq %mm7, %mm0
movd %mm6, -4(rp)
psrlq $32, %mm6
L(m3): paddq %mm0, %mm6
movd 4(up), %mm0
pmuludq %mm7, %mm0
movd %mm6, (rp)
psrlq $32, %mm6
paddq %mm0, %mm6
movd 8(up), %mm0
pmuludq %mm7, %mm0
movd %mm6, 4(rp)
psrlq $32, %mm6
paddq %mm0, %mm6
add $4, un
movd %mm6, 8(rp)
lea 16(up), up
js L(lm3)
psrlq $32, %mm6
movd %mm6, 12(rp)
inc n
C jz L(done)
lea -12(up), up
lea 4(rp), rp
jmp L(ol2)
C ================================================================
ALIGN(16)
L(lm0): movd (up), %mm0
pmuludq %mm7, %mm0
psrlq $32, %mm6
lea 16(rp), rp
L(of0): paddq %mm0, %mm6
movd 4(up), %mm0
pmuludq %mm7, %mm0
movd %mm6, (rp)
psrlq $32, %mm6
paddq %mm0, %mm6
movd 8(up), %mm0
pmuludq %mm7, %mm0
movd %mm6, 4(rp)
psrlq $32, %mm6
paddq %mm0, %mm6
movd 12(up), %mm0
pmuludq %mm7, %mm0
movd %mm6, 8(rp)
psrlq $32, %mm6
paddq %mm0, %mm6
add $4, un
movd %mm6, 12(rp)
lea 16(up), up
js L(lm0)
psrlq $32, %mm6
movd %mm6, 16(rp)
inc n
C jz L(done)
lea -8(up), up
lea 8(rp), rp
jmp L(ol3)
C ================================================================
ALIGN(16)
L(lm1): movd -12(up), %mm0
pmuludq %mm7, %mm0
psrlq $32, %mm6
lea 16(rp), rp
paddq %mm0, %mm6
movd -8(up), %mm0
pmuludq %mm7, %mm0
movd %mm6, -12(rp)
psrlq $32, %mm6
paddq %mm0, %mm6
movd -4(up), %mm0
pmuludq %mm7, %mm0
movd %mm6, -8(rp)
psrlq $32, %mm6
paddq %mm0, %mm6
movd (up), %mm0
pmuludq %mm7, %mm0
movd %mm6, -4(rp)
psrlq $32, %mm6
L(of1): paddq %mm0, %mm6
add $4, un
movd %mm6, (rp)
lea 16(up), up
js L(lm1)
psrlq $32, %mm6
movd %mm6, 4(rp)
inc n
jz L(done) C goes away when we add special n=2 code
lea -20(up), up
lea -4(rp), rp
jmp L(ol0)
C ================================================================
ALIGN(16)
L(lm2): movd -8(up), %mm0
pmuludq %mm7, %mm0
psrlq $32, %mm6
lea 16(rp), rp
paddq %mm0, %mm6
movd -4(up), %mm0
pmuludq %mm7, %mm0
movd %mm6, -8(rp)
psrlq $32, %mm6
paddq %mm0, %mm6
movd (up), %mm0
pmuludq %mm7, %mm0
movd %mm6, -4(rp)
psrlq $32, %mm6
L(of2): paddq %mm0, %mm6
movd 4(up), %mm0
pmuludq %mm7, %mm0
movd %mm6, (rp)
psrlq $32, %mm6
paddq %mm0, %mm6
add $4, un
movd %mm6, 4(rp)
lea 16(up), up
js L(lm2)
psrlq $32, %mm6
movd %mm6, 8(rp)
inc n
C jz L(done)
lea -16(up), up
C lea (rp), rp
C jmp L(ol1)
C ================================================================
L(ol1): lea 4(up,n,4), up
movd (up), %mm7 C read next U invariant limb
lea 8(rp,n,4), rp
mov n, un
movd 4(up), %mm1
pmuludq %mm7, %mm1
sar $2, un
movd %mm1, %ebx
inc un
jz L(re1)
movd 8(up), %mm0
pmuludq %mm7, %mm0
xor %edx, %edx C zero edx and CF
jmp L(a1)
L(la1): adc $0, %edx
add %ebx, 12(rp)
movd %mm0, %eax
pmuludq %mm7, %mm1
lea 16(rp), rp
psrlq $32, %mm0
adc %edx, %eax
movd %mm0, %edx
movd %mm1, %ebx
movd 8(up), %mm0
pmuludq %mm7, %mm0
adc $0, %edx
add %eax, (rp)
L(a1): psrlq $32, %mm1
adc %edx, %ebx
movd %mm1, %edx
movd %mm0, %eax
movd 12(up), %mm1
pmuludq %mm7, %mm1
adc $0, %edx
