Welcome, Guest. Please login or register.
Amiga Kit Amiga Store Iridium Banner AMIStore App Store A1200/A600 4xIDE Interface

AuthorTopic: TeamAROS Bounty #23 AROS Kickstart Replacement ROM  (Read 7104 times)

0 Members and 1 Guest are viewing this topic.

  • Guest
Re: TeamAROS Bounty #23 AROS Kickstart Replacement ROM
« Reply #45 on: January 05, 2005, 06:48:48 PM »
Quote

jdiffend wrote:
Quote

bloodline wrote:

What do you mean dreaming? It's prefectly feasable, and only really needs this bounty to be completed.


It may be feasable but more work needs done than this bounty requires.

The coldfire has some major differences in how some things in the supervisor mode work and it would require a lot more changes to the exec and boot code than just adding the instruction emulation code.


Maybe Elbox would be interested in AROS on their Coldfire boards?
 

Offline jdiffend

Re: TeamAROS Bounty #23 AROS Kickstart Replacement ROM
« Reply #46 on: January 05, 2005, 06:51:41 PM »
Quote

whabang wrote:

Well, there's this wee little detail of actually designing a Coldfire-board, making a Coldfire-version of AROS, make it trap the non-supported instructions (which would be limited to non-MMU, and possibly non-FPU applications),  and then write proper hardware drivers for the board, and any clones. But yes, it's doable.


Actually, if you want a coldfire/AROS based computer you could buy one of the development systems with built in VGA graphics and/or PCI card slots.  They just need an AROS port, case and power.
http://www.futureelectronics.com/promos/coldfire/MCF547x/

Or, instead of designing the entire Coldfire board for existing Amiga's you could build an adapter that holds one of the Coldfire module boards.  Then you just need the logic between the coldfire and the Amiga connector, the rest already works and you just program it.  I thought about doing this.

As for the MMU/FPU stuff... anything that hits the MMU or FPU itself isn't going to work.  If programs use the math.library then a Coldfire math.library would do the trick.  There's also an MMU.library and stuff that uses it might also work with a new version of the lib.

Something to think about... this year Freescale *should* start sampling the V5 Coldfore core to select customers.   It should offer twice the performance of the current 4e core chips.  That *might* be worth preparing for.

Reality check for you people drooling over the coldfire ...
When I first started looking at a Coldfire Amiga, Motorola hadn't even introduced the MCF5307 but they did have the 4e core on their roadmap.  At the time it looked promising because PC's weren't nearly as fast as they are now.  Since then the roadmap has slipped by something like 4 years and in that time the gap has widened so much (I just built my parents a nice 2.4 GHz PC for Christmas for $230) that I don't see the point in it anymore.  I checked pricing on the new Coldfire CPUs in small quantities (less than 10000) and it's as much as I can buy an Athlon for.  And the Socket A motherboards start at about $25.  To print a small 2 or 4 layer board as an adaptor for a coldfire module board will cost at least 2 times that in small numbers and you still need the module!

AROS might make a great embedded OS for the Coldfire and it would be good for settop boxes or low end machines for third world countries.  But even a V5 core is predicted to run only 610 Dhrystone mips at 333 MHz.  A 2.4GHz CPU like my parents have offers 4644 Dhrystone mips!  Even when the coldfire CPU speed gets bumped to 800+MHz like one of their roadmaps displayed (that roadmap isn't on the Freescale site) it will probably only be around 1500 mips and that will be at least two years from now when multi-core AMD and Intel CPUs should be everywhere.  (Motorola did mention a multi-core Coldfire at one point but all discussion of it has since disappeared just like the 800+MHz)

AROS for UAE makes sense.  For a real world miggy it makes sense for working to improve AROS compatibility and providing a migration path.  For the Coldfire it makes sense for embedded/settop box use.   For a way to bring a late 80's/early 90's machine into the new millenium in combination with the Coldfire... it's going to fall way short.

*IF* Freescale were to introduce a V5 Core Coldfire at 800+ MHz with MORE THAN 2 CORES then you could think about it.  But with the Coldfire targeted at embedded systems I just don't see that anytime soon.  (I do expect to see a Coldfire CPU with 800+MHz and multiple cores... just not for a few years)
 

  • Guest
Re: TeamAROS Bounty #23 AROS Kickstart Replacement ROM
« Reply #47 on: January 05, 2005, 07:00:41 PM »
Quote

whabang wrote:
Quote

bloodline wrote:
What do you mean dreaming? It's prefectly feasable, and only really needs this bounty to be completed.


