-
Notifications
You must be signed in to change notification settings - Fork 11
/
arith_amd64.s
521 lines (423 loc) · 9.14 KB
/
arith_amd64.s
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
// Copyright 2009 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE_go file.
//go:build !math_big_pure_go
// +build !math_big_pure_go
#include "textflag.h"
// This file provides fast assembly versions for the elementary
// arithmetic operations on vectors implemented in arith.go.
// func mulWW(x, y Word) (z1, z0 Word)
TEXT ·mulWW(SB),NOSPLIT,$0
MOVQ x+0(FP), AX
MULQ y+8(FP)
MOVQ DX, z1+16(FP)
MOVQ AX, z0+24(FP)
RET
// The carry bit is saved with SBBQ Rx, Rx: if the carry was set, Rx is -1, otherwise it is 0.
// It is restored with ADDQ Rx, Rx: if Rx was -1 the carry is set, otherwise it is cleared.
// This is faster than using rotate instructions.
// func addVV(z, x, y []Word) (c Word)
TEXT ·addVV(SB),NOSPLIT,$0
MOVQ z_len+8(FP), DI
MOVQ x+24(FP), R8
MOVQ y+48(FP), R9
MOVQ z+0(FP), R10
MOVQ $0, CX // c = 0
MOVQ $0, SI // i = 0
// s/JL/JMP/ below to disable the unrolled loop
SUBQ $4, DI // n -= 4
JL V1 // if n < 0 goto V1
U1: // n >= 0
// regular loop body unrolled 4x
ADDQ CX, CX // restore CF
MOVQ 0(R8)(SI*8), R11
MOVQ 8(R8)(SI*8), R12
MOVQ 16(R8)(SI*8), R13
MOVQ 24(R8)(SI*8), R14
ADCQ 0(R9)(SI*8), R11
ADCQ 8(R9)(SI*8), R12
ADCQ 16(R9)(SI*8), R13
ADCQ 24(R9)(SI*8), R14
MOVQ R11, 0(R10)(SI*8)
MOVQ R12, 8(R10)(SI*8)
MOVQ R13, 16(R10)(SI*8)
MOVQ R14, 24(R10)(SI*8)
SBBQ CX, CX // save CF
ADDQ $4, SI // i += 4
SUBQ $4, DI // n -= 4
JGE U1 // if n >= 0 goto U1
V1: ADDQ $4, DI // n += 4
JLE E1 // if n <= 0 goto E1
L1: // n > 0
ADDQ CX, CX // restore CF
MOVQ 0(R8)(SI*8), R11
ADCQ 0(R9)(SI*8), R11
MOVQ R11, 0(R10)(SI*8)
SBBQ CX, CX // save CF
ADDQ $1, SI // i++
SUBQ $1, DI // n--
JG L1 // if n > 0 goto L1
E1: NEGQ CX
MOVQ CX, c+72(FP) // return c
RET
// func subVV(z, x, y []Word) (c Word)
// (same as addVV except for SBBQ instead of ADCQ and label names)
TEXT ·subVV(SB),NOSPLIT,$0
MOVQ z_len+8(FP), DI
MOVQ x+24(FP), R8
MOVQ y+48(FP), R9
MOVQ z+0(FP), R10
MOVQ $0, CX // c = 0
MOVQ $0, SI // i = 0
// s/JL/JMP/ below to disable the unrolled loop
SUBQ $4, DI // n -= 4
JL V2 // if n < 0 goto V2
