'Merge a fast CRC32 snippet with it's CRC32 table help needed' Thread RSS Feed

Merge a fast CRC32 snippet with it's CRC32 table help needed

Thu, 01 Apr 2004 16:52:22 -0700

This is supposed to be a fast CRC 32 algorithmn. I would like some help to see if it really is faster compared to what I already have. The table is below this code. I don't know how to merge the two.

I don't recognize the enter 0,0, leave, and ret 16 statements.

Thanks.

; Copyright (c) 2002
; Pavel "EvilOne" Minayev
;
; Permission to use, copy, modify, distribute and sell this software
; and its documentation for any purpose is hereby granted without fee,
; provided that the above copyright notice appear in all copies and
; that both that copyright notice and this permission notice appear
; in supporting documentation. Author makes no representations about
; the suitability of this software for any purpose. It is provided
; "as is" without express or implied warranty.

bits 32

segment .text
start:
enter 0, 0
; load arguments into registers
mov ebx, [ebp+16] ; CRC-table
mov ecx, [ebp+12] ; string length
mov esi, [ebp+08] ; string
; initialize
mov eax, 0xFFFFFFFF
xor edx, edx
calcbyte:
; process a single byte:
; crc = table[(unsigned char)crc ^ byte] ^ (crc >> 8)
mov dl, al
xor dl, [esi]
shr eax, 8
xor eax, [ebx+edx*4]
inc esi
loop calcbyte
; clean up and return
leave
ret 16

; 32 bit reverse crc table for polynomial EDB88320h (Zmodem,PKZip)
; X^32+X^26+X^23+X^22+X^16+X^12+X^11+X^10+X^8+X^7+X^ 5+X^4+X^2+X^1+X^0

crc32tab dd 000000000h, 077073096h, 0ee0e612ch, 0990951bah, 0076dc419h
dd 0706af48fh, 0e963a535h, 09e6495a3h, 00edb8832h, 079dcb8a4h
dd 0e0d5e91eh, 097d2d988h, 009b64c2bh, 07eb17cbdh, 0e7b82d07h
dd 090bf1d91h, 01db71064h, 06ab020f2h, 0f3b97148h, 084be41deh
dd 01adad47dh, 06ddde4ebh, 0f4d4b551h, 083d385c7h, 0136c9856h
dd 0646ba8c0h, 0fd62f97ah, 08a65c9ech, 014015c4fh, 063066cd9h
dd 0fa0f3d63h, 08d080df5h, 03b6e20c8h, 04c69105eh, 0d56041e4h
dd 0a2677172h, 03c03e4d1h, 04b04d447h, 0d20d85fdh, 0a50ab56bh
dd 035b5a8fah, 042b2986ch, 0dbbbc9d6h, 0acbcf940h, 032d86ce3h
dd 045df5c75h, 0dcd60dcfh, 0abd13d59h, 026d930ach, 051de003ah
dd 0c8d75180h, 0bfd06116h, 021b4f4b5h, 056b3c423h, 0cfba9599h
dd 0b8bda50fh, 02802b89eh, 05f058808h, 0c60cd9b2h, 0b10be924h
dd 02f6f7c87h, 058684c11h, 0c1611dabh, 0b6662d3dh, 076dc4190h
dd 001db7106h, 098d220bch, 0efd5102ah, 071b18589h, 006b6b51fh
dd 09fbfe4a5h, 0e8b8d433h, 07807c9a2h, 00f00f934h, 09609a88eh
dd 0e10e9818h, 07f6a0dbbh, 0086d3d2dh, 091646c97h, 0e6635c01h
dd 06b6b51f4h, 01c6c6162h, 0856530d8h, 0f262004eh, 06c0695edh
dd 01b01a57bh, 08208f4c1h, 0f50fc457h, 065b0d9c6h, 012b7e950h
dd 08bbeb8eah, 0fcb9887ch, 062dd1ddfh, 015da2d49h, 08cd37cf3h
dd 0fbd44c65h, 04db26158h, 03ab551ceh, 0a3bc0074h, 0d4bb30e2h
dd 04adfa541h, 03dd895d7h, 0a4d1c46dh, 0d3d6f4fbh, 04369e96ah
dd 0346ed9fch, 0ad678846h, 0da60b8d0h, 044042d73h, 033031de5h
dd 0aa0a4c5fh, 0dd0d7cc9h, 05005713ch, 0270241aah, 0be0b1010h
dd 0c90c2086h, 05768b525h, 0206f85b3h, 0b966d409h, 0ce61e49fh
dd 05edef90eh, 029d9c998h, 0b0d09822h, 0c7d7a8b4h, 059b33d17h
dd 02eb40d81h, 0b7bd5c3bh, 0c0ba6cadh, 0edb88320h, 09abfb3b6h
dd 003b6e20ch, 074b1d29ah, 0ead54739h, 09dd277afh, 004db2615h
dd 073dc1683h, 0e3630b12h, 094643b84h, 00d6d6a3eh, 07a6a5aa8h
dd 0e40ecf0bh, 09309ff9dh, 00a00ae27h, 07d079eb1h, 0f00f9344h
dd 08708a3d2h, 01e01f268h, 06906c2feh, 0f762575dh, 0806567cbh
dd 0196c3671h, 06e6b06e7h, 0fed41b76h, 089d32be0h, 010da7a5ah
dd 067dd4acch, 0f9b9df6fh, 08ebeeff9h, 017b7be43h, 060b08ed5h
dd 0d6d6a3e8h, 0a1d1937eh, 038d8c2c4h, 04fdff252h, 0d1bb67f1h
dd 0a6bc5767h, 03fb506ddh, 048b2364bh, 0d80d2bdah, 0af0a1b4ch
dd 036034af6h, 041047a60h, 0df60efc3h, 0a867df55h, 0316e8eefh
dd 04669be79h, 0cb61b38ch, 0bc66831ah, 0256fd2a0h, 05268e236h
dd 0cc0c7795h, 0bb0b4703h, 0220216b9h, 05505262fh, 0c5ba3bbeh
dd 0b2bd0b28h, 02bb45a92h, 05cb36a04h, 0c2d7ffa7h, 0b5d0cf31h
dd 02cd99e8bh, 05bdeae1dh, 09b64c2b0h, 0ec63f226h, 0756aa39ch
dd 0026d930ah, 09c0906a9h, 0eb0e363fh, 072076785h, 005005713h
dd 095bf4a82h, 0e2b87a14h, 07bb12baeh, 00cb61b38h, 092d28e9bh
dd 0e5d5be0dh, 07cdcefb7h, 00bdbdf21h, 086d3d2d4h, 0f1d4e242h
dd 068ddb3f8h, 01fda836eh, 081be16cdh, 0f6b9265bh, 06fb077e1h
dd 018b74777h, 088085ae6h, 0ff0f6a70h, 066063bcah, 011010b5ch
dd 08f659effh, 0f862ae69h, 0616bffd3h, 0166ccf45h, 0a00ae278h
dd 0d70dd2eeh, 04e048354h, 03903b3c2h, 0a7672661h, 0d06016f7h
dd 04969474dh, 03e6e77dbh, 0aed16a4ah, 0d9d65adch, 040df0b66h
dd 037d83bf0h, 0a9bcae53h, 0debb9ec5h, 047b2cf7fh, 030b5ffe9h
dd 0bdbdf21ch, 0cabac28ah, 053b39330h, 024b4a3a6h, 0bad03605h
dd 0cdd70693h, 054de5729h, 023d967bfh, 0b3667a2eh, 0c4614ab8h
dd 05d681b02h, 02a6f2b94h, 0b40bbe37h, 0c30c8ea1h, 05a05df1bh
dd 02d02ef8dh

