NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/sys/amiga/programmer/552

Internet Message Format | 1996-08-05 | 13.6 KB

Path: easy.in-chemnitz.de!mkmk!floh
From: [e-mail] (Andre Weissflog)
Newsgroups: comp.sys.amiga.programmer
Subject: Re: amiga questions.  -  riscami.txt [1/1]
Message-ID: <X5kVx*[e-mail]>
Date: Mon, 08 Jan 1996 20:37:15 CET
Reply-To: [e-mail]
References: <4cr622$[e-mail]>
Distribution: world
Organization: private uucp site
X-Newsreader: Arn V 1.04

In article <4cr622$[e-mail]>, NEW COLLEGE writes:

[...about AT's RISC plans...]

> First about the OS porting:

>      If they port the OS to other systems it could be disastrous
> for the amiga in general, because although the OS is undoubtedly
> the best I have ever used if it is ported it will lose some of
> it's best features, most importantly because it could not longer
> be sure that each platform had the same base config, any extra
> features would have to be emulated in software, thus packing out
> and slowing down the kernal extremely i.e. the new RISC amiga
> may have a superfast 24-bit blitter (I hope! :-) but how many
> powerMacs have you seen with one, any portable os would have to
> emulate one which would be slow. This also raises another

Well, that's what a driver system is all about. It hides a
specific hardware from the programmer, which is a Good Thing.
Imagine a pc game which runs in 640x480 in 256 colors (which
is not a vga mode and is invoked differently with every
gfx chip set). Let's say, the game hits the hardware and
can only talk to ET4000w32 chips (that's the same situation
as with many Amiga games today). The game would be
completely useless to those users with an s3 chip on their gfx
card. The only option for a typical coder would be to support
a good dozen graphic chips directly.

Now imagine the game would use a driver system that provides
transparent access to bit blitting, double buffering etc....
On an ET4000w32 this game would possibly run a bit slower
due to the additional software layer, but with a good
driver and a 128 bit card the game runs ***much*** smoother then
it could ever do on an ET4000w32. And if somebody with an old
blitterless graphics card is not satisfied with the frame rates,
at least he has the chance to run the same game on his brand new
ultra vga card he has just bought.

> important question, if the os has to use drivers to provide
> emulations of some features which will be slow, any software
> which hopes to be portable will not be able to use many of the
> RISC amiga's feature because it would crawl like a snail with

Future Amigas *must*be*defined*only* by AmigaOS, or the
system won't survive (which doesn't mean of course, that there
shouldn't be cool default hardware in it, but the OS must provide
the interface to it).

By the way, the current AmigaOS *uses* drivers to access the
hardware (except the graphics system, but the current Amiga
graphics cards and CyberGfx prove that this is possible and
much faster then the original chip sets).