Well, there's this wee little detail of actually designing a Coldfire-board, making a Coldfire-version of AROS, make it trap the non-supported instructions (which would be limited to non-MMU, and possibly non-FPU applications),  and then write proper hardware drivers for the board, and any clones. But yes, it's doable.


It depends what kind of 68k compatibility you want.

Why bother trapping the missing 68k instructions etc, when native Coldfire AROS would work faster, and then we could use UAE for everything else.

Exactly the same way as Native x86 AROS.
 

Offline jdiffend

Re: TeamAROS Bounty #23 AROS Kickstart Replacement ROM
« Reply #48 on: January 05, 2005, 07:09:13 PM »
Quote

mdma wrote:

Why bother trapping the missing 68k instructions etc, when native Coldfire AROS would work faster, and then we could use UAE for everything else.

Exactly the same way as Native x86 AROS.


You still need the instruction traps if you are going to try to run Amiga apps on it since we can't recompile all of them to be Coldfire compatible.
 

  • Guest
Re: TeamAROS Bounty #23 AROS Kickstart Replacement ROM
« Reply #49 on: January 05, 2005, 07:15:49 PM »
Quote

jdiffend wrote:
Quote

mdma wrote:

Why bother trapping the missing 68k instructions etc, when native Coldfire AROS would work faster, and then we could use UAE for everything else.

Exactly the same way as Native x86 AROS.


You still need the instruction traps if you are going to try to run Amiga apps on it since we can't recompile all of them to be Coldfire compatible.


I should have been a bit clearer there i think.

I meant that there is no need for traps if AROS is compiled natively for the Coldfire CPU and UAE is used for 68k programs.

Once this AROS for UAE is completed, then work can start on the integration of this UAE into AROS to provide "transparent" emulation of system friendly 68k apps, as talked about on the dev-list and aros-exec.

Once all these things are achieved then we'll have a great little OS on our hands! :-)
 

Offline MskoDestny

Re: TeamAROS Bounty #23 AROS Kickstart Replacement ROM
« Reply #50 on: January 05, 2005, 08:15:29 PM »
Quote

mdma wrote:
I meant that there is no need for traps if AROS is compiled natively for the Coldfire CPU and UAE is used for 68k programs.

Using UAE will be dog slow compared to traps, especially since I don't believe there is a 68K -> Coldfire dynarec for UAE at the moment so it will have to run through the interpretter.  UAE would still be useful for programs that are dependent on behaviors that you can't easily simulate on Coldfire using the trap approach.
 

  • Guest
Re: TeamAROS Bounty #23 AROS Kickstart Replacement ROM
« Reply #51 on: January 05, 2005, 08:36:37 PM »
Quote

MskoDestny wrote:
Quote

mdma wrote:
I meant that there is no need for traps if AROS is compiled natively for the Coldfire CPU and UAE is used for 68k programs.

Using UAE will be dog slow compared to traps, especially since I don't believe there is a 68K -> Coldfire dynarec for UAE at the moment so it will have to run through the interpretter.  UAE would still be useful for programs that are dependent on behaviors that you can't easily simulate on Coldfire using the trap approach.


Someone who had a commercial interest in Coldfire Motherboards could have a vested interest in writing a 68k to ColdFire dynarec though! ;-)
 

Offline jdiffend

Re: TeamAROS Bounty #23 AROS Kickstart Replacement ROM
« Reply #52 on: January 05, 2005, 08:40:29 PM »
Quote

MskoDestny wrote:
Quote

mdma wrote:
I meant that there is no need for traps if AROS is compiled natively for the Coldfire CPU and UAE is used for 68k programs.

Using UAE will be dog slow compared to traps, especially since I don't believe there is a 68K -> Coldfire dynarec for UAE at the moment so it will have to run through the interpretter.  UAE would still be useful for programs that are dependent on behaviors that you can't easily simulate on Coldfire using the trap approach.


