| Will the 68050 Be 100% 68020 Compatible? | |
|---|
|
|---|
Morgan Johansson
Sweden
| |
Posts 45
11 Mar 2010 09:47
| Well, subject says it all really.. I don't mean clock-cycles but the instruction-set.
| |
Team Chaos Leader
USA
| | (Natami Team Member)
Posts 1199
11 Mar 2010 10:21
| We are planning for move.(l/w/b) SR, instruction to not be priveliged like it is on the 020. move.x SR, crashes a real 68020 in user mode. This means any dumb 68000 programs that use that instruction will not crash the 68050/70 like they would crash a 68020.
| |
Gabriele Budelacci
Italy
| |
Posts 92
11 Mar 2010 11:32
| Team Chaos Leader wrote:
|
We are planning for move.(l/w/b) SR, instruction to not be priveliged like it is on the 020.
move.x SR, crashes a real 68020 in user mode.
This means any dumb 68000 programs that use that instruction will not crash the 68050/70 like they would crash a 68020.
|
I totally disagree with this idea! BE WARN, please read these docs: EXTERNAL LINK EXTERNAL LINK
Introducing the unprivileged instruction MOVE.X SR in 68000, Motorola did a very bad mistake.
There is a reason why Motorola make this instruction privileged after 68000. I think the target for N050 may be the 68020+, please discard old 68000 'bad written' code.
| |
Phil G.
France
| | (Natami Team Member)
Posts 151
11 Mar 2010 11:36
| This 68000 problem was fixed in 68010 (I think). It prevents correct virtualization.
But it's not as bad as it looks for old code. For OS programs there are patches (like Degrader) which handle it, and for non-OS programs there is still Whdload stuff.
However, the 68050 will probably not be "100%" 68020 compatible, as I doubt it will implement callm/rtm instructions :D
| |
Gabriele Budelacci
Italy
| |
Posts 92
11 Mar 2010 11:46
| Phil G. wrote:
|
But it's not as bad as it looks for old code. For OS programs there are patches (like Degrader) which handle it, and for non-OS programs there is still Whdload stuff.
|
I think I can run WHDload on my A1200, so no unprivileged MOVE.X SR is needed.
| |
Deep Sub Micron
Germany
| | (Natami Team Member)
Posts 352
11 Mar 2010 12:30
| I think the move from SR is a good example, where some kind of configuration register could be useful. In this proposed configuration register will be bits for enabling different features. One feature is user mode executable move from SR without causing a privilege violation trap. Either returning the real SR or a faked SR value.
In that way its up to the user whether he likes to meet the virtualization requirements and/or keep compatibility.Remaining is just the question of how to set bits in this register and what's the default value.
| |
Angel of Paradise
Germany
| |
Posts 32
11 Mar 2010 12:52
| Gabriele Budelacci wrote:
|
Team Chaos Leader wrote:
|
We are planning for move.(l/w/b) SR, instruction to not be priveliged like it is on the 020.
move.x SR, crashes a real 68020 in user mode.
This means any dumb 68000 programs that use that instruction will not crash the 68050/70 like they would crash a 68020.
|
I totally disagree with this idea!
|
I don't want to run Linux on the NATAMI I want to run Amiga OS.
Is this virtualisation actually used or needed for AMIGA OS?
| |
Gabriele Budelacci
Italy
| |
Posts 92
11 Mar 2010 13:03
| Angel of Paradise wrote:
|
Is this virtualisation actually used or needed for AMIGA OS?
|
Today? Surely not!
But... in the future? Are you sure you want close this door now?
The reason? To play 1980' old game? The question is: when you need execute this instruction in user mode?
There is a shortcut to bypass your needs? All software for A1200, A3000 and A4000 surely doesn't needs this instruction in user mode.
Why you want it today in 2010? Cheers ;-)
| |
Thomas Richter
Germany
| |
Posts 695
11 Mar 2010 13:07
| Move from SR has been priviledged more than 20 years ago, exec does
have a function to get the content from the CCR.How old must an application be to fail on that? I don't remember any,
but there was Decigel to address this. So long,
Thomas
| |
Thomas Richter
Germany
| |
Posts 695
11 Mar 2010 13:11
| Gabriele Budelacci wrote:
|
Angel of Paradise wrote:
|
Is this virtualisation actually used or needed for AMIGA OS?
|
Today? Surely not!
But... in the future? Are you sure you want close this door now?
The reason? To play 1980' old game?
|
I think it is already closed. There is no MMU... )-:Greetings,
Thomas
| |
Gabriele Budelacci
Italy
| |
Posts 92
11 Mar 2010 13:22
| My hope is that, in future plans, a MMU (chip or fpga) can be added.
But this is out of this thread topic ;-)
| |
Ralph Eisermann
Germany
| |
Posts 58
11 Mar 2010 13:23
| Since FPGAs are reconfigurable, it could make sense to make both configurations available. I did not miss the 68k really, in my retro A500 my 68010 does its job like in the old days ;) Also the Minimigians do not miss the original behaviour.
add: There were some postings about designing a simple MMU, but at the moment it does not really make sense, because AOS does not support it. So, have a look at the coldfire/coldari instead.
| |
|
Welcome to the Natami lounge.