Path: utzoo!mnetor!uunet!portal!atari!apratt
From: apratt@atari.UUCP (Allan Pratt)
Newsgroups: comp.sys.atari.st
Subject: Re: ST hardware scrolling at hblank
Message-ID: <1050@atari.UUCP>
Date: 29 Apr 88 19:38:42 GMT
References: <4650@watdcsu.waterloo.edu>
Organization: Atari Corp., Sunnyvale CA
Lines: 21

From article <4650@watdcsu.waterloo.edu>, 
by magore@watdcsu.waterloo.edu (Mike Gore, Institute Computer Research - ICR):
> [ ... The only trouble would be ... ] interrupt overhead.

Well, interrupt overhead will cost you PLENTY, buster.  Just having the
hblank interrupt vectored to RTE costs about 18% of the CPU, last time we
tried it.

On a monochrome system, you get 400 lines * 50Hz = 20 KHz (20 Kilo
hblanks/sec); the exception takes 44 cycles, RTE takes 20, so that's (66
cycles per hblank) times 20 KHz = 1,320,000 cycles/sec in overhead;
that's 12% of the CPU mathematically -- there are lost cycles when the
CPU can't get the bus (during video fetch and refresh), and other
vagaries...  People here computed (years ago) that you lose more like
22% of the CPU with an RTE in Hblank. 

And now you want to do WORK during hblank?  Count those cycles!

============================================
Opinions expressed above do not necessarily	-- Allan Pratt, Atari Corp.
reflect those of Atari Corp. or anyone else.	  ...ames!atari!apratt