memcpy() vs for()

Originally Posted by mmbob

memcpy is (at least with Visual Studio) an optimized assembly function. Hand written assembly will nearly always be faster than C++, unless the assembly coder has no idea what he is doing. I believe memcpy copies in DWORD chunks also, which would make it much faster than a for loop.

Nigga knows shit.

Then may I suggest inline assembly using the "mov" assembly statement

Originally Posted by Departure

Then may I suggest inline assembly using the "mov" assembly statement

you can only copy/write 1, 2 or 4 bytes with mov.

Code:

mov byte ptr [address], al
mov word ptr [address], ax
mov dword ptr [address], eax

Yes true Gordon, if you do a straight mov... Look at example below to explain what I mean using REP MOVSB

Code:

procedure CopyMemory(Destination, Source:Pointer; dwSize:LongWord);
asm
  PUSH ECX
  PUSH ESI
  PUSH EDI
  MOV EDI, Destination
  MOV ESI, Source
  MOV ECX, dwSize
  REP MOVSB
  POP EDI
  POP ESI
  POP ECX
end;

All we do here is use MOVSB which will mov bytes from address in ESI to address in EDI, this will repeat using the REP statement Untill ECX = 0 allowing us to use any amount of bytes.

Reference
http://faydoc.tripod.com/cpu/movsb.htm

Post a Reply

Tags for this Thread