home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.os.msdos.programmer
- Path: sparky!uunet!caen!sol.ctr.columbia.edu!csa!kibirev
- From: kibirev@csa.bu.edu (oleg kibirev)
- Subject: Re: fast memcpy
- Organization: Computer Science Department, Boston University, Boston, MA, USA
- References: <2a66baaf@ralf> <1992Jul19.234636.25680@ccu1.aukuni.ac.nz>
- Message-ID: <1992Jul21.162618.25425@ctr.columbia.edu>
- Sender: news@ctr.columbia.edu (The Daily Lose)
- Date: Tue, 21 Jul 1992 16:26:18 GMT
- X-Posted-From: csa.bu.edu
- X-Posted-Through: sol.ctr.columbia.edu
- Lines: 17
-
- In article <1992Jul19.234636.25680@ccu1.aukuni.ac.nz> stuartw@ccu1.aukuni.ac.nz (Stuart Woolford) writes:
- >
- >in microsoft ( 5.0 and 6.0, atleast... ) a memcpy is done with an inline
- >string move instruction, ie:no call overhead, short of using DMA, there
- >is no way of going faster ( I think it uses word move, but could be wrong.. )
- >
- So, each time you use memcpy, optimization is turned off for the entire module?
- Or it's only Borland's reacrion to ASM? I think there is a way to do it faster
- (sometimes): checking if the start of both source and dest are misaligned and
- fixing it with single movsb before doing word move.
-
- DMA won't be faster on anything better than 4.77MHz PC: it operates at 4MHz on
- all later machines.
-
- Oleg Kibirev
-
-
-