Re: Merge a fast CRC32 snippet with it's CRC32 table help needed

Fri, 02 Apr 2004 07:46:52 -0700

It is a nice code, but LOOP instruction is slower then DEC/JNZ combination.

Re: Merge a fast CRC32 snippet with it's CRC32 table help needed

Fri, 02 Apr 2004 08:06:39 -0700

: It is a nice code, but LOOP instruction is slower then DEC/JNZ combination.
:
I looked at the code and can't figure out what condition causes the loop instruction to finish.

Thanks.

Re: Merge a fast CRC32 snippet with it's CRC32 table help needed

Fri, 02 Apr 2004 10:46:14 -0700

This message was edited by AsmGuru62 at 2004-4-2 10:47:7

: : It is a nice code, but LOOP instruction is slower then DEC/JNZ combination.
: :
: I looked at the code and can't figure out what condition causes the loop instruction to finish.
:
: Thanks.
:
This one:

mov ecx, [ebp+12] ; string length

You should know that LOOP will loop until ECX becomes zero. Also, always check ECX BEFORE running that LOOP - if it is already zero! - you will have a loop counter of 100000000H - and it ain't fun! The REP instruction, on the other hand, does check for zero BEFORE looping, so:

MOV ECX, 0
REP MOVSD

this code ^^^ will not cause any trouble. There will be no single MOVSD executed. Also, as I said, the pair

AGAIN:

; body of the loop

DEC  ECX
JNZ  AGAIN

is faster then LOOP.

Re: Merge a fast CRC32 snippet with it's CRC32 table help needed

Fri, 02 Apr 2004 13:21:00 -0700

: this code ^^^ will not cause any trouble. There will be no single MOVSD executed. Also, as I said, the pair
:

: AGAIN:
: 
: ; body of the loop
: 
: DEC  ECX
: JNZ  AGAIN
:

: is faster then LOOP.[/blue]

; Copyright (c) 2002
; Pavel "EvilOne" Minayev
;
; Permission to use, copy, modify, distribute and sell this software
; and its documentation for any purpose is hereby granted without fee,
; provided that the above copyright notice appear in all copies and
; that both that copyright notice and this permission notice appear
; in supporting documentation. Author makes no representations about
; the suitability of this software for any purpose. It is provided
; "as is" without express or implied warranty.

