home *** CD-ROM | disk | FTP | other *** search
- Path: comma.rhein.de!serpens!not-for-mail
- From: mlelstv@serpens.rhein.de (Michael van Elst)
- Newsgroups: comp.sys.amiga.programmer
- Subject: Re: Demo/game to OS friendly part II
- Date: 8 Feb 1996 00:40:19 +0100
- Organization: dis-
- Message-ID: <4fbd93$reo@serpens.rhein.de>
- References: <4eq5s3$mnn@serpens.rhein.de> <PETERM.96Feb5131910@tui.maths.irl.cri.nz> <4f4jir$5rn@serpens.rhein.de> <38232339@kone.fipnet.fi>
- NNTP-Posting-Host: serpens.rhein.de
-
- "Jyrki Saarinen" <jsaarinen@kone.fipnet.fi> writes:
-
- >> For the native graphics you would use QBlit() most often.
- >> You would also try to avoid the interrupt overhead for small blits.
- >> The cycles wasted in WaitBlit() could be less than the time spent
- >> for interrupts.
-
- >In what kind of class the cross-over point is?
-
- Depends on the display mode, CPU speed and wether you have FastRAM.
-
- An A3000/25 with FastRAM can do about 2270 pairs of task switches per
- second. So, for a Wait() to gain any CPU time for another task the queued
- blits have to take more than about 440 microseconds. With the current
- blitter that's about 800 words.
-
- One problem with this approach is that a single background task at the
- same priority will slow down the rendering task significantly if the blits
- are not larger than a full time slice. To keep rendering speed one would
- need to bump task priority temporarily.
-
- Regards,
- --
- Michael van Elst
-
- Internet: mlelstv@serpens.rhein.de
- "A potential Snark may lurk in every tree."
-