U2: // n >= 0
// regular loop body unrolled 4x
ADDQ CX, CX // restore CF
MOVQ 0(R8)(SI*8), R11
MOVQ 8(R8)(SI*8), R12
MOVQ 16(R8)(SI*8), R13
MOVQ 24(R8)(SI*8), R14
SBBQ 0(R9)(SI*8), R11
SBBQ 8(R9)(SI*8), R12
SBBQ 16(R9)(SI*8), R13
SBBQ 24(R9)(SI*8), R14
MOVQ R11, 0(R10)(SI*8)
MOVQ R12, 8(R10)(SI*8)
MOVQ R13, 16(R10)(SI*8)
MOVQ R14, 24(R10)(SI*8)
SBBQ CX, CX // save CF
ADDQ $4, SI // i += 4
SUBQ $4, DI // n -= 4
JGE U2 // if n >= 0 goto U2
V2: ADDQ $4, DI // n += 4
JLE E2 // if n <= 0 goto E2
L2: // n > 0
ADDQ CX, CX // restore CF
MOVQ 0(R8)(SI*8), R11
SBBQ 0(R9)(SI*8), R11
MOVQ R11, 0(R10)(SI*8)
SBBQ CX, CX // save CF
ADDQ $1, SI // i++
SUBQ $1, DI // n--
JG L2 // if n > 0 goto L2
E2: NEGQ CX
MOVQ CX, c+72(FP) // return c
RET
// func addVW(z, x []Word, y Word) (c Word)
TEXT ·addVW(SB),NOSPLIT,$0
MOVQ z_len+8(FP), DI
MOVQ x+24(FP), R8
MOVQ y+48(FP), CX // c = y
MOVQ z+0(FP), R10
MOVQ $0, SI // i = 0
// s/JL/JMP/ below to disable the unrolled loop
SUBQ $4, DI // n -= 4
JL V3 // if n < 4 goto V3
U3: // n >= 0
// regular loop body unrolled 4x
MOVQ 0(R8)(SI*8), R11
MOVQ 8(R8)(SI*8), R12
MOVQ 16(R8)(SI*8), R13
MOVQ 24(R8)(SI*8), R14
ADDQ CX, R11
ADCQ $0, R12
ADCQ $0, R13
ADCQ $0, R14
SBBQ CX, CX // save CF
NEGQ CX
MOVQ R11, 0(R10)(SI*8)
MOVQ R12, 8(R10)(SI*8)
MOVQ R13, 16(R10)(SI*8)
MOVQ R14, 24(R10)(SI*8)
ADDQ $4, SI // i += 4
SUBQ $4, DI // n -= 4
JGE U3 // if n >= 0 goto U3
V3: ADDQ $4, DI // n += 4
JLE E3 // if n <= 0 goto E3
L3: // n > 0
ADDQ 0(R8)(SI*8), CX
MOVQ CX, 0(R10)(SI*8)
SBBQ CX, CX // save CF
NEGQ CX
ADDQ $1, SI // i++
SUBQ $1, DI // n--
JG L3 // if n > 0 goto L3
E3: MOVQ CX, c+56(FP) // return c
RET
// func subVW(z, x []Word, y Word) (c Word)
// (same as addVW except for SUBQ/SBBQ instead of ADDQ/ADCQ and label names)
TEXT ·subVW(SB),NOSPLIT,$0
MOVQ z_len+8(FP), DI
MOVQ x+24(FP), R8
MOVQ y+48(FP), CX // c = y
MOVQ z+0(FP), R10
MOVQ $0, SI // i = 0
// s/JL/JMP/ below to disable the unrolled loop
SUBQ $4, DI // n -= 4
JL V4 // if n < 4 goto V4
U4: // n >= 0
// regular loop body unrolled 4x
MOVQ 0(R8)(SI*8), R11
MOVQ 8(R8)(SI*8), R12
MOVQ 16(R8)(SI*8), R13
MOVQ 24(R8)(SI*8), R14
SUBQ CX, R11
SBBQ $0, R12
SBBQ $0, R13
SBBQ $0, R14
SBBQ CX, CX // save CF
NEGQ CX
MOVQ R11, 0(R10)(SI*8)
