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: Difference between 020 & 030 optimisation?
- Date: 23 Mar 1996 05:58:15 GMT
- Organization: Technische Universitaet Muenchen, Germany
- Distribution: world
- Message-ID: <4j03tn$s3p@sunsystem5.informatik.tu-muenchen.de>
- References: <38232880@kone.fipnet.fi> <Do4DHD.By3@cix.compulink.co.uk> <4i3ndh$2bvs@columba.udac.uu.se> <cpc16-1803961527050001@mac002.joh.cam.ac.uk>
- NNTP-Posting-Host: hphalle5.informatik.tu-muenchen.de
- X-Newsreader: TIN [version 1.2 PL2]
-
- Chris Colman (cpc16@mp-s.phy.cam.ac.uk) wrote:
-
- : Data cache can make quite a large difference in e.g Wolfenstein-style
- : wallmapping games where rotating the textures by 90 degrees causes many
- : data cache hits.
-
- wow indeed with horiz reading wallmapper you read withing same
- longword again (when not unscaling more than about factor 1/3).
-
- how fast is a move.l (a0),d0 from datacache. somebody can test please ?
- 4 cycles ?
-
- what about (a0)+,(a1)+. "is faster on 030" only can mean 4 cycles
- as it normally needs 6.
-
- to test this, i suggest to use this loop
-
- REPT
- move.b (a0),(a1)+
- move.l d0,d0
-
- if both instructions need 6 cycles it's good.
-
- : Chris Colman (Findus)
- : cpc16@mp-s.phy.cam.ac.uk
- : A4000/030/10/120 ..060 on its way.......
-