Overview Features Instructions Performance Forum Downloads Products OrderV4 Reseller Contact

Welcome to the Apollo Forum

This forum is for people interested in the APOLLO CPU.
Please read the forum usage manual.
VISIT APOLLO IRC CHANNEL



All TopicsNewsPerformanceGamesDemosApolloVampireAROSWorkbenchATARIReleases
Documentation about the Vampire hardware

User Contribution to Development On Vampire?

Greg Thomas

Posts 24
08 Dec 2018 03:50


I'm looking forward to running an A1000 / Vampire. Also I want to do some development of my own with open cores. Maybe have DE10-nano MISTer as hacking development board before pushing perfected softcores to the Apollo board.

Amiga usrs are fans of efficiency and elegance in computing. Amiga multiprocessing though old, is still often smoother than a 3Ghz PC of literally 100 times its speed. Hopefully the Atari and 68K Mac users will join us for the ride.

Less than half of the 110K LE's will be used by the V4 Vampire leaving a lot of room for experimentation.

That experimentation would be useful to put back into the mainline so Amiga can move forward:
  from 30 years ago - (real ASIC 68k's + planer ECS/AGA),
  to 20 years ago - (080 with AMMX + SAGA/RTG),
  10 years ago - (hyperthreading, multicore + planer/RTG/2D SAGA + DSP),
  eventually to the modern - (many cores, 4K video),

Then we can move forward to the future of cheap low power massively parallel computing in the home.
   
When end users can contribute to development then we can help the Apollo guys to help us :-)
   
It would be nice to see what it takes to attach to:

SATA
  EXTERNAL LINK    EXTERNAL LINK    EXTERNAL LINK 
PCI
  EXTERNAL LINK 
PCIe
  EXTERNAL LINK    EXTERNAL LINK 
Open GPU (Kazan)
  EXTERNAL LINK 
HyperTransport (make Athlon Bridgeboard)
  EXTERNAL LINK    EXTERNAL LINK 

If Wishbone bus or TileLink bus are available inside the V4 and we can develop inside the unused space... It will be like the old days of nevously writing your own alterations to the startup-sequence of a Workbench floppy disk! ;-)
   
  TileLink bus seems to be the standard that the RISC V crowd will settle on in place of Wishbone bus.
   
  TileLink: A free and open-source, high-performance scalable cache-coherent fabric designed for RISC-V
  EXTERNAL LINK 
Diplomatic Design Patterns: A TileLink Case Study
  EXTERNAL LINK 
Amiga makes me excited to put all my time into learning how to do it.

The Apollo guys are great role models to admire! :-)

A good license requirement would be that contributed cores put onto the Apollo site are free for Apollo to sell commercially without any license headaches.


Vojin Vidanovic

Posts 1572
16 Dec 2018 19:14


While people love the most modern feats I am for a step by step evolution. As example, if V4 could reach  2 x PCI slots (or PCI and Zorro III), rudimentary h/w 3D and SATA I would be most satisfied. Remember that AmigaOS needs to support it too, and here some basis exists.

PCI-E and HT would be best left alone for some future incarnations.


Mr-Z EdgeOfPanic

Posts 185
17 Dec 2018 05:29


@Greg Thomas

The V4 is 76K LE not 110, and i'm sure we'll fill the whole 76K eventually :)


Gunnar von Boehn
(Apollo Team Member)
Posts 4335
17 Dec 2018 08:14


Your post has a lot enthusiasm.

Lets give some info to clarify a few things.

A) APOLLO 68080 architecture does already support SMP/MULTICORE
Its caches are designed to be fully coherent.
But MULTICORE is right now not usable on AMIGA OS, because lack of OS support.

B) PCI
We already have a good working PCI core.
PCI was used on the NATAMI.
The main bottleneck of PCI are missing drivers for cards.
To keep cost for the V4 lower, PCI is not on the V4 standalone.
The Vampire card have their own low cost expansion port, which allows for example to add WIFI upgrades.

C) SATA
Inexpensive SATA adapters are available
Using SATA over an inexpensive adapter gives the users a lower end cost than putting it on the V4 board.