MOVQ R12, 8(R10)(SI*8)
MOVQ R13, 16(R10)(SI*8)
MOVQ R14, 24(R10)(SI*8)
ADDQ $4, SI // i += 4
SUBQ $4, DI // n -= 4
JGE U4 // if n >= 0 goto U4
V4: ADDQ $4, DI // n += 4
JLE E4 // if n <= 0 goto E4
L4: // n > 0
MOVQ 0(R8)(SI*8), R11
SUBQ CX, R11
MOVQ R11, 0(R10)(SI*8)
SBBQ CX, CX // save CF
NEGQ CX
ADDQ $1, SI // i++
SUBQ $1, DI // n--
JG L4 // if n > 0 goto L4
E4: MOVQ CX, c+56(FP) // return c
RET
// func shlVU(z, x []Word, s uint) (c Word)
TEXT ·shlVU(SB),NOSPLIT,$0
MOVQ z_len+8(FP), BX // i = z
SUBQ $1, BX // i--
JL X8b // i < 0 (n <= 0)
// n > 0
MOVQ z+0(FP), R10
MOVQ x+24(FP), R8
MOVQ s+48(FP), CX
MOVQ (R8)(BX*8), AX // w1 = x[n-1]
MOVQ $0, DX
SHLQ CX, AX, DX // w1>>ŝ
MOVQ DX, c+56(FP)
CMPQ BX, $0
JLE X8a // i <= 0
// i > 0
L8: MOVQ AX, DX // w = w1
MOVQ -8(R8)(BX*8), AX // w1 = x[i-1]
SHLQ CX, AX, DX // w<<s | w1>>ŝ
MOVQ DX, (R10)(BX*8) // z[i] = w<<s | w1>>ŝ
SUBQ $1, BX // i--
JG L8 // i > 0
// i <= 0
X8a: SHLQ CX, AX // w1<<s
MOVQ AX, (R10) // z[0] = w1<<s
RET
X8b: MOVQ $0, c+56(FP)
RET
// func shrVU(z, x []Word, s uint) (c Word)
TEXT ·shrVU(SB),NOSPLIT,$0
MOVQ z_len+8(FP), R11
SUBQ $1, R11 // n--
JL X9b // n < 0 (n <= 0)
// n > 0
MOVQ z+0(FP), R10
MOVQ x+24(FP), R8
MOVQ s+48(FP), CX
MOVQ (R8), AX // w1 = x[0]
MOVQ $0, DX
SHRQ CX, AX, DX // w1<<ŝ
MOVQ DX, c+56(FP)
MOVQ $0, BX // i = 0
JMP E9
// i < n-1
L9: MOVQ AX, DX // w = w1
MOVQ 8(R8)(BX*8), AX // w1 = x[i+1]
SHRQ CX, AX, DX // w>>s | w1<<ŝ
MOVQ DX, (R10)(BX*8) // z[i] = w>>s | w1<<ŝ
ADDQ $1, BX // i++
E9: CMPQ BX, R11
JL L9 // i < n-1
// i >= n-1
X9a: SHRQ CX, AX // w1>>s
MOVQ AX, (R10)(R11*8) // z[n-1] = w1>>s
RET
X9b: MOVQ $0, c+56(FP)
RET
// func mulAddVWW(z, x []Word, y, r Word) (c Word)
TEXT ·mulAddVWW(SB),NOSPLIT,$0
MOVQ z+0(FP), R10
MOVQ x+24(FP), R8
MOVQ y+48(FP), R9
MOVQ r+56(FP), CX // c = r
MOVQ z_len+8(FP), R11
MOVQ $0, BX // i = 0
CMPQ R11, $4
JL E5
U5: // i+4 <= n
// regular loop body unrolled 4x
MOVQ (0*8)(R8)(BX*8), AX
MULQ R9
ADDQ CX, AX
ADCQ $0, DX
MOVQ AX, (0*8)(R10)(BX*8)
MOVQ DX, CX
MOVQ (1*8)(R8)(BX*8), AX
MULQ R9
ADDQ CX, AX
ADCQ $0, DX
MOVQ AX, (1*8)(R10)(BX*8)
MOVQ DX, CX
MOVQ (2*8)(R8)(BX*8), AX
MULQ R9
ADDQ CX, AX
ADCQ $0, DX
MOVQ AX, (2*8)(R10)(BX*8)
MOVQ DX, CX
MOVQ (3*8)(R8)(BX*8), AX