> arthritis on any other machine so we will end up with large
> podgy, slow apps. Also if the OS is to be portable it will
> immediately wipe out one of the key features of the amiga,
> hardware bashing, if each platform is different (as it will be)
  ~~~~~~~~~~~~~~~~~
Hitting the hardware directly was and is a Bad Thing.

> then the program cannot use things such a the copper or blitter

At least for 2D bit blit stuff there exists a complete
software interface for 20 years or so called bitblt. Why should
I not use it? A good driver talking to a good hardware with low
overhead is the only way to go. Standard PC cards have 50MByte/sec
blitters, and we can expect 500MByte/sec in a few months, without
software drivers you can have those 50MB/sec today but forget
about the 500 tomorrow because the hardware interfaces WILL
be different because you can't predict which manufacturer
will have the chip set out first and which one will sell
the most chips.

Copper is a nice idea but completely obsolete with todays cpus,
color depths and display resolutions.

It's all a bit more complicated for 3d stuff. The ultimate lowcost
3d rendering device is still not here and there's much room for a
new machine that would be the 90's 3d-equivalent of the
original Amiga-idea.

> directly because it cannot be sure it is there, instead it will
> have to go  through the OS which will slow it down tremendously,
> for example if you feel like a challenge program a game like
> stardust without using the blitter or copper and only using OS
> calls, hard isn't it? Finally for this section the amiga OS gets a lot
>  of its speed from the fact that most of intuiton`s buttons etc. are
          ^^^^^
> stored in rom along with most of the major OS code, if the OS became
  ^^^^^^^^^^^^^
Wrong, ROM code massively slows down execution, that's why
there exist tools that remap ROM into FastRAM. And "intuition's
buttons" aren't "stored" in ROM either, they're rendered on the fly
using graphics primitives like line, rect and text.

>  portable all of this would have to be loaded from  disk, slowing down
>  screen redraws ect. or it would all have to be loaded into ram, giving
> lightening fast redraws (faster than from rom) but would swallow large
>  chunks of memory. At present AT seem to be trying to form some
>  kind of wierd amalgam of windows, system 7 and workbench (Maybe
>  they should call it Windows system workbench?), if they were to

That's bullshit. MacOS sucked from the beginning, and Windows
always was a failed try to copy the Mac. IMHO that's because both
systems are designed from the wrong end. The Mac wanted to provide
a good human interface as it's main design goal, thus the low level
stuff suffered from this and the OS as a whole became a perversion of
the original idea. Well, and Windows is just the same, just
that Microsoft doesn't have any design goals at all except ruling
the world. Because there's no solid ground to build on, enhancements
will need loads of new code, and possibly some dirty tricks to
function at all. That's why Win95 and MacOS are so bloated, not
new "features".

AmigaOS is extremely well (ok, pretty well) designed on the lower
levels, namely exec.library with the device driver system. So
there's a slightly different situation.

>  release the OS as a software package, as they seem to want to do
> it has a big chance of failing, just look at OS/2 that was a nice (sort-of)
>  multitasking OS from IBM, one of the worlds largest computer companies
>  but it only sold about 3 copies because of  windows.

Maybe that's because OS/2 does many things, but nothing really
well. And it's also an image question. When running office
applications, why should I use OS/2 instead of Windows? When
I want to do 3d modeling and design, why should I want to use
OS/2 instead of (low budget option, Amiga; high budget option,
SGI with Alias)?

>    With regard to the statement that they are dropping the AAA
> chipset and farming out the new chipset designs to a new company WHY!!!,
> the chipset is one of the amiga`s most powerful features which the OS uses
> extensively to give full pre-emptive (and FAST!) multitasking and dropping

Bullshit again. The chip set has nothing to do with multitasking,
and asynchronous blitting can be done with any $25 vga chip set
and the right OS. I haven't seen AAA, but reading the specs I must
say it would have been a killer by 1991 and maybe it would have been
impressive in early 1993 but today every $150 PCI graphics card
is way better.

> this in favour of a cheap and nasty svga card would be the worst decision
> ever made. Apparently for the new RISC amiga AT will farm out the design of
> the chipset to a outside company. This is not a bad idea if the outside
> company are any good but a far more sensible idea would be to give them the
> designs for the AAA chipset and tell them to modify them to suit AT`s needs.

To little, much to late and way to expensive. There are N companies out
there doing nothing else then designing graphic chips all day,
AT would have no chance. An interesting option would be to go to
one of the established chip makers and ask for a few
specialized enhancements and/or integrated parts (like, a 1 chip
A1200).

> This is sensible because when Commodore died the AAA chipset prototype
> was 96% complete (figures from the deathbed vigil video), dropping a chipset
>
The design might have been as complete, but according to They Lord
Dave Haynie's postings in c.s.a.hardware some time ago, the overall
process was somewhere around 60% (tons of test runs and bug fixes
still to be done).

> this close to completion is like shooting yourself in the foot, even tough
> there was no proto. OS support this. In fact the lack of OS support would be
> an advantage because it will help AT to make a fresh start and rewrite the
                                                 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
> OS from the ground up. Without the tricks the chipset can do the amiga will
  ~~~~~~~~~~~~~~~~~~~~~~
Haha, very funny.


