From: larstr AT colargol DOT idb DOT hist DOT no (Lars Troen)
Newsgroups: comp.os.msdos.djgpp
Subject: Re: memcpy(); is there something faster?
Date: 26 Feb 1997 02:04:13 GMT
Organization: UNINETT news service
Lines: 23
Message-ID: <5f05mt$s7s$1@doffen.uninett.no>
References: <59g08k$758_001 AT cpe DOT Maroochydore DOT aone DOT net DOT au> <5euboi$296 AT flex DOT uunet DOT pipex DOT com>
NNTP-Posting-Host: colargol.idb.hist.no
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp

nikki (nikki AT gameboutique DOT co) wrote:
: > Actually... The fpu memcpy and MMX memcpy which both copy 8 bytes at a time
: > are basically the same. They're both using the same processor and fpu stack.
: > 
: > hmm.. gotta learn djgpp style asm so I can port it to djgpp..

: if y'all want an fpu memcopy in inline asm for djgpp i wrote one here :)
: sadly it's not as good as the old rep movsd though :( basically fpu memcopy
: will outperform anything given that you can guarentee none of the 8bytes you
: are moving will cause an fpu exception. if they do you have to do it the

how can they cause an exception? When moving 8 bytes at a time (with
fild/fistp) you're using integers all the way. If you on the other hand use
fld/fstp (and can copy up to 10bytes at a time, but loose accuracy and get
exceptions and additional cache misses) I understand..


: 'safe' way which gives you a transfer rate of 16bytes/16cycles and 1/2 as
: many cache write misses. if it's 686 or higher you get 1/4 write cache misses.
: sadly, if it's 686 or higher the rep movsd will go faster ;(
: basically fpu memcopy is not all it's cracked up to be.

who's got a 686 anyway (or PPro?) ?