Hi, Paul.

PE> So are you proud of me?  :-)

Of course! But it's the buffer move problem your solving here, not the
move until \n problem. And I'm glad to see you remembered to cld, I
assumed/forgot that in my code.

However, go back to my Test 9 and Test 10 - the lodsd/stosd loop was
slower than loading and storing eax; it may also be slower in the
rep movsd version. I think that should be tested.

PE> memcpy is designed to copy from a buffer to another buffer, for
PE> a length of whatever.  No overrun allowed.  Destination to be
PE> returned to caller.  Parameter order is dest, source, length...

You should also mention that it returns *dest (if I've read that 'C'
code properly).

PE> ; stringa.asm - string assembler functions
PE> ;
PE> ; This program written by Paul Edwards, Fidonet 3:711/934
PE> ; Released to the public domain

PE> .386p

PE> .model flat

PE> _DATA   segment dword public use32 "DATA"
PE> _DATA   ends
PE> _BSS    segment dword public use32 "BSS"
PE> _BSS    ends

AAMOI why do specify these segments here?

PE> DGROUP  group   _DATA,_BSS
PE>     assume cs:_TEXT,ds:DGROUP

PE> _TEXT segment "CODE"

PE> public memcpy
PE> memcpy proc
PE> push ebp
PE> mov ebp, esp
PE> pushf
PE> push edi
PE> push esi
PE> push ecx
PE> mov edi, [ebp+8]
PE> mov esi, [ebp+12]
PE> mov ecx, [ebp+16]
PE> cld
PE> shr ecx, 2
PE> rep movsd
PE> mov ecx, [ebp+16]
PE> and ecx, 00000003H
PE> rep movsb
PE> mov eax, [ebp+8]
PE> pop ecx
PE> pop esi
PE> pop edi
PE> popf
PE> pop ebp
PE> ret
PE> memcpy endp

PE> _TEXT ends

PE> end

Nice code.

Regards, FIM.

 * * Want the right answer?  Ask the right question!
@EOT:

---