add %ebx, 4(rp)
psrlq $32, %mm0
adc %edx, %eax
movd %mm0, %edx
movd %mm1, %ebx
lea 16(up), up
movd (up), %mm0
adc $0, %edx
add %eax, 8(rp)
psrlq $32, %mm1
adc %edx, %ebx
movd %mm1, %edx
pmuludq %mm7, %mm0
inc un
movd 4(up), %mm1
jnz L(la1)
adc un, %edx C un is zero here
add %ebx, 12(rp)
movd %mm0, %eax
pmuludq %mm7, %mm1
lea 16(rp), rp
psrlq $32, %mm0
adc %edx, %eax
movd %mm0, %edx
movd %mm1, %ebx
adc un, %edx
add %eax, (rp)
psrlq $32, %mm1
adc %edx, %ebx
movd %mm1, %eax
adc un, %eax
add %ebx, 4(rp)
adc un, %eax
mov %eax, 8(rp)
inc n
C ================================================================
L(ol0): lea (up,n,4), up
movd 4(up), %mm7 C read next U invariant limb
lea 4(rp,n,4), rp
mov n, un
movd 8(up), %mm0
pmuludq %mm7, %mm0
sar $2, un
movd 12(up), %mm1
movd %mm0, %eax
pmuludq %mm7, %mm1
xor %edx, %edx C zero edx and CF
jmp L(a0)
L(la0): adc $0, %edx
add %ebx, 12(rp)
movd %mm0, %eax
pmuludq %mm7, %mm1
lea 16(rp), rp
psrlq $32, %mm0
adc %edx, %eax
movd %mm0, %edx
movd %mm1, %ebx
movd 8(up), %mm0
pmuludq %mm7, %mm0
adc $0, %edx
add %eax, (rp)
psrlq $32, %mm1
adc %edx, %ebx
movd %mm1, %edx
movd %mm0, %eax
movd 12(up), %mm1
pmuludq %mm7, %mm1
adc $0, %edx
add %ebx, 4(rp)
L(a0): psrlq $32, %mm0
adc %edx, %eax
movd %mm0, %edx
movd %mm1, %ebx
lea 16(up), up
movd (up), %mm0
adc $0, %edx
add %eax, 8(rp)
psrlq $32, %mm1
adc %edx, %ebx
movd %mm1, %edx
pmuludq %mm7, %mm0
inc un
movd 4(up), %mm1
jnz L(la0)
adc un, %edx C un is zero here
add %ebx, 12(rp)
movd %mm0, %eax
pmuludq %mm7, %mm1
lea 16(rp), rp
psrlq $32, %mm0
adc %edx, %eax
movd %mm0, %edx
movd %mm1, %ebx
adc un, %edx
add %eax, (rp)
psrlq $32, %mm1
adc %edx, %ebx
movd %mm1, %eax
adc un, %eax
add %ebx, 4(rp)
adc un, %eax
mov %eax, 8(rp)
inc n
C ================================================================
L(ol3): lea 12(up,n,4), up
movd -8(up), %mm7 C read next U invariant limb
lea (rp,n,4), rp C put rp back
mov n, un
movd -4(up), %mm1
pmuludq %mm7, %mm1
sar $2, un
movd %mm1, %ebx
movd (up), %mm0
xor %edx, %edx C zero edx and CF
jmp L(a3)
L(la3): adc $0, %edx
add %ebx, 12(rp)
movd %mm0, %eax
pmuludq %mm7, %mm1
lea 16(rp), rp
psrlq $32, %mm0
adc %edx, %eax
movd %mm0, %edx
movd %mm1, %ebx
movd 8(up), %mm0
pmuludq %mm7, %mm0
adc $0, %edx
add %eax, (rp)
psrlq $32, %mm1
adc %edx, %ebx
movd %mm1, %edx
movd %mm0, %eax
movd 12(up), %mm1
pmuludq %mm7, %mm1
adc $0, %edx
add %ebx, 4(rp)
psrlq $32, %mm0
adc %edx, %eax
movd %mm0, %edx
movd %mm1, %ebx
lea 16(up), up
movd (up), %mm0
adc $0, %edx
add %eax, 8(rp)
L(a3): psrlq $32, %mm1
adc %edx, %ebx
movd %mm1, %edx
pmuludq %mm7, %mm0
inc un
movd 4(up), %mm1
jnz L(la3)
adc un, %edx C un is zero here
add %ebx, 12(rp)
movd %mm0, %eax
pmuludq %mm7, %mm1
lea 16(rp), rp
psrlq $32, %mm0
adc %edx, %eax
movd %mm0, %edx
movd %mm1, %ebx
adc un, %edx
add %eax, (rp)
psrlq $32, %mm1
adc %edx, %ebx
movd %mm1, %eax
