forked from intel/isa-l
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathcrc32_iscsi_by16_10.asm
513 lines (420 loc) · 14.9 KB
/
crc32_iscsi_by16_10.asm
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
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Copyright(c) 2011-2020 Intel Corporation All rights reserved.
;
; Redistribution and use in source and binary forms, with or without
; modification, are permitted provided that the following conditions
; are met:
; * Redistributions of source code must retain the above copyright
; notice, this list of conditions and the following disclaimer.
; * Redistributions in binary form must reproduce the above copyright
; notice, this list of conditions and the following disclaimer in
; the documentation and/or other materials provided with the
; distribution.
; * Neither the name of Intel Corporation nor the names of its
; contributors may be used to endorse or promote products derived
; from this software without specific prior written permission.
;
; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; Function API:
; UINT32 crc32_iscsi_by16_10(
; UINT32 init_crc, //initial CRC value, 32 bits
; const unsigned char *buf, //buffer pointer to calculate CRC on
; UINT64 len //buffer length in bytes (64-bit data)
; );
;
; Authors:
; Erdinc Ozturk
; Vinodh Gopal
; James Guilford
;
; Reference paper titled "Fast CRC Computation for Generic Polynomials Using PCLMULQDQ Instruction"
; URL: http://www.intel.com/content/dam/www/public/us/en/documents/white-papers/fast-crc-computation-generic-polynomials-pclmulqdq-paper.pdf
;
;
%include "reg_sizes.asm"
%ifndef FUNCTION_NAME
%define FUNCTION_NAME crc32_iscsi_by16_10
%endif
%if (AS_FEATURE_LEVEL) >= 10
[bits 64]
default rel
section .text
%ifidn __OUTPUT_FORMAT__, win64
%xdefine arg1 r8
%xdefine arg2 rcx
%xdefine arg3 rdx
%xdefine arg1_low32 r8d
%else
%xdefine arg1 rdx
%xdefine arg2 rdi
%xdefine arg3 rsi
%xdefine arg1_low32 edx
%endif
align 16
mk_global FUNCTION_NAME, function
FUNCTION_NAME:
endbranch
%ifidn __OUTPUT_FORMAT__, win64
sub rsp, (16*10 + 8)
; push the xmm registers into the stack to maintain
vmovdqa [rsp + 16*0], xmm6
vmovdqa [rsp + 16*1], xmm7
vmovdqa [rsp + 16*2], xmm8
vmovdqa [rsp + 16*3], xmm9
vmovdqa [rsp + 16*4], xmm10
vmovdqa [rsp + 16*5], xmm11
vmovdqa [rsp + 16*6], xmm12
vmovdqa [rsp + 16*7], xmm13
vmovdqa [rsp + 16*8], xmm14
vmovdqa [rsp + 16*9], xmm15
%endif
; check if smaller than 256B
cmp arg3, 256
jl .less_than_256
; load the initial crc value
vmovd xmm10, arg1_low32 ; initial crc
; receive the initial 64B data, xor the initial crc value
vmovdqu8 zmm0, [arg2+16*0]
vmovdqu8 zmm4, [arg2+16*4]
vpxorq zmm0, zmm10
vbroadcasti32x4 zmm10, [rk3] ;xmm10 has rk3 and rk4
;imm value of pclmulqdq instruction will determine which constant to use
sub arg3, 256
cmp arg3, 256
jl .fold_128_B_loop
vmovdqu8 zmm7, [arg2+16*8]
vmovdqu8 zmm8, [arg2+16*12]
vbroadcasti32x4 zmm16, [rk_1] ;zmm16 has rk-1 and rk-2
sub arg3, 256
align 16
.fold_256_B_loop:
add arg2, 256
vpclmulqdq zmm1, zmm0, zmm16, 0x10
vpclmulqdq zmm0, zmm0, zmm16, 0x01
vpternlogq zmm0, zmm1, [arg2+16*0], 0x96
vpclmulqdq zmm2, zmm4, zmm16, 0x10
vpclmulqdq zmm4, zmm4, zmm16, 0x01
vpternlogq zmm4, zmm2, [arg2+16*4], 0x96
vpclmulqdq zmm3, zmm7, zmm16, 0x10
vpclmulqdq zmm7, zmm7, zmm16, 0x01
vpternlogq zmm7, zmm3, [arg2+16*8], 0x96
vpclmulqdq zmm5, zmm8, zmm16, 0x10
