home *** CD-ROM | disk | FTP | other *** search
- > > I have one (two) small question. How fast is it on the Afterburner040?
- >
- > E1M1 (standard benchmark on startup):
- >
- > 15 FPS -> 40Mhz 68040 / 50Mhz DSP / compat_level #2
- > 9 FPS -> 32Mhz 68040 / 32Mhz DSP / compat_level #99
- >
- > I can only get away with compat_level #2 because my DSP is so speedy. This way,
- > the bus never catches up with it and the CPU barely needs to handshake at all.
-
- But 15 fps isn't bad at all. :-) I think it's even a bit faster then before, right
- (I think it was about 11-12 fps on the AB)?
-
- I guess you have plenty of time left on the 040 to put in some fancy audio effects.
- Multi channeld 16bit 3D sound. :-))
-
- > > Will you do any modifications to speed it up further on the AB040?
- >
- > Not much I can do that doesn't involve a big rewrite. :)
-
- :-) What about your own free domain engine your were working on? :-))
- It would be funny to see such a engine on the AB040 running faster and better
- looking than Qukae. :-0
-
- > > Johan talked about drawing the screen in Fast RAM and then copy it
- > > to 'video RAM' (ST-RAM). Or am I totatly wrong here?
- >
- > I tried that (used 'move16' to transfer 16-byte DPHRASE chunks on each instruction)
- > but it's just a fraction slower than normal. It is much faster than normal when
- > there are transparent textures on the screen though - makes a big difference there.
- > Might be a worthwhile option for nasty WADs. :)
-
- Well, if it's faster with transperant textures I can't see why not the option should
- be there (if it doesn't mean to much work).
-
- > The Falcon's 16-bit bus is so crap that the screencopy takes just as long as drawing
- > all of the pixels one by one in vertical columns. I imagine the fact that I pipelined
- > the drawing loops to death has contributed to this invariable screen access time.
- > Badly pipelined code would have gained more from a FastRAM buffer because the loops
- > would 'choke' less on the delays caused by plotting to quick ram.
- >
- > If the Falcon had a 32-bit bus then the difference would have been marked, but as
- > it stands there's not much hope for a fastram screenbuffer making a hell of an
- > improvement in most cases.
-
- Well, if there will be a cheap graphics card then we can start talking... :-)))
-
- //Magnus Kollberg
-
-