Poster | Thread |
Petah
| |
[POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 5-Dec-2019 23:56:06
| | [ #1 ] |
|
|
|
Regular Member |
Joined: 10-Mar-2003 Posts: 432
From: EU <3 ❤️ | | |
|
| Anyone who wants to run AmigaOS 4 on a legacy Amiga sporting a Motorola compatible 680x0 CPU has to upgrade their system, kitting it with a PowerPC accelerator board. This usually means resorting to 20 year old solutions, likely involving hardware from the (now defunct) German developer Phase 5. A lot of potential users are, however, deterred by the prices. Second hand boards are usually very, very expensive - prices north of €10 000 aren't unheard of, as this this example shows.
With the advent of new technical breakthroughs, a solution may be on the horizon. As so called FPGA-technology are getting cheaper and cheaper, creating a 100% compatible PowerPC board for Zorro-slots and the Amiga 500/Amiga 1200/CDTV proprietary expansion ports should be viable. Such a implementation will, in theory, give Amiga users a modern, cheap way of taking the step up to AmigaOS 4 without having to depend on Phase 5 hardware.
In this AmigaWorld.net Special Poll, you are kindly asked to raise your voice and give an explanation why there aren't (yet) any FPGA-based solutions commercially available on the Amiga-market. _________________ That'll Put Marzipan In Your Pie Plate, Bingo 💻 Pro-Amiga, 🌍 Pro-Globalism, 🍅 Pro-Vegan, 🛦 Pro-NATO & 🇪🇺 Pro-Joint EU Defense Intervention Initiative |
|
Status: Offline |
|
|
simplex
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 6-Dec-2019 0:42:38
| | [ #2 ] |
|
|
|
Cult Member |
Joined: 5-Oct-2003 Posts: 896
From: Hattiesburg, MS | | |
|
| @Petah
1) Didn't they only open up the architecture recently? (as in, a few weeks back?) Or did I misunderstand that announcement?
2) Are FPGAs fast enough to keep up with a PowerPC CPU? or do I misunderstand what you're getting at? _________________ I've decided to follow an awful lot of people I respect and leave AmigaWorld. If for some reason you want to talk to me, it shouldn't take much effort to find me. |
|
Status: Offline |
|
|
billt
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 6-Dec-2019 0:57:42
| | [ #3 ] |
|
|
|
Elite Member |
Joined: 24-Oct-2003 Posts: 3205
From: Maryland, USA | | |
|
| @simplex
Quote:
1) Didn't they only open up the architecture recently? (as in, a few weeks back?) Or did I misunderstand that announcement? |
Yes, that sounds about right. But there had been at least one beginning of an open-source PPC softcore some time back anyway (zCore). I don't know if it got to point of being useful or not, or just an academic experiment. I saved a copy to github a while ago, but I would guess the official release very recently is likely much better.
New release from IBM Team, softcore named Microwatt: https://github.com/antonblanchard/microwatt
Quote:
2) Are FPGAs fast enough to keep up with a PowerPC CPU? or do I misunderstand what you're getting at? |
Some may be, but some are several thousand $$ each chip. Are affordable ones fast enough? Not sure. We would need to start experimenting with place/route of the recently opened softcore code with various models and see what clock speed they come out working at... And even the fastest FPGAs will not give you a PPC core as fast as a T4240 or T2080 at speed. The FPGA overhead will get in the way of clock speeds that high. It works well for 68k, as you are looking at Motorola chips topping out at about 66MHz, a few have overclocked up to 100MHz if I recall. FPGas can get you more clock speed than that, but they cannot get you a couple GHz CPU core...Last edited by billt on 06-Dec-2019 at 09:18 AM. Last edited by billt on 06-Dec-2019 at 09:15 AM. Last edited by billt on 06-Dec-2019 at 09:14 AM.
_________________ All glory to the Hypnotoad! |
|
Status: Offline |
|
|
bison
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 6-Dec-2019 4:18:38
| | [ #4 ] |
|
|
|
Elite Member |
Joined: 18-Dec-2007 Posts: 2112
From: N-Space | | |
|
| @Petah
Tangentially related: An FPGA FPU for Tabor might turn out to be useful. I don't know if that's possible or not.
Last edited by bison on 06-Dec-2019 at 04:19 AM.
