NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/sys/amiga/misc-part1/3851

Text File | 1996-08-05 | 5.6 KB | 114 lines

Newsgroups: comp.sys.amiga.misc,comp.sys.be
Path: newsstand.cit.cornell.edu!ub!dsinc!scala!news
From: [e-mail] (Dave Haynie)
Subject: Re: 1995 Production Status?
Sender: [e-mail] (Usenet administrator)
Message-ID: <[e-mail]>
Date: Thu, 1 Feb 1996 20:10:36 GMT
Reply-To: [e-mail] (Dave Haynie)
References: <4de07p$[e-mail]> <[e-mail]> <4dk1mb$[e-mail]> <4dp18h$[e-mail]> <[e-mail]> <4e7kfg$[e-mail]> <[e-mail]>
Nntp-Posting-Host: gator
Organization: Scala Computer Television, US Research Center

In <[e-mail]>, [e-mail] (Neil O'Rourke) writes:
>In article <4e7kfg$[e-mail]>, [e-mail] (Dave
>Haynie) wrote:
>
>> >How can the CHRP specifications be so well-designed that any OS will find
>> >the way to use the chips of a CHRP platform ?

>> CHRP defines three basic methods for dealing with the hardware
>> resources....


>> A second level of motherboard resources live under the RTAS (Run-Time
>> Abstraction Services) system. These services allow
>> hardware-independent interaction with NVRAM, clocks, PCI, power
>> management, cache control, and SMP.

>All these abstraction layers, they have a heck of an impact on performance
>don't they?

Not really. First of all, there's not THAT much that's really nested
in abstraction. You do need a way to talk to low-level hardware in any
system. Even on the Amiga we had this here and there, that's what a
"resource" is on the Amiga, a small abstraction layer that sits in
front of a piece of hardware. It should have no more overhead than a C
function call. And 99% of the time, a programmer would write pretty
much the same function anyway, since he knows damn well that the next
time around, something else might be living there. Resources were not
used as much as they should have been on the Amiga. And on PCs, their
approach, the BIOS, isn't as well developed or flexible, but it's more
standardized. The idea of the PPCP HAL is to create a formalized way
to talk to hardware on any system.

The reason you need a ROM on the Amiga, and a different one for each
type of system, is because there was no such general abstraction
layer. This, along with PCI, was a goal for the post-AA generation,
and I wouldn't be surprised to see it surface again in Power Amigas,
even without considering their support of PPCP. It's a plain old good
idea.

>I've been following in Byte about the next MacOS release with
>full hardware abstraction, and it seems that they have to crank up the
>clock speed just to keep applications running at the same apparent
>speed.

I think Apple's been wining about supporting standard platforms since
day one, but don't believe it. If anything, they have to crank up the
clock speed to support Copland, that I can believe. It's not a big
deal to use the PPCP HAL.

>It seems to me that while dropping AmigaOS on to a powerPC based
>platform is a step in the right direction, making it CHRP complient
>will take the "Amiga-ness" out of the system (no blitter to bang for
>moving stuff, no audio chips to make sing, etc. etc)

What radio? Foon quitzit herm? Klaatu barrada nicto!

CHRP isn't going to castrate AmigaOS. First of all, consider that the
lamest PPC system around is still going to have hardware that's 2x-50x
faster than any Amiga on the planet. You can't buy a PCI-based
graphics card, for instance, that doesn't have a blitter at least 10x
faster than the AA blitter. Systems will certainly have audio; even if
it weren't a CHRP requirement (it is -- CHRP requires stereo 16-bit
44.1kHz audio out, and some kind of audio in), it would certainly be
a feature of any Power Amiga. I hear this kind of thing all the time,
but I wish people would think first. After all, virtually every PClone
shipped today comes with a faster blitter and better audio than the
best Amiga. That is a simple fact of life and a 3-year+ shutdown on
Amiga development. There's no reason to believe anyone would build a
CHRP system that took these advances in system minimums away from
you.

>> The final level is Open Firmware (which itself is managed by RTAS).
>> Open Firmware is a CPU-independent device driver specification, based
>> on Forth and now a standard driver type under PCI. Custom resources
>> live on the PCI bus and are supported via Open Firmware drivers.

>Ok, farm out the Amiga's chipset on to a PCI card?

Yes.

>I know PCI is fast, but doesn't this decouple the chips from the
>processor, and the processor from the chips?

It's no different than the system you have now, except the PCI bus is
5x-45x faster than any Amiga local bus. Don't go looking for magic
where there isn't any -- a bus is a bus is a bus. Some are fast, some
aren't. A good PCI-based Amiga chipset client would be more efficient
for CPU access than today's Amiga chipset, in order to be a well
behaved resident of the PCI bus. Of course, you may not even want "AA
on a PCI card" with all the cool PCI chips that are out there
today. Some, true, aren't much more than an SVGA chip with blitter on
a 64-bit bus -- and even that's nothing to sneeze at. But others are,
well, as far above the Amiga chipset as the Amiga chipset was above PC
and Mac displays back in '85.

Dave Haynie          | ex-Commodore Engineering |   for DiskSalv 3 &
Sr. Systems Engineer |  Hardwired Media Company | "The Deathbed Vigil"
Scala Inc., US R&D   |    Ki No Kawa Aikido     |     [e-mail]

         "Feeling ... Pretty ... Psyched" -R.E.M.