vpclmulqdq zmm8, zmm8, zmm16, 0x01
vpternlogq zmm8, zmm5, [arg2+16*12], 0x96
sub arg3, 256
jge .fold_256_B_loop
;; Fold 256 into 128
add arg2, 256
vpclmulqdq zmm1, zmm0, zmm10, 0x01
vpclmulqdq zmm2, zmm0, zmm10, 0x10
vpternlogq zmm7, zmm1, zmm2, 0x96 ; xor ABC
vpclmulqdq zmm5, zmm4, zmm10, 0x01
vpclmulqdq zmm6, zmm4, zmm10, 0x10
vpternlogq zmm8, zmm5, zmm6, 0x96 ; xor ABC
vmovdqa32 zmm0, zmm7
vmovdqa32 zmm4, zmm8
add arg3, 128
jmp .less_than_128_B
; at this section of the code, there is 128*x+y (0<=y<128) bytes of buffer. The fold_128_B_loop
; loop will fold 128B at a time until we have 128+y Bytes of buffer
; fold 128B at a time. This section of the code folds 8 xmm registers in parallel
align 16
.fold_128_B_loop:
add arg2, 128
vpclmulqdq zmm2, zmm0, zmm10, 0x10
vpclmulqdq zmm0, zmm0, zmm10, 0x01
vpternlogq zmm0, zmm2, [arg2+16*0], 0x96
vpclmulqdq zmm5, zmm4, zmm10, 0x10
vpclmulqdq zmm4, zmm4, zmm10, 0x01
vpternlogq zmm4, zmm5, [arg2+16*4], 0x96
sub arg3, 128
jge .fold_128_B_loop
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
add arg2, 128
align 16
.less_than_128_B:
;; At this point, the buffer pointer is pointing at the last
;; y bytes of the buffer, where 0 <= y < 128.
;; The 128 bytes of folded data is in 2 of the zmm registers:
;; zmm0 and zmm4
cmp arg3, -64
jl .fold_128_B_register
vbroadcasti32x4 zmm10, [rk15]
;; If there are still 64 bytes left, folds from 128 bytes to 64 bytes
;; and handles the next 64 bytes
vpclmulqdq zmm2, zmm0, zmm10, 0x10
vpclmulqdq zmm0, zmm0, zmm10, 0x01
vpternlogq zmm0, zmm2, zmm4, 0x96
add arg3, 128
jmp .fold_64B_loop
align 16
.fold_128_B_register:
; fold the 8 128b parts into 1 xmm register with different constants
vmovdqu8 zmm16, [rk9] ; multiply by rk9-rk16
vmovdqu8 zmm11, [rk17] ; multiply by rk17-rk20, rk1,rk2, 0,0
vpclmulqdq zmm1, zmm0, zmm16, 0x01
vpclmulqdq zmm2, zmm0, zmm16, 0x10
vextracti64x2 xmm7, zmm4, 3 ; save last that has no multiplicand
vpclmulqdq zmm5, zmm4, zmm11, 0x01
vpclmulqdq zmm6, zmm4, zmm11, 0x10
vmovdqa xmm10, [rk1] ; Needed later in reduction loop
vpternlogq zmm1, zmm2, zmm5, 0x96 ; xor ABC
vpternlogq zmm1, zmm6, zmm7, 0x96 ; xor ABC
vshufi64x2 zmm8, zmm1, zmm1, 0x4e ; Swap 1,0,3,2 - 01 00 11 10
vpxorq ymm8, ymm8, ymm1
vextracti64x2 xmm5, ymm8, 1
vpxorq xmm7, xmm5, xmm8
; instead of 128, we add 128-16 to the loop counter to save 1 instruction from the loop
; instead of a cmp instruction, we use the negative flag with the jl instruction
add arg3, 128-16
jl .final_reduction_for_128
; now we have 16+y bytes left to reduce. 16 Bytes is in register xmm7 and the rest is in memory
; we can fold 16 bytes at a time if y>=16
; continue folding 16B at a time
align 16
.16B_reduction_loop:
vpclmulqdq xmm8, xmm7, xmm10, 0x1
vpclmulqdq xmm7, xmm7, xmm10, 0x10
vpternlogq xmm7, xmm8, [arg2], 0x96
add arg2, 16
sub arg3, 16
; instead of a cmp instruction, we utilize the flags with the jge instruction
; equivalent of: cmp arg3, 16-16
; check if there is any more 16B in the buffer to be able to fold
jge .16B_reduction_loop
;now we have 16+z bytes left to reduce, where 0<= z < 16.
;first, we reduce the data in the xmm7 register
align 16
.final_reduction_for_128:
add arg3, 16
je .128_done
; here we are getting data that is less than 16 bytes.
; since we know that there was data before the pointer, we can offset
; the input pointer before the actual point, to receive exactly 16 bytes.
