home *** CD-ROM | disk | FTP | other *** search
- Nntp-Posting-Host: kjerringa.ifi.uio.no
- Newsgroups: comp.sys.amiga.programmer
- Path: sparky!uunet!mcsun!sunic!aun.uninett.no!nuug!ifi.uio.no!olavka
- From: olavka@ifi.uio.no (Olav Lur}s Kalgraf)
- Subject: Re: New hardware reference guide?
- Message-ID: <1992Nov16.134153.6356@ifi.uio.no>
- Sender: olavka@ifi.uio.no (Olav Lur}s Kalgraf)
- Organization: Dept. of Informatics, University of Oslo, Norway
- References: <1992Nov12.215020.20036@ultb.isc.rit.edu>
- Date: Mon, 16 Nov 1992 13:41:53 GMT
- Lines: 69
- Originator: olavka@kjerringa.ifi.uio.no
-
-
- In article <1992Nov12.215020.20036@ultb.isc.rit.edu>, eas3714@ultb.isc.rit.edu (E.A. Story) writes:
- > In olavka@ifi.uio.no (Olav Lur}s Kalgraf) writes:
- > >
- > > Although I agree that it is a step in the right direction, there is a
- > > small problem with using the OS for games etc. Last weekend I tried
- > > to make a small vector routine using the graphics.library. The results were
- > > absolutly PATHETIC! And I use an A4000 for gods sake! After that big
- > > dissapointment I rewrote the code to some giant pseudo OS/hardware hack.
- >
- > Code please? It doesn't mean anything if you don't know HOW to program
- > fast graphics under 3.0... (and you don't, most likely, since MOST people
- > don't have any documentation on 3.0 as of yet).
- >
- Actually I used the 2.0 includes, and used calls to old graphics lib
- fillpoly functions. BUT if 3.0 was coded in a sencible way, these functions would
- be optimized, and recoded to use the AGA chipset, NO? Anyway all I
- did was making a 1 BITPLANE rotating polygon and timing it with the raster
-
-
- > > This worked quite well, but anyway I came to a few conclusions:
- > >
- > > 1. The graphics.library is incredibly slow and most likely
- > > not updated much in release3. This needs to be rewritten
- > > entirely to satisfy the needs for speed in todays games.
- >
- > Again, graphics.library isn't gonna help you much if you insist on using
- > old routines in slow ways.
-
- And again, my result didn't have anything to do with the way I am using
- gfxlib. ONE filled polygon, five points rotated about 3 axis......
- OK, I used the fillpoly or something, function. And I am sure I could
- have done it faster by using the blitter functions, BUT won't these
- functions be kind of obsolete when DIG is introduced. They are assuming
- things about the hardware you know.......
-
-
- >
- > > 2. The dream of DIG can not include the blitter unless 1.
- > > comes true. We really need the blazing speed of optimized
- > > assembly to take advantage of the amigas chips with todays
- > > incredibly slow OS graphics support.
- >
- > We need blazing speed, not necessarily optimized assembly. Two different
- > things.
- >
- Different, true. But also very closely connected, optimized assembly
- means blazing speed.
-
-
- > > 3. Anyone claiming that the speed difference between using
- > > OS and optimized assembly+hardware hacking is minimal
- > > does not know what he is talking about.
- >
- > Depends on what you're doing... I'd suggest that you optimize the way you
- > USE graphics.library, not the way graphics.library is written.
- >
- Have you ever tried hitting the hardware with pure optimized assebly????
- I can tell you that it can't be compared to OS snailcode. I mean, there is
- a reason why most games are written in 'trash the metal' code......
-
-
- Yours, Olav Kalgraf..
-
-
- PS: I really hope you are right about 3.0 gfxlib, it would be nice if
- there were some functions for fast gfxoperations there, in
- pre 3.0 stuff like that was NONEXISTANT!!!
- pre
-