> be severely limited because it will be lacking most of the things it was
> originally designed for, for instance using present SVGA chips it will not
  ~~~~~~~~~~~~~~~~~~~~~~~
AmigaOS was oriniginally designed for being enhanceable, not
beeing limited to 1 and only 1 hardware. Some components are
more portable then the others, but the basic philosophy behind
AmigaOS is very ok.

> be able to overlay screens of different resolutions, so no more menus like
> those in Dpaint, and no nice high resolution status bars in games, everything
>
If you can have a 1280x1024x24 bit display and can do complex
operations on pixel level in 50fps, why would you need to do such
tricks?

> will have to run on the same screen, slowing things down, and no more fancy
> chipset tricks for demo coders, everything will have to use the cpu for
> tricks, if you want an idea of what this will be like just look at pc demos,
>
> boring arn`t they?.

As far as games and similar fun stuff goes, I'd rather have a high
resolution 50fps 3d environment around me instead of a 50fps 2d
parallax scroller.

> Almost finished now, just a couple more things.
> If they are serious about porting the OS I hope that the process of writing
> the code will not take up so much time that they will not be able to
> significantly rewrite the OS, because it is beginning to look a little
> dated against Win95.

To make AmigaOS 'look like Win95' requires no complete rewrite, just
a little bad taste and a few changes in Intuition. To make AmigaOS
behave like Win95 you had to add a few busy loops here and there
of course.

As far as rewriting goes, you obviously have no idea how much work
would be involved with this (ignoring the fact that massive rewritings
are not necessary (well, except rewriting the assembly parts to C of
course)). IMHO, the only parts that need work on the current OS are
graphics (obviousely, for complete device independance) and a clear
decision should be made, what the gui direction of choice will be in
the future (BOOPSI instead of gadtools I hope). The current datatypes
implementations should be improved, and a central oop system needs to
be hammered out (those NewDTObject()/SetGadgetAttrs() function mutants
make me nervous for the future.

> Finally if the new amiga uses PCI slots this would enable cheap PC expansion
> cards to be used, but this raises a major problem, all amiga cards have a
> small rom chip in them to say what they are, so that the autoconfig system
> can work but PC cards do not have this so the autoconfig will not work for
> them, this means that in the startup sequence a software system will have
> to be used to recognise and mount such cards, this will be aukward and
> slow (just look at the bodge Win95's software autoconfig does of some
> things).

PC PCI cards *have* autoconfig code on them, a minor problem is just
that most cards don't use the processor independent OpenBoot standard
but x86 code hunks (which is perfectly within PCI specs). The Win95
hardware config dilemma is mainly due to conventional devices which
can't report about their behaviour and possibly ISA plug'n'play toys.
Don't expect the same chaos on a future PCI Amiga, in fact, chances
are good that you wouldn't notice a difference at all. Just plug your
card in, switch the Amiga on and if necessary, install some devices by
double clicking the 'Install' icon.

[stuff deleted]

>
> P.S. If AT release a new interim amiga (or an updated 1200) , some sort of
> version of the present software pack but aimed at programmers would be
> a great sell I'm sure, because a lot of people who buy an amiga want to
> program it but at present are put off by all the different choices and
> the expense of the software (I'm only just learning myself). The pack could
> contain things like:
>
>
> A1200/Interim Amiga
> 250mb+ Hard Drive as an option
> A69K + A tutorial
> A C/C++ compiler like DICE, the new Storm Compiler or
> otherwise something like Amiga E.
> Includes, Autodocs and A tutorial for the above.
> Paint Package.
> A Gui construction system such as E's EasyGUI.
> A version of the RKM's fo the new machine.
>

A CD-ROM with every machine containing a complete development
environment would be nice, that's true.

> P.P.P.S (Last one!) Does anyone have any information on the new Storm
> C compiler, a URl/email-address would be nice :-).

It seems to be a MaxonC++ spin off, and I can't remember having heard
any good critics about it, sorry.

Bye,
-Floh.

[I have a vague feeling like having seen most of this posting
several months ago?]

====//=== Andre Weissflog <[e-mail]> =======
...// Sep'95: Return Of The Living Death...................
\\// 90% of everything is crap (Sturgeon's Law)...........
=\\===============================================Amiga!=