adc un, %eax
add %ebx, 4(rp)
adc un, %eax
mov %eax, 8(rp)
inc n
C ================================================================
L(ol2): lea 8(up,n,4), up
movd -4(up), %mm7 C read next U invariant limb
lea 12(rp,n,4), rp
mov n, un
movd (up), %mm0
pmuludq %mm7, %mm0
xor %edx, %edx
sar $2, un
movd 4(up), %mm1
test un, un C clear carry
movd %mm0, %eax
pmuludq %mm7, %mm1
inc un
jnz L(a2)
jmp L(re2)
L(la2): adc $0, %edx
add %ebx, 12(rp)
movd %mm0, %eax
pmuludq %mm7, %mm1
lea 16(rp), rp
L(a2): psrlq $32, %mm0
adc %edx, %eax
movd %mm0, %edx
movd %mm1, %ebx
movd 8(up), %mm0
pmuludq %mm7, %mm0
adc $0, %edx
add %eax, (rp)
psrlq $32, %mm1
adc %edx, %ebx
movd %mm1, %edx
movd %mm0, %eax
movd 12(up), %mm1
pmuludq %mm7, %mm1
adc $0, %edx
add %ebx, 4(rp)
psrlq $32, %mm0
adc %edx, %eax
movd %mm0, %edx
movd %mm1, %ebx
lea 16(up), up
movd (up), %mm0
adc $0, %edx
add %eax, 8(rp)
psrlq $32, %mm1
adc %edx, %ebx
movd %mm1, %edx
pmuludq %mm7, %mm0
inc un
movd 4(up), %mm1
jnz L(la2)
adc un, %edx C un is zero here
add %ebx, 12(rp)
movd %mm0, %eax
pmuludq %mm7, %mm1
lea 16(rp), rp
psrlq $32, %mm0
adc %edx, %eax
movd %mm0, %edx
movd %mm1, %ebx
adc un, %edx
add %eax, (rp)
psrlq $32, %mm1
adc %edx, %ebx
movd %mm1, %eax
adc un, %eax
add %ebx, 4(rp)
adc un, %eax
mov %eax, 8(rp)
inc n
jmp L(ol1)
C ================================================================
L(re2): psrlq $32, %mm0
movd (up), %mm7 C read next U invariant limb
adc %edx, %eax
movd %mm0, %edx
movd %mm1, %ebx
adc un, %edx
add %eax, (rp)
lea 4(rp), rp
psrlq $32, %mm1
adc %edx, %ebx
movd %mm1, %eax
movd 4(up), %mm1
adc un, %eax
add %ebx, (rp)
pmuludq %mm7, %mm1
adc un, %eax
mov %eax, 4(rp)
movd %mm1, %ebx
L(re1): psrlq $32, %mm1
add %ebx, 4(rp)
movd %mm1, %eax
adc un, %eax
xor n, n C make n zeroness assumption below true
mov %eax, 8(rp)
L(done): C n is zero here
mov 24(%esp), up
mov 28(%esp), %eax
movd (up), %mm0
inc %eax
pmuludq %mm0, %mm0
lea 4(up), up
mov 20(%esp), rp
shr %eax
movd %mm0, (rp)
psrlq $32, %mm0
lea -12(rp), rp
mov %eax, 28(%esp)
jnc L(odd)
movd %mm0, %ebp
movd (up), %mm0
lea 8(rp), rp
pmuludq %mm0, %mm0
lea -4(up), up
add 8(rp), %ebp
movd %mm0, %edx
adc 12(rp), %edx
rcr n
jmp L(ent)
C ALIGN(16) C alignment seems irrelevant
L(top): movd (up), %mm1
adc n, n
movd %mm0, %eax
pmuludq %mm1, %mm1
movd 4(up), %mm0
adc (rp), %eax
movd %mm1, %ebx
pmuludq %mm0, %mm0
psrlq $32, %mm1
adc 4(rp), %ebx
movd %mm1, %ebp
movd %mm0, %edx
adc 8(rp), %ebp
adc 12(rp), %edx
rcr n C FIXME: isn't this awfully slow on atom???
adc %eax, (rp)
adc %ebx, 4(rp)
L(ent): lea 8(up), up
adc %ebp, 8(rp)
psrlq $32, %mm0
adc %edx, 12(rp)
L(odd): decl 28(%esp)
lea 16(rp), rp
jnz L(top)
L(end): adc n, n
movd %mm0, %eax
adc n, %eax
mov %eax, (rp)
L(rtn): emms
pop %ebp
pop %ebx
pop %esi
pop %edi
ret
L(one): pmuludq %mm7, %mm7
movq %mm7, -4(rp)
emms
pop %esi
pop %edi
ret
EPILOGUE()
|