_________________ "Unix is supposed to fix that." -- Jay Miner |
|
Status: Offline |
|
|
logicalheart
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 6-Dec-2019 5:59:01
| | [ #5 ] |
|
|
|
Cult Member |
Joined: 2-Dec-2003 Posts: 696
From: Sandy, Utah. USA | | |
|
| A real PowerPC CPU is cheaper and faster than FPGA hardware and programming. The problem is the cost to engineer and develop the hardware and software to interface with the rest of the computer. The best solution might be a Tabor or X5000 computer with one of the classic Amiga chipsets on FPGA?
_________________ http://www.hostcove.com http://www.youtube.com/hostcove Sam460 : X1000 : X5000 |
|
Status: Offline |
|
|
OlafS25
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 6-Dec-2019 10:40:10
| | [ #6 ] |
|
|
|
Elite Member |
Joined: 12-May-2010 Posts: 6339
From: Unknown | | |
|
| @Petah
simply it makes no sense
There are still CPUs available that are faster than anything in FPGA. 68k was different because CPUs were no longer produced and CPU in FPGA is faster than all existing CPUs ever produced (I mean 68000er, not Coldfire). When there is no PowerPC any more available and FPGAs fast enough that might change. But still the question if there would be need and if there would be anyone doing it. |
|
Status: Offline |
|
|
PhantomInterrogative
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 6-Dec-2019 11:42:41
| | [ #7 ] |
|
|
|
Cult Member |
Joined: 10-Sep-2004 Posts: 809
From: The Interrogative Lair | | |
|
| @Petah
Why not a custom Risc V with the most used instructions from both 68k and PPC to ease emulation of these processors? _________________ I sold my SAM460ex lite... waiting for money to buy a Raspberry Pi... or a Classic A1000 with Buffee... or an A1222... and OS4.3 FE update 11 |
|
Status: Offline |
|
|
Deniil715
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 6-Dec-2019 14:21:18
| | [ #8 ] |
|
|
|
Elite Member |
Joined: 14-May-2003 Posts: 4236
From: Sweden | | |
|
| @logicalheart
Quote:
logicalheart wrote: A real PowerPC CPU is cheaper and faster than FPGA hardware and programming. The problem is the cost to engineer and develop the hardware and software to interface with the rest of the computer. The best solution might be a Tabor or X5000 computer with one of the classic Amiga chipsets on FPGA?
|
Yes, that would make sense. Classic as a "plugin" PCI-card to an NG machine rather than the other way around. Problem is that few people would buy it I suppose. Those who want the Classic, want the Classic shell/case even if the cost would be the same for, say Tabor+Classic-on-PCI with a vast improvement in performance.
Are there any portable PPCs in production? If not, it might be diffiult to make a PPC-accelerator small enough to fit in an A1200._________________ - Don't get fooled by my avatar, I'm not like that (anymore, mostly... maybe only sometimes) > Amiga Classic and OS4 developer for OnyxSoft. |
|
Status: Offline |
|
|
Kronos
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 6-Dec-2019 14:44:50
| | [ #9 ] |
|
|
|
Elite Member |
Joined: 8-Mar-2003 Posts: 2562
From: Unknown | | |
|
| @Deniil715
Adding an (FPGA) AGA to a NG system makes about as much sense as adding a FPGA PPC to an classic Amiga.
None at all if you wonder....
FPGA+NG:
What SW needs the chipset but still works in an NG OS (running on Phase5-HW or UAE)? Little to none I'll assume. You'd also have a hard time placing that PCI-based AGA at the right addresses. -> total waste of time, run UAE
"Classic"+FPGA-PPC: Putting an 68k core into the same FPGA will yield the same performance (at least) and will run far more SW. Unless you emulate a full Cyberstorm/BlizzardPPC you won't get an NG to run without active support by it's developers, what a waste of time when you can PPC 20x faster for 10€ that are either supported or in the same limbo (depending on choice of NG-OS). -> total waste of time, buy an XBox360 _________________ - We don't need good ideas, we haven't run out on bad ones yet - blame Canada |
|
Status: Offline |
|
|
coldacid
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 6-Dec-2019 16:42:09
| | [ #10 ] |
|
|
|
Member |
Joined: 27-Oct-2019 Posts: 39
From: Candinavia | | |
|
| An FPGA PPC with enough performance to match real chips would be prohibitively expensive for regular people like us. It'd actually be cheaper to source proper PPC or POWERx chipsets. The only thing an FPGA would gain is having a pre-built board to run it on, if one's available for the actual chip selected, whereas building a system around real Power processors would mean requiring the money and expertise to design and construct motherboard and all from the ground up. (Unless I'm mistaken, there's nothing like reference implementations like in the Mac clone days for modern PPC/Power chipsets.)