D) BUS
APOLLO does support industry standard bus protocols already

E) GPU
The Team is already working on a new GPU core for the Vampire.

The V4 tries to be a sensible compromise of features and cost.



Vojin Vidanovic

Posts 1572
17 Dec 2018 11:37


Gunnar von Boehn wrote:

    .
    But MULTICORE is right now not usable on AMIGA OS, because lack of OS support.
   

   
    Understood, even OS4 has failed to move to that area with SMP scheduler. What would be the best candicate "for future" - backporting AROS x64 to AROS ,m68k SMP or m68k Linux that does not have it, but could be backported Kernel (and MMU 68k support needed)?
   
   
Gunnar von Boehn wrote:

    B) PCI
    We already have a good working PCI core.
    PCI was used on the NATAMI.
    The main bottleneck of PCI are missing drivers for cards.
    To keep cost for the V4 lower, PCI is not on the V4 standalone.
    The Vampire card have their own low cost expansion port, which allows for example to add WIFI upgrades.
     

   
    I do understand use of standard expansion connector for FPGAs is a compromise. V5+ could have at least one PCI - Radeon 9250, Sound Blaster etc. drivers do exist (thanks to Mediator and other older solutions). I am sure some V2/V4 expansions will appear, but at least one PCI slot would enable Voodo or Radeon 9xxx cards, thus enabling Warp3D quickly and completing 68k AmigaOS softwarebase, alongside MMU. WOS/MOS 1.x/PUP software will be seen as a bit non standad solution, and one day 603e emulator (backport of UAE solution?) could do even that, but that is for 300Mhz 080 to sweat :-)
   
   
Gunnar von Boehn wrote:

    C) SATA
    Inexpensive SATA adapters are available
    Using SATA over an inexpensive adapter gives the users a lower end cost than putting it on the V4 board.
     

   
    I would not mind a bit more expensive Vamp with JUST SATA. Easuer HDD,DVD-RW and SDD availiability makes making a new system a lot easier, plus some expected performance boost with decent driver.
   
   
Gunnar von Boehn wrote:

    E) GPU
    The Team is already working on a new GPU core for the Vampire.
     

   
    Great news, will it fit V4 and how does it fare compared to Voodo 3 3000?
   
   
Gunnar von Boehn wrote:

    The V4 tries to be a sensible compromise of features and cost.
     

   
    Fully understood, but + model with more backblated memory card reader, SATA, at least +2 USBs, and at least one more standardized expansion (be Zorro or PCI) would be a great bonus and a lot of users would opt for it. If Wi-Fi solution could be bundled with it, that is one cute and expandable beast.
   


Greg Thomas

Posts 24
17 Dec 2018 16:51


Gunnar von Boehn wrote:

  A) APOLLO 68080 architecture does already support SMP/MULTICORE
  Its caches are designed to be fully coherent.
  But MULTICORE is right now not usable on AMIGA OS, because lack of OS support.

Fabulous news! More than Commodore ever gave us after A3000!
So L1 cache coherency or does it have L2 cache too, which is the bit with the cache coherency? how many cores can be coherent with cache? 2 like hyper-threading or even more?

Only a matter of time before we have multi-core Workbench. AROS already has working SMP, but only on the nightly builds at this stage. Other OS's such as BSD, Haiku, can use it now. The main thing is the provision. Like how the A1200 has the clockport to allow possible things which were impossible otherwise.

Gunnar von Boehn wrote:
  B) PCI
  We already have a good working PCI core.
  PCI was used on the NATAMI.
  The main bottleneck of PCI are missing drivers for cards.
  To keep cost for the V4 lower, PCI is not on the V4 standalone.
  The Vampire card have their own low cost expansion port, which allows for example to add WIFI upgrades.

That is fantastic. But oh yes the silicon real estate inside the FGPA will be the limitation. Like having 50KB to write to a WorkBench floppy Boot disk.

Gunnar von Boehn wrote:
  C) SATA
  Inexpensive SATA adapters are available
  Using SATA over an inexpensive adapter gives the users a lower end cost than putting it on the V4 board.
 
  D) BUS
  APOLLO does support industry standard bus protocols already
 
  E) GPU
  The Team is already working on a new GPU core for the Vampire.
 
