home *** CD-ROM | disk | FTP | other *** search
- Path: informatik.tu-muenchen.de!fischerj
- From: fischerj@informatik.tu-muenchen.de (Juergen "Rally" Fischer)
- Newsgroups: comp.sys.amiga.programmer
- Subject: Re: Demo/game to OS frien
- Date: 12 Mar 1996 16:55:57 GMT
- Organization: Technische Universitaet Muenchen, Germany
- Distribution: world
- Message-ID: <4i4aat$7o8@sunsystem5.informatik.tu-muenchen.de>
- References: <38232498@kone.fipnet.fi> <4gctlm$bcn@oreig.uji.es> <38232604@kone.fipnet.fi> <4gj3ve$f9c@sunsystem5.informatik.tu-muenchen.de> <642.6629T546T2374@login.eunet.no> <4gpjgi$19c@serpens.rhein.de> <265.6630T583T838@login.eunet.no> <4gt2ub$ban@serpens.rhein.de> <4gtm3n$mu0@sunsystem5.informatik.tu-muenchen.de> <4h6ag5$541@serpens.rhein.de>
- NNTP-Posting-Host: hphalle10.informatik.tu-muenchen.de
- Originator: fischerj@hphalle10.informatik.tu-muenchen.de
-
-
- |> If there is no buffer then the CPU databus is occupied for the
- |> complete cycle. No way to interleave that with other accesses
- |> and not lose cycles.
-
- but a infinite fast cpu+fastmem can aproximate 7mb/sec that way :)
-
- |>
- |> The best compromise for all systems is to fetch as many words
- |> into registers and then write them to chip memory. Of course that
- |> loses the A3000's (and other's) advantage but the overhead is
- others ? other "chipbuffer" amigas ?
-
- well, on 020-14 normal copy is also better because it does in 12 cycles
- vs. 8 + "somewhat about 5" when doing movem load.
-
- 030 boards are said to do move.l (An)+,dn faster than movem load. funny.
-
- |> minimized (especially for machines with 16bit chipmem).
- |>
- |> Regards,
- |> --
- |> Michael van Elst
- |>
- |> Internet: mlelstv@serpens.rhein.de
- |> "A potential Snark may lurk in every tree."
- ------------------------------------------------------------------------
- fischerj@Informatik.TU-Muenchen.DE (Juergen "Rally" Fischer) =:)
-
-