Overview Features Coding ApolloOS Performance Forum Downloads Products Order Contact

Welcome to the Apollo Forum

This forum is for people interested in the APOLLO CPU.
Please read the forum usage manual.
Please visit our Apollo-Discord Server for support.



All TopicsNewsPerformanceGamesDemosApolloVampireAROSWorkbenchATARIReleases
Information about the Apollo CPU and FPU.

Since AROS Approaches SMP - Can Apollo Do MP?page  1 2 

Cyber Gorf
(Needs Verification)
Posts 39/ 1
31 May 2017 14:17


I know it is far from stable and maybe not even usable right now, but (s)MP on Aros64 is coming!

How difficult would it be to get more than one Apollo-Core running in the fpga?
Could special Inter-Processor-Signals help/assist smp-exec and the scheduler?

Apart from SMP there is always a simpler MP approach, by running a full OS-kernel on each core, with its own region of memory - and communication via network-style message passing and/or shared memory.


Vojin Vidanovic

Posts 770
31 May 2017 14:29


Thats even "in galaxy far far away question". I am sure others will answer more technical, but this should be that "what if"
 
  Apollo core FPGA design is made to be scalable,and in the very end of Vampire500 World Premiere subject of future ASIC is explored. But that is the goal after many many core revisions and FPGA Vampires largely succesfull and on sale for a500,a600,a1200 and standalone.
 
  Then MAYBE a donor could be found to fund the ASIC redesign and production, and MAYBE MAYBE after its success a need for multicore
  will exist.
 
  And then, YES, Backporting state of SMP from AROSx64 to AROS 68k (which is kind of recompile and bugfix) would THEN be a solution.
 
  Currently, we are at polishing SAGA chipset and having more Amiga delers with Vampire A500/A600. Its "few lightyears away" and we dont have a Warp engine yet :)


Cyber Gorf
(Needs Verification)
Posts 39/ 1
31 May 2017 14:38


@Vojin:
No - I am not talking about ASIC!
Since Apollo-Core is in VHDL it is "easy" to put more than one core on the FPGA (more or less copy-paste) - if your FPGA is big enough.

The hard part would be, to make it somehow useful!

So clearly multicore in FPGA comes years before any ASIC!


Gunnar von Boehn
(Apollo Team Member)
Posts 6197
31 May 2017 14:39


Ye, APOLLO is designed for SMP.



Cyber Gorf
(Needs Verification)
Posts 39/ 1
31 May 2017 14:44


Gunnar von Boehn wrote:

Ye, APOLLO is designed for SMP.
 

That is great!

Could this be considered for a standalone-vampire? Required FPGA size?


Vojin Vidanovic

Posts 770
31 May 2017 14:58


Cyber Gorf wrote:
So clearly multicore in FPGA comes years before any ASIC!

If that ever comes to the table ("big FPGA standalone"), as expressed at other place, some ... Virge3D or 3Dfx Voodo or Radeon 8500 core could be nice too, just to bypass all 3D instructions development, if possible and bring Warp3D/MESA at least in 90s state.


Samuel Crow

Posts 424
31 May 2017 15:01


@Cyber Gorf
A second core 68080 is possible in a large FPGA but I don't think it would fit in a Cyclone series FPGA.  So it is a slightly trick question:  Apollo can go multicore but not in the current generation of Vampire cards.

@thread
As for what it would be used for, back in the days of the NatAmi prototypes, we discussed things like upgrading the blitter to a full fledged slave core that would mix audio as well as fancy alpha-blending effects.  You could do that already using one core but background processes would be a load off the first CPU.


Cyber Gorf
(Needs Verification)
Posts 39/ 1
31 May 2017 15:14


@Vojin

Please stop hijacking!
You tried to bring up this subject of yours in several threads already - but nobody except you is interested.
This thread is not about 3D. Please stop it.



Cyber Gorf
(Needs Verification)
Posts 39/ 1
31 May 2017 15:20


@Samuel

Thanks. I was almost certain, that right now vampires would be to small.
You confirmed my suspicion.

So the question for a large-fpga standalone with two Apollo-Cores is still valid.

Sure it is always nice to see some well-designed upgrades to copper and blitter.
Looking forward to this as well.


Wawa T

Posts 695
31 May 2017 16:46



Could this be considered for a standalone-vampire? Required FPGA size?

right now aros smp capabilities are disabled on amiga-m68k target for compatibility reasons. it remains to be seen how would that play off. so far no hardware is available for testing so its all highly hypothetical.




Cyber Gorf
(Needs Verification)
Posts 39/ 1
31 May 2017 16:59


wawa t wrote:

 

  Could this be considered for a standalone-vampire? Required FPGA size?
 

  right now aros smp capabilities are disabled on amiga-m68k target for compatibility reasons. it remains to be seen how would that play off. so far no hardware is available for testing so its all highly hypothetical.
 

 
  The chicken or the egg ....
 
  once a FPGA with two Apollo-Cores exists, re-enabeling of SMP can begin.
  Aros-devs and Apolloteam need to work together on that.
 
  How about adding 2 (or more) extra "shared"-registers to the 68080?
  Lets call them A and B for now. So CPU1 could write to A1, which is B2 on CPU2 and the oder way round..
 
  That could be a nice way to tell exec, or the scheduler about messages or new tasks etc...
 


Vojin Vidanovic

Posts 770
03 Jun 2017 11:42


Cyber Gorf wrote:
  That could be a nice way to tell exec, or the scheduler about messages or new tasks etc...

Its nice to think ahead (I do the same in couple of "unexplored" areas), and to point it out for planning sake, but seems real expanded core isnt anz time now.

