T.R | Title | User | Personal Name | Date | Lines |
---|
3343.1 | MMU needed | FROCKY::BALZER | Christian Balzer DTN:785-1029 | Tue Jan 16 1990 09:03 | 6 |
| You'll need a MMU, like the one that comes with the A2620 board
or the one inside the 68030.
Regards,
<CB>
|
3343.2 | Thanks <CB> | MSVAX::BARRETT | Who is Steve Dahl? | Tue Jan 16 1990 13:20 | 14 |
| Well, that raises a few questions then.
1) Does this imply that you need memory expansion; that it will
not work with the stock 1meg?
2) If I get MMU from somewhere, will it work without the additional
hardware? Will SETCPU?
3) Is there a way I can get MMU to try this?
Some games (the original ARCHON for example) require kickstart V1.1,
so I would love to get this to work. It will also be great when
Kickstart V1.4 comes out, as the last time the disk was out
several months before the chip.
|
3343.3 | | DICKNS::MACDONALD | WA1OMM 7.093/145.05/223.58 AX.25 | Tue Jan 16 1990 15:01 | 28 |
| < Note 3343.2 by MSVAX::BARRETT "Who is Steve Dahl?" >
-< Thanks <CB> >-
Well, that raises a few questions then.
1) Does this imply that you need memory expansion; that it will
not work with the stock 1meg?
>> The memory expansion is not the critical piece.
2) If I get MMU from somewhere, will it work without the additional
hardware? Will SETCPU?
>> STECPU V1.5 was designed for use with the Commodore A2620 and A2630
>> coprocessor boards. It may work with other coprocessor boards, but
>> there are no guarantees.
3) Is there a way I can get MMU to try this?
Some games (the original ARCHON for example) require kickstart V1.1,
so I would love to get this to work. It will also be great when
Kickstart V1.4 comes out, as the last time the disk was out
several months before the chip.
>> Yep, buy an A2620 board.
|
3343.4 | Is there no way then? | MSVAX::BARRETT | Who is Steve Dahl? | Tue Jan 16 1990 15:17 | 4 |
| Is there no way to boot a kickstart disk on a 2000 without
a coprocessor board? It would seem to me that it might be possible
to load kickstart somewhere in a 256k chunk of memory, then boot
to it.
|
3343.5 | I believe the problem is Kickstart is not relocatable | TLE::RMEYERS | Randy Meyers | Tue Jan 16 1990 17:25 | 15 |
| Re: .4
> It would seem to me that it might be possible
> to load kickstart somewhere in a 256k chunk of memory, then boot
> to it.
No. The problem is that Kickstart is a non-relocatable program
with non-relocatable data. It has in it absolute addresses that point
to particular addresses in the ROM. You can't just move that code
and data to another address and have it work.
SetCPU works by setting up the memory management unit to translate
all the addresses in the ROM space to point to the RAM that the alternate
version of Kickstart is loaded into. After that, the system thinks that
the alternate Kickstart is located at its expected address.
|
3343.6 | | WILARD::BARRETT | Is my PC a PicoVAX? | Tue Jan 16 1990 17:35 | 2 |
| Is this possible without a co-processor, or does memory management
require it?
|
3343.7 | Not with the 68000 | TALLIS::MCAFEE | Steve McAfee | Wed Jan 17 1990 08:58 | 7 |
| My understanding is that the 68000 by itself doesn't work
well with an MMU. Something about the interrupts which the
MMU needs to issue. I understand that early on Apollo among
others built 68000 systems with two 68000's, one running a
few cycles (one?) ahead of the other.
- steve
|
3343.8 | MMU is overkill without 68020 | LEVERS::PLOUFF | can't memorize Zen... | Wed Jan 17 1990 09:22 | 31 |
| re: .6
The double-68000 machines were designed that way for virtual memory.
Problem with a 68000 is: if there is a VM page fault in the middle
of an instruction, there's no way to recover the internal state
of the processor. This was fixed on the 68010.
re: adding an MMU
The memory management unit sits between the processor chip and
everything else on the memory bus (excluding other coprocessors).
It translates processor addresses to memory addresses and introduces
a performance hit to the system. Motorola's MMU chip adds 1 cycle
to every memory access.
The reason it's on the A2620 card is to support Unix's virtual memory
features. SETCPU's Kickstart relocation feature is just gravy.
But this is an expensive feature if you just want to have more than
one Kickstart available. Dave Haynie has said in public that the
reason he put the feature into SETCPU was because a 68020-equipped
Amiga ran slightly faster when Kickstart was in 32-bit wide memory
as opposed to being accessed across the 16-bit backplane bus.
If all you want is to have more than one Kickstart option, I think
you should stick with the hacks that put extra Kickstarts into EPROMs
and switch select between them.
My hardware 2�.
Wes
|
3343.9 | sigh...ROM upgrades aren't fun... | NAC::BRANNON | value added | Wed Jan 17 1990 15:50 | 7 |
| re .6:
Or buy a used Amiga 1000 to run the games that haven't followed the
rules.
regards,
dennis
|
3343.10 | In the future... | FROCKY::BALZER | Christian Balzer DTN:785-1029 | Thu Jan 18 1990 04:17 | 6 |
| For what it's worth, the 1.4 version of Kickstart will *probably*
be relocatable, so you could use it w/o a MMU.
Regards,
<CB>
|