Re m68k or FPGA-m68k: I could see the use in a m68k board in a NG machine, but it would require pretty deep changes to how code for that architecture is run now and may not actually offer any performance gain over current 68k emulation on PPC. Definitely would be a fun hack, but not much more than that.
@Kronos Quote:
-> total waste of time, buy an XBox360 :P |
If I could install OS 4.1 on my 360, I would. It's not like I use it for games anymore! |
|
Status: Offline |
|
|
Kronos
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 6-Dec-2019 17:10:15
| | [ #11 ] |
|
|
|
Elite Member |
Joined: 8-Mar-2003 Posts: 2562
From: Unknown | | |
|
| @coldacid
Quote:
If I could install OS 4.1 on my 360, I would. It's not like I use it for games anymore! |
Kinda forgot that the 360 was the PPC one....
But the problem of not being able to install OS4.1 would still be the same with an FPGA-PPC.
_________________ - We don't need good ideas, we haven't run out on bad ones yet - blame Canada |
|
Status: Offline |
|
|
coldacid
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 6-Dec-2019 17:34:03
| | [ #12 ] |
|
|
|
Member |
Joined: 27-Oct-2019 Posts: 39
From: Candinavia | | |
|
| @Kronos
Yeah the problem isn't the hardware itself so much as no boot support for anything Hyperion hasn't blessed. Otherwise we'd be able to use PPC Macs for running 4.1 if we wanted, or any other Power based systems. |
|
Status: Offline |
|
|
NutsAboutAmiga
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 6-Dec-2019 19:59:10
| | [ #13 ] |
|
|
|
Elite Member |
Joined: 9-Jun-2004 Posts: 12817
From: Norway | | |
|
| @Petah
As everyone else pointed out it be a new platform that needs to be supported.
Unless you like to remake a “Classic PPC” (BlizzardPPC+A1200) or do a Sam460.. I guess you want something in-between the two really, so you can have Paula sound, CAA/CIAB timers and PCIe slot’s for modern graphics card, and USB3 controller, SATA controllers. And for that I think you end up with slower PPC then on you can get as a shelf item.
AGA is not necessary to run Classic software that is well written, frankly AGA was almost outdated when it was born, the issue we run into is the blitter, timing, and sound, and lack of joystick ports, this are the biggest problems for running classic software, in my option… Yes AGA in FPGA might be great too, but most of the time you want promote it to RTG anyway.
I have small dream that RetroMode.library can form bases for system friendly API for all Amiga systems, so you can do Amiga graphics in a more system friendly way… but I don’t see any point in doing it for Amos Kittens, when Amos Pro X is being worked on, I see no value in competing on that. But retroMode.library can become a new kind of SDL for all Amiga systems.
Last edited by NutsAboutAmiga on 06-Dec-2019 at 08:17 PM. Last edited by NutsAboutAmiga on 06-Dec-2019 at 08:04 PM. Last edited by NutsAboutAmiga on 06-Dec-2019 at 08:02 PM.
_________________ http://lifeofliveforit.blogspot.no/ Facebook::LiveForIt Software for AmigaOS |
|
Status: Offline |
|
|
Samurai_Crow
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 6-Dec-2019 20:17:33
| | [ #14 ] |
|
|
|
Elite Member |
Joined: 18-Jan-2003 Posts: 2320
From: Minnesota, USA | | |
|
| @Petah
FPGAs are likely to clock below 100 MHz on consumer grade PPC designs. It makes no sense to use a sub BlizzardPPC performance card that costs as much as an X1000 did.
