home *** CD-ROM | disk | FTP | other *** search
- Path: www.cybercity.dk!usenet
- From: ccc6004@vip.cybercity.dk (Hans Henrik Happe)
- Newsgroups: comp.sys.amiga.programmer
- Subject: Re: One hardware-basher's manifesto
- Date: 26 Feb 1996 12:24:34 GMT
- Organization: CyberCity of Denmark
- Message-ID: <1429.6629T1374T1599@vip.cybercity.dk>
- References: <751.6612T1001T302@vip.cybercity.dk> <04000205714080640503@BIRDLAND> <4ge8na$vhe@ar.ar.com.au>
- NNTP-Posting-Host: 194.16.56.108
- X-Newsreader: THOR 2.22 (Amiga;TCP/IP) *UNREGISTERED*
-
-
- Storm ( storm@ar.ar.com.au ) wrote:
-
- >> ALL Caches should be turned on :) Else it is a speed-waste...
-
- > Is it not true that data burst mode will slow you down if you are "randomly"
- > accessing memory? I've heard that it will, although I have not experimented
- > with it. Logic seems to dictate that it could.
-
- > Although, you could argue, that you should leave the caches alone, and leave
- > it up to user to fiddle with them if he so desires.
-
- Then you have another good reason to do it OS-friendly. Because then the user
- can fiddle with the caches when a demo is changing from a rutine which are doing
- random access to a rutine doing lineary access :)
-
- No cache is not an easy one ;( But what about giving the user some choises before
- intering the demo. It would depend on the demo, but consider somthing like:
-
- Burst on/off
- Data cache on/off (040 :( )
- Inteligent cache handling ( The coder knows what's best for you,
- because he/she did the thing. )
-
- But most inportant of all (and there is no excuse): Make a prober timing rutine,
- because then a bad cache setup will not affect the frame sync (meaning - if one
- state of a rutine takes 6 frames (hopefully not ;) other states will take the same.
- This is ofcause hard to achive the states are very different over a long period of
- time :(
-
- Hans Henrik Happe
- Goat / Sumpen
-
- You can kill the human race, but you can't kill the AMIGA.
-
-