 
  The V4 tries to be a sensible compromise of features and cost.

This is all great news.
I may have mislead my point a little however. If there were a GPIO expansion headder for the unused IO's then stuff can be added later subject to FPGA core space. Maybe its PCI and a ribbon cable comes from Vampire expansion header to a "PCI Mediator" type board.

If Commodore never put the A1200 clock header then it could never have had the Squirrel SCSI and whatever other ingenious hacks were made.

mSATA, PCIe WIFI cards, and of these kind of things would be great. Just having that cheap GPIO port or edge connector is the difference between complex development work and impossible dream.

But understood that specifically the V4 its all a matter of best use of limited resources for now :-)

All great news to have the Amiga we would have dreamed to have in 2004! Good even in 2014, we are catching up quickly from our 1994 technology Amiga's :-D
2024 here we come in a couple of generations.

I already have two old time Amiga people who want a stand alone! There will be many more.


Gunnar von Boehn
(Apollo Team Member)
Posts 4335
17 Dec 2018 21:49


Greg Thomas wrote:

  how many cores can be coherent with cache? 2 like hyper-threading or even more?
 

APOLLO does support Hyper-threading already - today.

In addition to this the caches of APOLLO are designed with coherency in mind. This coherence design allows Multicore support.
The cohherence is also why APOLLO is compatible to 68k "selfmodifying" code - as APOLLOs-Caches do support this as side-effect of the coherence snooping.
 
 
 
Greg Thomas wrote:

  Only a matter of time before we have multi-core Workbench.
 

I think it could be done for the real AMIGA OS.
Our advantage is that we control the whole system - the Chipset, and the CPU. We can add some tricks here. ;)
 
 
 
Greg Thomas wrote:
 
  If there were a GPIO expansion headder for the unused IO's then stuff can be added later subject to FPGA core space.
 

The V4 has such ports.
You can use them today to add a WIFI card.
They might be used tomorrow to add a legacy floppy.

 


Greg Thomas

Posts 24
23 Dec 2018 11:47


This is really great news. So does the cache mark an area of memory as paged to cache and locked to the core which requested, but if another core wants that memory then the copy back can be dropped until the new core is done with that memory page. Is that how it works? How many levels of cache are there? For how many cores can this work well?

OK so that card is future proofed as far as the capabilities of the chip go. More good news :-)

Will the end user be able to do anything, or does the FPGA compile work as an all or nothing situation like writing a .DMS floppy disk. You can't add a bit on the end to any unused space? And so would have to compile the entire thing each time?

Well what about the GPIO's? Can we configure those somehow or will those address lines be unconnected and waiting for future core utilisation?

Whatever the answers, we all wish we had this in 1993! :-)


Alan Haynes

Posts 132
24 Dec 2018 07:38


Gunnar von Boehn wrote:

 
Greg Thomas wrote:
 
  If there were a GPIO expansion headder for the unused IO's then stuff can be added later subject to FPGA core space.
 

  The V4 has such ports.
  You can use them today to add a WIFI card.
  They might be used tomorrow to add a legacy floppy.
 

Hi Gunnar and Merry Christmas to you and the team and of course your families from Australia.
Really good news in Renaud's report especially about the full implementation of the CIA's. Does this mean as per what you said above that we might even be able to connect a real Amiga floppy Drive or even a normal PC one that will read Amiga Disks?
If that is the case and from what I read then it is a possible future opportunity. Would that also mean the potential to use a Gotek instead of a floppy?
I have on order one of Steve Jones's cases for the very purpose of installing the V4 so am looking forward to this. Am quite happy to wait for it to be ready in your time. Currently I have an a500 V2 plus in an a500+ but will soon transfer to my A2000.

Thank you to you and the team for all the hard work and sacrifice to bring this New Amiga to market.
Merry Christmas and a Happy New Year to all.



Gunnar von Boehn
(Apollo Team Member)
Posts 4335
24 Dec 2018 19:48


Alan Haynes wrote:

  Hi Gunnar and Merry Christmas to you and the team and of course your families from Australia.

Merry Christmas to you too and to all Amiga fans!

posts 10