I hope I did this right.

bits 32

segment .text
start:
enter 0, 0
; load arguments into registers
mov ebx, [ebp+16] ; CRC-table
;mov ecx, [ebp+12] ; string length

MOV ECX, 0
REP MOVSD
mov esi, [ebp+08] ; string
; initialize
mov eax, 0xFFFFFFFF
xor edx, edx
calcbyte:
; process a single byte:
; crc = table[(unsigned char)crc ^ byte] ^ (crc >> 8)

mov dl, al
xor dl, [esi]
shr eax, 8
xor eax, [ebx+edx*4]
inc esi
DEC ECX
JNZ calcbyte

; clean up and return
leave
ret 16

Re: Merge a fast CRC32 snippet with it's CRC32 table help needed

Fri, 02 Apr 2004 14:13:39 -0700

This message was edited by AsmGuru62 at 2004-4-2 14:14:58

:
: : this code ^^^ will not cause any trouble. There will be no single MOVSD executed. Also, as I said, the pair
: :

: : AGAIN:
: : 
: : ; body of the loop
: : 
: : DEC  ECX
: : JNZ  AGAIN
: :

: : is faster then LOOP.[/blue]
:
: ; Copyright (c) 2002
: ; Pavel "EvilOne" Minayev
: ;
: ; Permission to use, copy, modify, distribute and sell this software
: ; and its documentation for any purpose is hereby granted without fee,
: ; provided that the above copyright notice appear in all copies and
: ; that both that copyright notice and this permission notice appear
: ; in supporting documentation. Author makes no representations about
: ; the suitability of this software for any purpose. It is provided
: ; "as is" without express or implied warranty.
:
: I hope I did this right.
:
: bits 32
:
: segment .text
: start:
: enter 0, 0
: ; load arguments into registers
: mov ebx, [ebp+16] ; CRC-table
: ;mov ecx, [ebp+12] ; string length
:
: MOV ECX, 0
: REP MOVSD
I used it ^^^ as an example - you do not need these two lines.
In fact - these lines will make your code not working properly...

: mov esi, [ebp+08] ; string
: ; initialize
: mov eax, 0xFFFFFFFF
: xor edx, edx
: calcbyte:
: ; process a single byte:
: ; crc = table[(unsigned char)crc ^ byte] ^ (crc >> 8)
:
:
: mov dl, al
: xor dl, [esi]
: shr eax, 8
: xor eax, [ebx+edx*4]
: inc esi
: DEC ECX
: JNZ calcbyte
:
: ; clean up and return
: leave
: ret 16
:
:

Re: Merge a fast CRC32 snippet with it's CRC32 table help needed

Fri, 02 Apr 2004 15:41:49 -0700

:
: : MOV ECX, 0
: : REP MOVSD
: I used it ^^^ as an example - you do not need these two lines.
: In fact - these lines will make your code not working properly...

I'll give what I think this code is doing.

The mov ebx points to the CRC table
String length maybe is the length of the file I want to do the CRC of

Am I close.

bits 32

segment .text
start:
enter 0, 0
; load arguments into registers
mov ebx, [ebp+16] ; CRC-table
mov ecx, [ebp+12] ; string length

mov esi, [ebp+08] ; string
; initialize
mov eax, 0xFFFFFFFF
xor edx, edx
calcbyte:
; process a single byte:
; crc = table[(unsigned char)crc ^ byte] ^ (crc >> 8)

mov dl, al
xor dl, [esi]
shr eax, 8
xor eax, [ebx+edx*4]
inc esi
DEC ECX
JNZ calcbyte

; clean up and return
leave
ret 16

Re: Merge a fast CRC32 snippet with it's CRC32 table help needed

Sun, 04 Apr 2004 06:58:29 -0700

: :
: : : MOV ECX, 0
: : : REP MOVSD
: : I used it ^^^ as an example - you do not need these two lines.
: : In fact - these lines will make your code not working properly...
:
: I'll give what I think this code is doing.
:
: The mov ebx points to the CRC table
: String length maybe is the length of the file I want to do the CRC of
:
: Am I close.
:
: bits 32
:
: segment .text
: start:
: enter 0, 0
: ; load arguments into registers
: mov ebx, [ebp+16] ; CRC-table
: mov ecx, your data size goes here
:
: mov esi, address of a first byte of your data
: ; initialize
: mov eax, 0xFFFFFFFF
: xor edx, edx
: calcbyte:
: ; process a single byte:
: ; crc = table[(unsigned char)crc ^ byte] ^ (crc >> 8)
:
:
: mov dl, al
: xor dl, [esi]
: shr eax, 8
: xor eax, [ebx+edx*4]
: inc esi
: DEC ECX
: JNZ calcbyte
:
: ; clean up and return
: leave
: ret 16
:
: