1 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
2 ; Copyright(c) 2011-2016 Intel Corporation All rights reserved.
4 ; Redistribution and use in source and binary forms, with or without
5 ; modification, are permitted provided that the following conditions
7 ; * Redistributions of source code must retain the above copyright
8 ; notice, this list of conditions and the following disclaimer.
9 ; * Redistributions in binary form must reproduce the above copyright
10 ; notice, this list of conditions and the following disclaimer in
11 ; the documentation and/or other materials provided with the
13 ; * Neither the name of Intel Corporation nor the names of its
14 ; contributors may be used to endorse or promote products derived
15 ; from this software without specific prior written permission.
17 ; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
18 ; "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
19 ; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
20 ; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
21 ; OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
22 ; SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
23 ; LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24 ; DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25 ; THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26 ; (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
27 ; OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
31 ; uint64_t crc64_ecma_norm_by8(
32 ; uint64_t init_crc, //initial CRC value, 64 bits
33 ; const unsigned char *buf, //buffer pointer to calculate CRC on
34 ; uint64_t len //buffer length in bytes (64-bit data)
37 ; yasm -f x64 -f elf64 -X gnu -g dwarf2 crc64_ecma_norm_by8
38 %include "reg_sizes.asm"
40 %define fetch_dist 1024
47 %ifidn __OUTPUT_FORMAT__, win64
58 %ifidn __OUTPUT_FORMAT__, win64
60 %define VARIABLE_OFFSET 16*10+8
62 %define VARIABLE_OFFSET 16*2+8
65 global crc64_ecma_norm_by8:function
70 sub rsp,VARIABLE_OFFSET
72 %ifidn __OUTPUT_FORMAT__, win64
73 ; push the xmm registers into the stack to maintain
74 movdqa [rsp + XMM_SAVE + 16*0], xmm6
75 movdqa [rsp + XMM_SAVE + 16*1], xmm7
76 movdqa [rsp + XMM_SAVE + 16*2], xmm8
77 movdqa [rsp + XMM_SAVE + 16*3], xmm9
78 movdqa [rsp + XMM_SAVE + 16*4], xmm10
79 movdqa [rsp + XMM_SAVE + 16*5], xmm11
80 movdqa [rsp + XMM_SAVE + 16*6], xmm12
81 movdqa [rsp + XMM_SAVE + 16*7], xmm13
85 ; check if smaller than 256
88 ; for sizes less than 256, we can't fold 128B at a time...
92 ; load the initial crc value
93 movq xmm10, arg1 ; initial crc
95 ; crc value does not need to be byte-reflected, but it needs to be moved to the high part of the register.
96 ; because data will be byte-reflected and will align with initial crc at correct place.
99 movdqa xmm11, [SHUF_MASK]
100 ; receive the initial 128B data, xor the initial crc value
101 movdqu xmm0, [arg2+16*0]
102 movdqu xmm1, [arg2+16*1]
103 movdqu xmm2, [arg2+16*2]
104 movdqu xmm3, [arg2+16*3]
105 movdqu xmm4, [arg2+16*4]
106 movdqu xmm5, [arg2+16*5]
107 movdqu xmm6, [arg2+16*6]
108 movdqu xmm7, [arg2+16*7]
111 ; XOR the initial_crc value
121 movdqa xmm10, [rk3] ;xmm10 has rk3 and rk4
122 ;imm value of pclmulqdq instruction will determine which constant to use
123 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
124 ; we subtract 256 instead of 128 to save one instruction from the loop
127 ; at this section of the code, there is 128*x+y (0<=y<128) bytes of buffer. The _fold_128_B_loop
128 ; loop will fold 128B at a time until we have 128+y Bytes of buffer
131 ; fold 128B at a time. This section of the code folds 8 xmm registers in parallel
134 ; update the buffer pointer
135 add arg2, 128 ; buf += 128;
137 prefetchnta [arg2+fetch_dist+0]
138 movdqu xmm9, [arg2+16*0]
139 movdqu xmm12, [arg2+16*1]
144 pclmulqdq xmm0, xmm10, 0x0
145 pclmulqdq xmm8, xmm10 , 0x11
146 pclmulqdq xmm1, xmm10, 0x0
147 pclmulqdq xmm13, xmm10 , 0x11
153 prefetchnta [arg2+fetch_dist+32]
154 movdqu xmm9, [arg2+16*2]
155 movdqu xmm12, [arg2+16*3]
160 pclmulqdq xmm2, xmm10, 0x0
161 pclmulqdq xmm8, xmm10 , 0x11
162 pclmulqdq xmm3, xmm10, 0x0
163 pclmulqdq xmm13, xmm10 , 0x11
169 prefetchnta [arg2+fetch_dist+64]
170 movdqu xmm9, [arg2+16*4]
171 movdqu xmm12, [arg2+16*5]
176 pclmulqdq xmm4, xmm10, 0x0
177 pclmulqdq xmm8, xmm10 , 0x11
178 pclmulqdq xmm5, xmm10, 0x0
179 pclmulqdq xmm13, xmm10 , 0x11
185 prefetchnta [arg2+fetch_dist+96]
186 movdqu xmm9, [arg2+16*6]
187 movdqu xmm12, [arg2+16*7]
192 pclmulqdq xmm6, xmm10, 0x0
193 pclmulqdq xmm8, xmm10 , 0x11
194 pclmulqdq xmm7, xmm10, 0x0
195 pclmulqdq xmm13, xmm10 , 0x11
203 ; check if there is another 128B in the buffer to be able to fold
205 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
208 ; at this point, the buffer pointer is pointing at the last y Bytes of the buffer, where 0 <= y < 128
209 ; the 128B of folded data is in 8 of the xmm registers: xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7
212 ; fold the 8 xmm registers to 1 xmm register with different constants
216 pclmulqdq xmm0, xmm10, 0x11
217 pclmulqdq xmm8, xmm10, 0x0
223 pclmulqdq xmm1, xmm10, 0x11
224 pclmulqdq xmm8, xmm10, 0x0
230 pclmulqdq xmm2, xmm10, 0x11
231 pclmulqdq xmm8, xmm10, 0x0
237 pclmulqdq xmm3, xmm10, 0x11
238 pclmulqdq xmm8, xmm10, 0x0
244 pclmulqdq xmm4, xmm10, 0x11
245 pclmulqdq xmm8, xmm10, 0x0
251 pclmulqdq xmm5, xmm10, 0x11
252 pclmulqdq xmm8, xmm10, 0x0
256 movdqa xmm10, [rk1] ;xmm10 has rk1 and rk2
259 pclmulqdq xmm6, xmm10, 0x11
260 pclmulqdq xmm8, xmm10, 0x0
265 ; instead of 128, we add 112 to the loop counter to save 1 instruction from the loop
266 ; instead of a cmp instruction, we use the negative flag with the jl instruction
268 jl _final_reduction_for_128
270 ; now we have 16+y bytes left to reduce. 16 Bytes is in register xmm7 and the rest is in memory
271 ; we can fold 16 bytes at a time if y>=16
272 ; continue folding 16B at a time
276 pclmulqdq xmm7, xmm10, 0x11
277 pclmulqdq xmm8, xmm10, 0x0
284 ; instead of a cmp instruction, we utilize the flags with the jge instruction
285 ; equivalent of: cmp arg3, 16-16
286 ; check if there is any more 16B in the buffer to be able to fold
287 jge _16B_reduction_loop
289 ;now we have 16+z bytes left to reduce, where 0<= z < 16.
290 ;first, we reduce the data in the xmm7 register
293 _final_reduction_for_128:
294 ; check if any more data to fold. If not, compute the CRC of the final 128 bits
298 ; here we are getting data that is less than 16 bytes.
299 ; 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.
300 ; after that the registers need to be adjusted.
304 movdqu xmm1, [arg2 - 16 + arg3]
307 ; get rid of the extra data that was loaded before
308 ; load the shift constant
309 lea rax, [pshufb_shf_table + 16]
313 ; shift xmm2 to the left by arg3 bytes
316 ; shift xmm7 to the right by 16-arg3 bytes
319 pblendvb xmm1, xmm2 ;xmm0 is implicit
324 pclmulqdq xmm7, xmm10, 0x11
325 pclmulqdq xmm8, xmm10, 0x0
330 ; compute crc of a 128-bit value
331 movdqa xmm10, [rk5] ; rk5 and rk6 in xmm10
335 pclmulqdq xmm7, xmm10, 0x01 ; H*L
341 movdqa xmm10, [rk7] ; rk7 and rk8 in xmm10
346 pclmulqdq xmm7, xmm10, 0x01
349 pclmulqdq xmm7, xmm10, 0x11
355 %ifidn __OUTPUT_FORMAT__, win64
356 movdqa xmm6, [rsp + XMM_SAVE + 16*0]
357 movdqa xmm7, [rsp + XMM_SAVE + 16*1]
358 movdqa xmm8, [rsp + XMM_SAVE + 16*2]
359 movdqa xmm9, [rsp + XMM_SAVE + 16*3]
360 movdqa xmm10, [rsp + XMM_SAVE + 16*4]
361 movdqa xmm11, [rsp + XMM_SAVE + 16*5]
362 movdqa xmm12, [rsp + XMM_SAVE + 16*6]
363 movdqa xmm13, [rsp + XMM_SAVE + 16*7]
365 add rsp, VARIABLE_OFFSET
368 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
369 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
370 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
371 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
376 ; check if there is enough buffer to be able to fold 16B at a time
379 movdqa xmm11, [SHUF_MASK]
381 ; if there is, load the constants
382 movdqa xmm10, [rk1] ; rk1 and rk2 in xmm10
384 movq xmm0, arg1 ; get the initial crc value
385 pslldq xmm0, 8 ; align it to its correct place
386 movdqu xmm7, [arg2] ; load the plaintext
387 pshufb xmm7, xmm11 ; byte-reflect the plaintext
391 ; update the buffer pointer
394 ; update the counter. subtract 32 instead of 16 to save one instruction from the loop
397 jmp _16B_reduction_loop
400 ; mov initial crc to the return value. this is necessary for zero-length buffers.
