Re: [steem] Processors
- From: "Steem Authors" <steem@xxxxxxx>
- To: "Steem Discussion" <steem@xxxxxxxxxxxxx>
- Date: Wed, 28 Aug 2002 17:50:50 +0100
Hi Summa,
> > Unfortunately the increased code size could slow the whole of Steem down
> > quite a bit, plus all the checks we would need for what type of
processor is
> > being used (24-bit or 32-bit addresses, that would have to be checked
> > everywhere).
>
> The 68k already works with 32-Bit addresses and just ignores the most
> significant byte of the address registers. Your emulation shouldn't
> work any different otherwise you would have serious compatiblity
> problems, so no checks should be needed since I/O adresses and
> everything else stays the same...
Yes but whenever there is a read or write the high byte is stripped, this
means that if we had 030 emulation we will have to check whether to strip
and then check different values when range checking. This may not seem like
much, but when you have 60,000 lines of code to check and very little CPU
time to work with for a lot of users it is not a good idea. Especially as
emulating the chip accurately would be impossible for us.
> I'm not sure if I understand how the code size would affect the speed
> since my guess is that Steem already don't fit completely into the
> processor cache...
No, but the slightest change can make a huge difference, we try to fiddle
with the core as little as possible.
> > Well the later Amigas were more popular than the later STs, so there is
more
> > software available and more people with a reason to want them. I'm not
sure
> > but I would think it would be easier to implement different processors
in an
> > emulator that isn't trying to be 100% cycle accurate too.
>
> I think the Amiga Emulator is cycle accurate as well, since I don't
> think the demo programmers on this computer did a more protable code
> than the ones on the ST.
I read due to the Amigas complex bus arbitration it is very difficult to do
accurate timings, depending on what is going on at the time any routine can
take a completely different amount of cycles. I heard an emulator was being
attempted, but I don't know how successfully.
> For the 030 emulation mode it don't have to
> be cycle accurate. I'm not sure since I'm far to long haven't done
> any ST programming but I think it might be enough for most of the
> cases to check for the 030 instructions when normaly an illegal
> instruction would occour since only some minor changes were made to
> the 68k set.
I think to do this half-arsed is just asking for trouble, and at the same
time almost totally pointless due to the lack of software that would run on
such an unusual hardware combination. The only way to do this emulation
properly is to write TT emulator including a cycle accurate 030 core and
*all* the TT specific chips. That is a huge task, and it think it can only
been done in a new emulator by people who have access to the original
hardware. Whether it is worth doing is another question, there really
weren't that many TTs about.
Sincerely,
Russ
--
Steem - http://www.blimey.strayduck.com/
Manage your list membership - //www.freelists.org/
Click here to unsubscribe -
mailto:steem-request@xxxxxxxxxxxxx?subject=unsubscribe
Other related posts:
