Re: Fastcode : function won't work with length>65535 ?

I too was searching for the fastest CRC function and found the folowing to
be faster than your function :)

have fun!

function CalculateCRC32(var ABuffer; const BufferSize:LongWord): LongWord;
asm
push esi
push edi
push ebx

mov edi,edx
mov esi,eax
xor ebx,ebx
mov eax,$ffffffff

mov ecx,edi
shr ecx,2
jecxz @Rest

@Loop:
mov edx,[esi]

mov bl,al
xor bl,dl
shr eax,8
xor eax,dword ptr [CRC32table+ebx*4]

mov bl,al
xor bl,dh
shr eax,8
xor eax,dword ptr [CRC32table+ebx*4]

shr edx,16

mov bl,al
xor bl,dl
shr eax,8
xor eax,dword ptr [CRC32table+ebx*4]

mov bl,al
xor bl,dh
shr eax,8
xor eax,dword ptr [CRC32table+ebx*4]

add esi,4

loop @Loop

@Rest:
mov ecx,edi
and ecx,3
jecxz @End

@Loop_Rest:
mov bl,al
xor bl,[esi]
shr eax,8
inc esi
xor eax,dword ptr [CRC32table+ebx*4]
loop @Loop_Rest

@End:

xor eax,$ffffffff

pop ebx
pop edi
pop esi
end;
"Nibbler" <no@spam> wrote in message news:451b8148@xxxxxxxxxxxxxxxxxxxxxxxxx
Hello everyone,

Searching for a fast CRC function I found on google groups the Aleksandr
Sharahov optimized one :

function GetShaCRC11(OldCRC: cardinal; StPtr: pointer; StLen: integer):
cardinal;
asm
jcxz @ret
push ebx
xor ebx, ebx
@loop:
movzx ebx, byte ptr [edx]
inc edx
xor bl, al
// shl ebx, 02
shr eax, 08
xor eax, dword ptr [CRCtable+ebx*4]
dec ecx
jne @loop
pop ebx
@ret:
ret
end;

It works great and faster than the previous one I was using, but it seems
to be limited to a StLen<=$FFFF

Using a length > 65535 and an OldCRC=$FFFFFFFF (to be compatible with the
zip CRC) will result in an immediate exit.

I really don't have any asm knowledge, would you please help me to make it
work with a StLen>$FFFF ?

Thank you in advance,
Nibbler.



.

Relevant Pages