🚀 go-pugleaf

RetroBBS NetNews Server

Inspired by RockSolid Light RIP Retro Guy

Article View: comp.arch
Article #22777

Re: BitBlt, new instructions for RISC.

#22777
From: Randell Jesup
Date: Mon, 05 Mar 1990 09:20
46 lines
2776 bytes
[Sorry of people see this twice, news was broken the first time I posted it.]

In article <AZua2yC00WB5Qqd2FJ@andrew.cmu.edu> zs04+@andrew.cmu.edu (Zachary T. Smith) writes:
>until the (semiconductor) industry is capable of putting a blitter on
>a 1 megabit VRAM chip (or better yet on a 4 megabit chip), and then
>just do that. It'll get done eventually, so why come up with hacks in
>the meantime?
...
>PS: Why wouldn't it get done? A blitter that operates on 1024 or 2048
>bit lines per op and which can cache fonts in the DRAM as well would
>even a very fast blit processor running w/standard VRAM look like a
>toy.

	I wouldn't hold my breath.  It may happen, but the cost will probably
be prohibitive even if it does.

	Some off-the-cuff technical problems I can think of are:
1) The blitter can only access data on the same "BRAM".
2) It complicates video system design _a lot_, since the font/etc data must
   be stored in non-visible portions of each ram.  This gets bad, since
   nothing normally forces characters to appear on any given alignment.
why go further, I think those will kill the idea already.

	Personally, I think the correct solution is not to modify RISC cpu
designs to make blitting easy, but to use custom blitter/etc chips in the
video sub-system.  They can be optimized for their intended function more than
any RISC cpu can (since it must also be a good general-purpose cpu).  They
can be isolated from the main CPU, allowing limited multi-processing (tell the
blitter to start, and go off and do other things while waiting).  Isolating
the video system from the main CPU bus means less bus contention (seen by the
CPU), etc.

	This follows the methodology that hardware (CPU/whatever) should be
optimized for it's task whenever possible.  You look at the speed (measured
in both blitting speed and overall system throughput) and the cost, and then
decide.

	Of course, I could be biased, since the Amiga uses a custom video
coprocessor, and runs much faster when the CPU runs from ROM/fast-ram, while
the blitter (and display "copper" (coprocessor)) play with "chip" (video) ram.


-- 
Randell Jesup, Keeper of AmigaDos, Commodore Engineering.
{uunet|rutgers}!cbmvax!jesup, jesup@cbmvax.cbm.commodore.com  BIX: rjesup  
Common phrase heard at Amiga Devcon '89: "It's in there!"
Technical Details
Message-ID: <9990@cbmvax.commodore.com>
Path: arc1-ca.pugleaf.net!utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!cbmvax!jesup
References: <AZua2yC00WB5Qqd2FJ@andrew.cmu.edu>

Show Full Headers
Complete Article Headers
Path: arc1-ca.pugleaf.net!utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!cbmvax!jesup
From: jesup@cbmvax.commodore.com (Randell Jesup)
Newsgroups: comp.arch
Subject: Re: BitBlt, new instructions for RISC.
Message-ID: <9990@cbmvax.commodore.com>
Date: 5 Mar 90 09:20:38 GMT
References: <AZua2yC00WB5Qqd2FJ@andrew.cmu.edu>
Reply-To: jesup@cbmvax.cbm.commodore.com (Randell Jesup)
Organization: Commodore, West Chester, PA
Lines: 46