Stadalone is likely to also come with finished v2 product, just with SAGA over DIGITAL-VIDEO, 14-bit SAGA audio, SD card boot and if we get FPU and some polished AROS package - thanks to larger FPGA and CURRENT EFFORTS - that would be great result. That is the end of current roadmap, and only after that all done and marketed, we could expect some new Apollo roadmap.

Buying current Vampires is only way to go ahead. I plan standalone for next year, dropping any other hardware related plans.

AROSx86 SMP and couple of CPUs is soo ahead, we need at least single core 500Mhz Apollo (or 1gHz) and APPS THAT DO NEED THAT KIND OF POWER.

I would prefer enhencements to stability and single task crash isolation. And some productivity apps that would "sweat" the standalone and single core ASIC designs when done. Then it come sto table.

Options are open, we just need to journey so long, beyond 2020. Like in 90s Sci-Fi programmes.

You are well aware this is all hobbiest effort with no real companies, legal entities, factories ... but with strong coomunity drive and good people.


Cyber Gorf
(Needs Verification)
Posts 39/ 1
03 Jun 2017 18:33


Vojin Vidanovic wrote:

  Its nice to think ahead (I do the same in couple of "unexplored" areas), and to point it out for planning sake, but seems real expanded core isnt anz time now.

Absolutely. This is just some talk about a distant future ...
 
 
  Buying current Vampires is only way to go ahead. I plan standalone for next year, dropping any other hardware related plans.

I will definitely buy the standalone as soon as it is out there.
Same goes for a A3000/4000 version - finally AGA on my A3000 :-))


  I would prefer enhencements to stability and single task crash isolation. And some productivity apps that would "sweat" the standalone and single core ASIC designs when done. Then it come sto table.

Well - the single address-space design of the OS is prone to instabilities... even Aros/OS4/MorphOS could not change that... but it has its advantages as well - mainly incredible speed and low overhead.

A multicore-design could help there a litte bit:
running a new instance of Exec on every core with its own memory-space.
One instance my crash and reboot while the others are still live and kicking.
Think of it as running XEN or KVM or several VirtualBoxes ...

So you start all the stage Apps in the first instance and for testing, playing you start the programs in the next instance and so on...



Cyber Gorf
(Needs Verification)
Posts 39/ 1
03 Jun 2017 18:34


Vojin Vidanovic wrote:

    Its nice to think ahead (I do the same in couple of "unexplored" areas), and to point it out for planning sake, but seems real expanded core isnt anz time now.

 
  Absolutely. This is just some talk about a distant future ...
   
 
 
    Buying current Vampires is only way to go ahead. I plan standalone for next year, dropping any other hardware related plans.

 
  I will definitely buy the standalone as soon as it is out there.
  Same goes for a A3000/4000 version - finally AGA on my A3000 :-))
 
 
 
    I would prefer enhencements to stability and single task crash isolation. And some productivity apps that would "sweat" the standalone and single core ASIC designs when done. Then it come sto table.
 

 
  Well - the single address-space design of the OS is prone to instabilities... even Aros/OS4/MorphOS could not change that... but it has its advantages as well - mainly incredible speed and low overhead.
 
  A multicore-design could help there a litte bit:
  running a new instance of Exec on every core with its own memory-space.
  One instance may crash and reboot while the others are still live and kicking.
  Think of it as running XEN or KVM or several VirtualBoxes ...
 
  So you start all the stable apps in the first instance and for testing or playing you start the programs in the next instance and so on...
 


Wawa T

Posts 695
04 Jun 2017 03:11


>Well - the single address-space design of the OS is prone to instabilities... even Aros/OS4/MorphOS could not change that... but it has its advantages as well - mainly incredible speed and low overhead.

thats what it boils down to.


Gregthe Canuck

Posts 274
04 Jun 2017 03:26


*If* AROS keeps improving its support of the Vampire/Apollo systems I can see it and the Apollo core growing up together.

That is the nice part with the open source operating system, if/when the Apollo team decides to start experimenting with multiple cores some basic support should be there in the O/S to start from.

Everyone just needs to be patient. Very hard for sure. Pent-up frustration from 20 years of stagnation is boiling up.

Cheers!



Wawa T

Posts 695
04 Jun 2017 07:58



Everyone just needs to be patient.

no. people should help make it happen if they want that. posting on forums doesnt count.


Gregthe Canuck

Posts 274
04 Jun 2017 08:22


Wawa - please don't bring a grating tone into these discussions.
 
There are going to be people with certain skillsets that will be able to help out given time and motivation. Others will be able to support with their wallets by purchasing products and/or contributing towards bounties. This is the reality.
 
In my case I am trying to contribute to thoughtful discussions as time allows and plan to contribute towards bounties that seem appropriate. At some point I will be buying Vampire hardware as well.
 
I am up to my armpits in alligators with my own software business. I am managing a team of developers actually being paid out of my and my partner's own pockets. I spend more on software/manpower than Hyperion, A-Eon, and all the other Amiga companies put together. I understand probably better than most the issues that go along with getting a platform going and making the hard choices where to spend time, effort and capital.
 
In short, be a bit more realistic.
 


Kalamatee (AROS)

Posts 15
05 Jun 2017 18:28


Sorry to be the bearer of bad news - but there won't be support for smp in the m68k build ever. The m68k platform has no provision for working in an smp-like environment, and changes would make it binary incomparable with existing amigaos/aros for m68k.

If the vampire/apollo team create some "smp like" m68k core, it would need its own custom n68k flavour of aros, and most m68k apps won't work on it.


Gunnar von Boehn
(Apollo Team Member)
Posts 6197
05 Jun 2017 22:38


Kalamatee (AROS) wrote:

Sorry to be the bearer of bad news - but there won't be support for smp in the m68k build ever.

This is not a problem, we knew this and this is also fully OK.



posts 29page  1 2