; after that the registers need to be adjusted.
align 16
.get_last_two_xmms:
vmovdqa xmm2, xmm7
vmovdqu xmm1, [arg2 - 16 + arg3]
; get rid of the extra data that was loaded before
; load the shift constant
lea rax, [rel pshufb_shf_table]
add rax, arg3
vmovdqu xmm0, [rax]
vpshufb xmm7, xmm0
vpxor xmm0, [mask3]
vpshufb xmm2, xmm0
vpblendvb xmm2, xmm2, xmm1, xmm0
;;;;;;;;;;
vpclmulqdq xmm8, xmm7, xmm10, 0x1
vpclmulqdq xmm7, xmm7, xmm10, 0x10
vpternlogq xmm7, xmm8, xmm2, 0x96
align 16
.128_done:
; compute crc of a 128-bit value
xor rax, rax
vmovq r11, xmm7
crc32 rax, r11
vpextrq r11, xmm7, 1
crc32 rax, r11
align 16
.cleanup:
%ifidn __OUTPUT_FORMAT__, win64
vmovdqa xmm6, [rsp + 16*0]
vmovdqa xmm7, [rsp + 16*1]
vmovdqa xmm8, [rsp + 16*2]
vmovdqa xmm9, [rsp + 16*3]
vmovdqa xmm10, [rsp + 16*4]
vmovdqa xmm11, [rsp + 16*5]
vmovdqa xmm12, [rsp + 16*6]
vmovdqa xmm13, [rsp + 16*7]
vmovdqa xmm14, [rsp + 16*8]
vmovdqa xmm15, [rsp + 16*9]
add rsp, (16*10 + 8)
%endif
ret
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
align 16
.less_than_256:
; check if there is enough buffer to be able to fold 16B at a time
cmp arg3, 32
jl .less_than_32
vmovd xmm1, arg1_low32 ; get the initial crc value
cmp arg3, 64
jl .less_than_64
;; receive the initial 64B data, xor the initial crc value
vmovdqu8 zmm0, [arg2]
vpxorq zmm0, zmm1
add arg2, 64
sub arg3, 64
cmp arg3, 64
jb .reduce_64B
vbroadcasti32x4 zmm10, [rk15]
align 16
.fold_64B_loop:
vmovdqu8 zmm4, [arg2]
vpclmulqdq zmm2, zmm0, zmm10, 0x10
vpclmulqdq zmm0, zmm0, zmm10, 0x01
vpternlogq zmm0, zmm2, zmm4, 0x96
add arg2, 64
sub arg3, 64
cmp arg3, 64
jge .fold_64B_loop
align 16
.reduce_64B:
; Reduce from 64 bytes to 16 bytes
vmovdqu8 zmm11, [rk17]
vpclmulqdq zmm1, zmm0, zmm11, 0x01
vpclmulqdq zmm2, zmm0, zmm11, 0x10
vextracti64x2 xmm7, zmm0, 3 ; save last that has no multiplicand
vpternlogq zmm1, zmm2, zmm7, 0x96
vmovdqa xmm10, [rk_1b] ; Needed later in reduction loop
vshufi64x2 zmm8, zmm1, zmm1, 0x4e ; Swap 1,0,3,2 - 01 00 11 10
vpxorq ymm8, ymm8, ymm1
vextracti64x2 xmm5, ymm8, 1
vpxorq xmm7, xmm5, xmm8
sub arg3, 16
jns .16B_reduction_loop ; At least 16 bytes of data to digest
jmp .final_reduction_for_128
align 16
.less_than_64:
;; if there is, load the constants
vmovdqa xmm10, [rk_1b]
vmovdqu xmm7, [arg2] ; load the plaintext
vpxor xmm7, xmm1 ; xmm1 already has initial crc value
;; update the buffer pointer
add arg2, 16
;; update the counter
;; - subtract 32 instead of 16 to save one instruction from the loop
sub arg3, 32
jmp .16B_reduction_loop
align 16
.less_than_32:
; mov initial crc to the return value. this is necessary for zero-length buffers.