MULQ R9
ADDQ CX, AX
ADCQ $0, DX
MOVQ AX, (3*8)(R10)(BX*8)
MOVQ DX, CX
ADDQ $4, BX // i += 4
LEAQ 4(BX), DX
CMPQ DX, R11
JLE U5
JMP E5
L5: MOVQ (R8)(BX*8), AX
MULQ R9
ADDQ CX, AX
ADCQ $0, DX
MOVQ AX, (R10)(BX*8)
MOVQ DX, CX
ADDQ $1, BX // i++
E5: CMPQ BX, R11 // i < n
JL L5
MOVQ CX, c+64(FP)
RET
// func addMulVVW(z, x []Word, y Word) (c Word)
TEXT ·addMulVVW(SB),NOSPLIT,$0
CMPB ·support_adx(SB), $1
JEQ adx
MOVQ z+0(FP), R10
MOVQ x+24(FP), R8
MOVQ y+48(FP), R9
MOVQ z_len+8(FP), R11
MOVQ $0, BX // i = 0
MOVQ $0, CX // c = 0
MOVQ R11, R12
ANDQ $-2, R12
CMPQ R11, $2
JAE A6
JMP E6
A6:
MOVQ (R8)(BX*8), AX
MULQ R9
ADDQ (R10)(BX*8), AX
ADCQ $0, DX
ADDQ CX, AX
ADCQ $0, DX
MOVQ DX, CX
MOVQ AX, (R10)(BX*8)
MOVQ (8)(R8)(BX*8), AX
MULQ R9
ADDQ (8)(R10)(BX*8), AX
ADCQ $0, DX
ADDQ CX, AX
ADCQ $0, DX
MOVQ DX, CX
MOVQ AX, (8)(R10)(BX*8)
ADDQ $2, BX
CMPQ BX, R12
JL A6
JMP E6
L6: MOVQ (R8)(BX*8), AX
MULQ R9
ADDQ CX, AX
ADCQ $0, DX
ADDQ AX, (R10)(BX*8)
ADCQ $0, DX
MOVQ DX, CX
ADDQ $1, BX // i++
E6: CMPQ BX, R11 // i < n
JL L6
MOVQ CX, c+56(FP)
RET
adx:
MOVQ z_len+8(FP), R11
MOVQ z+0(FP), R10
MOVQ x+24(FP), R8
MOVQ y+48(FP), DX
MOVQ $0, BX // i = 0
MOVQ $0, CX // carry
CMPQ R11, $8
JAE adx_loop_header
CMPQ BX, R11
JL adx_short
MOVQ CX, c+56(FP)
RET
adx_loop_header:
MOVQ R11, R13
ANDQ $-8, R13
adx_loop:
XORQ R9, R9 // unset flags
MULXQ (R8), SI, DI
ADCXQ CX,SI
ADOXQ (R10), SI
MOVQ SI,(R10)
MULXQ 8(R8), AX, CX
ADCXQ DI, AX
ADOXQ 8(R10), AX
MOVQ AX, 8(R10)
MULXQ 16(R8), SI, DI
ADCXQ CX, SI
ADOXQ 16(R10), SI
MOVQ SI, 16(R10)
MULXQ 24(R8), AX, CX
ADCXQ DI, AX
ADOXQ 24(R10), AX
MOVQ AX, 24(R10)
MULXQ 32(R8), SI, DI
ADCXQ CX, SI
ADOXQ 32(R10), SI
MOVQ SI, 32(R10)
MULXQ 40(R8), AX, CX
ADCXQ DI, AX
ADOXQ 40(R10), AX
MOVQ AX, 40(R10)
MULXQ 48(R8), SI, DI
ADCXQ CX, SI
ADOXQ 48(R10), SI
MOVQ SI, 48(R10)
MULXQ 56(R8), AX, CX
ADCXQ DI, AX
ADOXQ 56(R10), AX
MOVQ AX, 56(R10)
ADCXQ R9, CX
ADOXQ R9, CX
ADDQ $64, R8
ADDQ $64, R10
ADDQ $8, BX
CMPQ BX, R13
JL adx_loop
MOVQ z+0(FP), R10
MOVQ x+24(FP), R8
CMPQ BX, R11
JL adx_short
MOVQ CX, c+56(FP)
RET
adx_short:
MULXQ (R8)(BX*8), SI, DI
ADDQ CX, SI
ADCQ $0, DI
ADDQ SI, (R10)(BX*8)
ADCQ $0, DI
MOVQ DI, CX
ADDQ $1, BX // i++
CMPQ BX, R11
JL adx_short
MOVQ CX, c+56(FP)
RET