405 movdqa xmm11, [SHUF_MASK]
407 movq xmm0, arg1 ; get the initial crc value
408 pslldq xmm0, 8 ; align it to its correct place
412 jl _less_than_16_left
414 movdqu xmm7, [arg2] ; load the plaintext
415 pshufb xmm7, xmm11 ; byte-reflect the plaintext
416 pxor xmm7, xmm0 ; xor the initial crc value
419 movdqa xmm10, [rk1] ; rk1 and rk2 in xmm10
420 jmp _get_last_two_xmms
423 ; use stack space to load data less than 16 bytes, zero-out the 16B in memory first.
428 ; backup the counter value
471 pxor xmm7, xmm0 ; xor the initial crc value
474 lea rax, [pshufb_shf_table + 16]
488 ; Right shift (8-length) bytes in XMM
498 pxor xmm7, xmm0 ; xor the initial crc value
504 ; precomputed constants
510 DQ 0x4eb938a7d257740e
514 DQ 0x001067e571d7d5c2
518 DQ 0x0000000000000000
520 DQ 0x578d29d06cc4f872
522 DQ 0x42f0e1eba9ea3693
524 DQ 0xe464f4df5fb60ac1
526 DQ 0xb649c5b35a759cf2
528 DQ 0x9af04e1eff82d0dd
530 DQ 0x6e82e609297f8fe8
536 DQ 0x5f6843ca540df020
538 DQ 0xddf4b6981205b83f
540 DQ 0x54819d8713758b2c
542 DQ 0x4a6b90073eb0af5a
544 DQ 0x571bee0a227ef92b
546 DQ 0x44bef2a201b5200c
550 dq 0x8080808080808080, 0x8080808080808080
552 dq 0xFFFFFFFFFFFFFFFF, 0x00000000FFFFFFFF
554 dq 0x0000000000000000, 0xFFFFFFFFFFFFFFFF
557 dq 0x08090A0B0C0D0E0F, 0x0001020304050607
560 ; use these values for shift constants for the pshufb instruction
561 ; different alignments result in values as shown:
562 ; dq 0x8887868584838281, 0x008f8e8d8c8b8a89 ; shl 15 (16-1) / shr1
563 ; dq 0x8988878685848382, 0x01008f8e8d8c8b8a ; shl 14 (16-3) / shr2
564 ; dq 0x8a89888786858483, 0x0201008f8e8d8c8b ; shl 13 (16-4) / shr3
565 ; dq 0x8b8a898887868584, 0x030201008f8e8d8c ; shl 12 (16-4) / shr4
566 ; dq 0x8c8b8a8988878685, 0x04030201008f8e8d ; shl 11 (16-5) / shr5
567 ; dq 0x8d8c8b8a89888786, 0x0504030201008f8e ; shl 10 (16-6) / shr6
568 ; dq 0x8e8d8c8b8a898887, 0x060504030201008f ; shl 9 (16-7) / shr7
569 ; dq 0x8f8e8d8c8b8a8988, 0x0706050403020100 ; shl 8 (16-8) / shr8
570 ; dq 0x008f8e8d8c8b8a89, 0x0807060504030201 ; shl 7 (16-9) / shr9
571 ; dq 0x01008f8e8d8c8b8a, 0x0908070605040302 ; shl 6 (16-10) / shr10
572 ; dq 0x0201008f8e8d8c8b, 0x0a09080706050403 ; shl 5 (16-11) / shr11
573 ; dq 0x030201008f8e8d8c, 0x0b0a090807060504 ; shl 4 (16-12) / shr12
574 ; dq 0x04030201008f8e8d, 0x0c0b0a0908070605 ; shl 3 (16-13) / shr13
575 ; dq 0x0504030201008f8e, 0x0d0c0b0a09080706 ; shl 2 (16-14) / shr14
576 ; dq 0x060504030201008f, 0x0e0d0c0b0a090807 ; shl 1 (16-15) / shr15
577 dq 0x8786858483828100, 0x8f8e8d8c8b8a8988
578 dq 0x0706050403020100, 0x0f0e0d0c0b0a0908
579 dq 0x8080808080808080, 0x0f0e0d0c0b0a0908
580 dq 0x8080808080808080, 0x8080808080808080
582 ;;; func core, ver, snum
583 slversion crc64_ecma_norm_by8, 01, 00, 001a