mov eax, arg1_low32
test arg3, arg3
je .cleanup
vmovd xmm0, arg1_low32 ; get the initial crc value
cmp arg3, 16
je .exact_16_left
jl .less_than_16_left
vmovdqu xmm7, [arg2] ; load the plaintext
vpxor xmm7, xmm0 ; xor the initial crc value
add arg2, 16
sub arg3, 16
vmovdqa xmm10, [rk1] ; rk1 and rk2 in xmm10
jmp .get_last_two_xmms
align 16
.less_than_16_left:
cmp arg3, 4
jl .only_less_than_4
xor r10, r10
bts r10, arg3
dec r10
kmovw k2, r10d
vmovdqu8 xmm7{k2}{z}, [arg2]
vpxor xmm7, xmm0 ; xor the initial crc value
lea rax, [rel pshufb_shf_table]
vmovdqu xmm0, [rax + arg3]
vpshufb xmm7,xmm0
jmp .128_done
align 16
.exact_16_left:
vmovdqu xmm7, [arg2]
vpxor xmm7, xmm0 ; xor the initial crc value
jmp .128_done
align 16
.only_less_than_4:
mov eax, arg1_low32
cmp arg3, 2
jb .only_1_left
je .only_2_left
; 3 bytes left
crc32 eax, word [arg2]
crc32 eax, byte [arg2 + 2]
jmp .cleanup
align 16
.only_2_left:
crc32 eax, word [arg2]
jmp .cleanup
align 16
.only_1_left:
crc32 eax, byte [arg2]
jmp .cleanup
section .data
align 32
%ifndef USE_CONSTS
; precomputed constants
rk_1: dq 0x00000000b9e02b86
rk_2: dq 0x00000000dcb17aa4
rk1: dq 0x00000000493c7d27
rk2: dq 0x0000000ec1068c50
rk3: dq 0x0000000206e38d70
rk4: dq 0x000000006992cea2
rk5: dq 0x00000000493c7d27
rk6: dq 0x00000000dd45aab8
rk7: dq 0x00000000dea713f0
rk8: dq 0x0000000105ec76f0
rk9: dq 0x0000000047db8317
rk10: dq 0x000000002ad91c30
rk11: dq 0x000000000715ce53
rk12: dq 0x00000000c49f4f67
rk13: dq 0x0000000039d3b296
rk14: dq 0x00000000083a6eec
rk15: dq 0x000000009e4addf8
rk16: dq 0x00000000740eef02
rk17: dq 0x00000000ddc0152b
rk18: dq 0x000000001c291d04
rk19: dq 0x00000000ba4fc28e
rk20: dq 0x000000003da6d0cb
rk_1b: dq 0x00000000493c7d27
rk_2b: dq 0x0000000ec1068c50
dq 0x0000000000000000
dq 0x0000000000000000
%else
INCLUDE_CONSTS
%endif
pshufb_shf_table:
; use these values for shift constants for the pshufb instruction
; different alignments result in values as shown:
; dq 0x8887868584838281, 0x008f8e8d8c8b8a89 ; shl 15 (16-1) / shr1
; dq 0x8988878685848382, 0x01008f8e8d8c8b8a ; shl 14 (16-3) / shr2
; dq 0x8a89888786858483, 0x0201008f8e8d8c8b ; shl 13 (16-4) / shr3
; dq 0x8b8a898887868584, 0x030201008f8e8d8c ; shl 12 (16-4) / shr4
; dq 0x8c8b8a8988878685, 0x04030201008f8e8d ; shl 11 (16-5) / shr5
; dq 0x8d8c8b8a89888786, 0x0504030201008f8e ; shl 10 (16-6) / shr6
; dq 0x8e8d8c8b8a898887, 0x060504030201008f ; shl 9 (16-7) / shr7
; dq 0x8f8e8d8c8b8a8988, 0x0706050403020100 ; shl 8 (16-8) / shr8
; dq 0x008f8e8d8c8b8a89, 0x0807060504030201 ; shl 7 (16-9) / shr9
; dq 0x01008f8e8d8c8b8a, 0x0908070605040302 ; shl 6 (16-10) / shr10
; dq 0x0201008f8e8d8c8b, 0x0a09080706050403 ; shl 5 (16-11) / shr11
; dq 0x030201008f8e8d8c, 0x0b0a090807060504 ; shl 4 (16-12) / shr12
; dq 0x04030201008f8e8d, 0x0c0b0a0908070605 ; shl 3 (16-13) / shr13
; dq 0x0504030201008f8e, 0x0d0c0b0a09080706 ; shl 2 (16-14) / shr14
; dq 0x060504030201008f, 0x0e0d0c0b0a090807 ; shl 1 (16-15) / shr15
dq 0x8786858483828100, 0x8f8e8d8c8b8a8988
dq 0x0706050403020100, 0x000e0d0c0b0a0908
mask: dq 0xFFFFFFFFFFFFFFFF, 0x0000000000000000
mask2: dq 0xFFFFFFFF00000000, 0xFFFFFFFFFFFFFFFF
mask3: dq 0x8080808080808080, 0x8080808080808080
%else ; Assembler doesn't understand these opcodes. Add empty symbol for windows.
%ifidn __OUTPUT_FORMAT__, win64
global no_ %+ FUNCTION_NAME
no_ %+ FUNCTION_NAME %+ :
%endif
%endif ; (AS_FEATURE_LEVEL) >= 10