I think the confusion comes from expectations raised by UAE running on PCs.  It may be fast on modern PC's but it won't be on a Coldfire.  Any app that can run without UAE will be many times faster.

UAE should only be used as a last resort on apps that don't run with the traps on a Coldfire CPU.  If you want to know how fast UAE will run on a Coldfire, try running it on a PC from 1996 or 1997.  I'd guess a 266MHz to 333MHz Pentium II should be pretty close to the same speed as a 4e Coldfire.  

It will be easier to write a dynarec for the coldfire since many of the instructions would stay as is or would require only a couple to duplicate and the register model is the same on the 68K and 4e core.
 

Offline jdiffend

Re: TeamAROS Bounty #23 AROS Kickstart Replacement ROM
« Reply #53 on: January 05, 2005, 10:50:00 PM »
Thanks to some of the stuff on Aminet I should be able to support a boot device.  An AROS version of the driver will take time though.
 

  • Guest
Re: TeamAROS Bounty #23 AROS Kickstart Replacement ROM
« Reply #54 on: January 05, 2005, 11:26:15 PM »
Quote

jdiffend wrote:
Thanks to some of the stuff on Aminet I should be able to support a boot device.  An AROS version of the driver will take time though.


2005 could well turn out to be "The Year of AROS" :-D

20 years after Amiga was first shown to the public too.
 

Offline jdiffend

Re: TeamAROS Bounty #23 AROS Kickstart Replacement ROM
« Reply #55 on: January 06, 2005, 02:06:57 AM »
Quote

mdma wrote:

2005 could well turn out to be "The Year of AROS" :-D

20 years after Amiga was first shown to the public too.


Don't jump to conclusions.  I'm just starting to look through everything, I haven't even tried to compile a single line of code yet and I've already found a lot of potential problems.  The scope of this project seems to grow in leaps and bounds.

 

  • Guest
Re: TeamAROS Bounty #23 AROS Kickstart Replacement ROM
« Reply #56 on: January 06, 2005, 02:10:57 AM »
Quote

jdiffend wrote:
Quote

mdma wrote:

2005 could well turn out to be "The Year of AROS" :-D

20 years after Amiga was first shown to the public too.


Don't jump to conclusions.  I'm just starting to look through everything, I haven't even tried to compile a single line of code yet and I've already found a lot of potential problems.  The scope of this project seems to grow in leaps and bounds.



Well, I'm trying to be optimistic.  New years resolution you see.  Usually it's to stop smoking, but i never manage.  Hopefully i can stick to this one! :-D
 

Offline jdiffend

Re: TeamAROS Bounty #23 AROS Kickstart Replacement ROM
« Reply #57 on: January 06, 2005, 03:11:10 AM »
The biggest problem I've found is that AROS was designed to be a hosted OS rather than a rommable OS.  At this point it makes more sense to make the Amiga AROS ROM more like the PC bios than the Amiga OS ROM.
 

  • Guest
Re: TeamAROS Bounty #23 AROS Kickstart Replacement ROM
« Reply #58 on: January 06, 2005, 03:13:39 AM »
Quote

jdiffend wrote:
The biggest problem I've found is that AROS was designed to be a hosted OS rather than a rommable OS.  At this point it makes more sense to make the Amiga AROS ROM more like the PC bios than the Amiga OS ROM.


Or maybe openfirmware-ish?

Is that possible/feasible/needed?
 

Offline bloodline

  • Master Sock Abuser
  • Hero Member
  • *****
  • Join Date: Mar 2002
  • Posts: 12113
  • Total likes: 0
    • http://www.troubled-mind.com
Re: TeamAROS Bounty #23 AROS Kickstart Replacement ROM
« Reply #59 on: January 06, 2005, 07:46:44 AM »
Quote

jdiffend wrote:
The biggest problem I've found is that AROS was designed to be a hosted OS rather than a rommable OS.  At this point it makes more sense to make the Amiga AROS ROM more like the PC bios than the Amiga OS ROM.


That's not quite true, great care has been taken to make sure that the AROS code rommable, in fact this requirement has (and still is) causing quite a few headaches :-D

I suggest you look at the x86 Native flavour as that has the AROS kernel loaded into RAM which is then protected and treated as ROM.

-Edit- Porting Openfirmware to the Amiga would be a brilliant and very popular idea though!!!