Apollo team had to greatly simplify the MMU architecture to get the 68080 faster than a 68060, after all. PPC would also be the same speed on that hardware unless you could simplify something. |
|
Status: Offline |
|
|
billt
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 6-Dec-2019 20:18:48
| | [ #15 ] |
|
|
|
Elite Member |
Joined: 24-Oct-2003 Posts: 3205
From: Maryland, USA | | |
|
| Quote:
coldacid wrote:
Yeah the problem isn't the hardware itself so much as no boot support for anything Hyperion hasn't blessed. Otherwise we'd be able to use PPC Macs for running 4.1 if we wanted, or any other Power based systems. |
One of the benefits of an FPGA based system over an Xbox360 or PS3, is that you can make the FPGA system look like an already supported Amiga platform, such as CSPPC based classic or Sam460ex, just like WinEUA and Qemu have done. Then an existing OS4 install CD doesn't know the difference, and you don' t need to wait for Hyperion's blessing or driver coding...
It can be tricky to make hardware to fit the old software, but UAE, Qemy and Minimig are examples of that having been done in different ways to build from.
So, the Microwatt PPC softcore is based on Wishbone bus. So is the ao68000 and aoOCS FPGA stuff that is comparable to Minimig, but these ones also use Wishbone bus...
https://github.com/antonblanchard/microwatt
https://github.com/alfikpl?tab=repositories
OK, so no AGA, and only a 68000 style, but a starting point... Look to tg68 and the Suska 68K30 for ways to uupdate the 68k CPU part, and minimig-AGA to enhance that stuff, or go the other way and make a minimig->Wishbone bridge and connect the Microwatt softcore on the other side of that, and to WinUAE for how the PPC fits into things...
Or, look to make an FPGA clone of Sam460/440 of some sort, without any Classic stuff in it, looking to the CPU databooks and Qemu for inspiration..._________________ All glory to the Hypnotoad! |
|
Status: Offline |
|
|
Lou
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 6-Dec-2019 20:52:02
| | [ #16 ] |
|
|
|
Elite Member |
Joined: 2-Nov-2004 Posts: 4169
From: Rhode Island | | |
|
| Quote:
This project is an attempt at reimplementation of WarpOS for Sonnet Crescendo 7200 and other PPC PCI cards.
The main part of the project is a library, which aims at API and ABI compatibility with WarpOS powerpc.library.
This project is in a beta stage of development.
|
https://github.com/Sakura-IT/SonnetAmiga |
|
Status: Offline |
|
|
NutsAboutAmiga
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 6-Dec-2019 22:00:11
| | [ #17 ] |
|
|
|
Elite Member |
Joined: 9-Jun-2004 Posts: 12817
From: Norway | | |
|
| @Lou
But is not hardware reimplementation, this software reimplementation, this because there is software API, that WarpUP / PowerUP programs talk to, that it how this work.. but old Amiga programs, don’t use software API’s they poke the hardware directly in many cases, even so there is bag of tricks that can be used, MMU trapping, and software patching and so on, but all this stuff end up making the system unstable and unpredictable, it’s better to not too. It’s better to run the sh*ty software in UAE, if they can’t play by the operating systems rules.
Barking down the FPGA route, you can do a lot more to fix incompatibility issues, but I think ExecSG team need to make special version of ExecSG that knows how to boot on operating system like that, without the 68K processor, but with some or all of the classic chipsets. what drivers need to loaded, and so on... can't expect the users to know how to do this... so its not that simple.
Last edited by NutsAboutAmiga on 06-Dec-2019 at 10:25 PM.
_________________ http://lifeofliveforit.blogspot.no/ Facebook::LiveForIt Software for AmigaOS |
|
Status: Offline |
|
|
Lou
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 7-Dec-2019 0:40:21
| | [ #18 ] |
|
|
|
Elite Member |
Joined: 2-Nov-2004 Posts: 4169
From: Rhode Island | | |
|
| @NutsAboutAmiga
For complex addon hardware - you'd be foolish to bang the hardware. It's probably not even documented on how...
If you click thru some links you'll see the software compatibility is really high. |
|
Status: Offline |
|
|
OlafS25
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 7-Dec-2019 10:11:28
| | [ #19 ] |
|
|
|
Elite Member |
Joined: 12-May-2010 Posts: 6339
From: Unknown | | |
|
| @NutsAboutAmiga
even new software (expecially games) directly hit the hardware
for some that is even the fun part of developing and the reason they do it (financially even "classic" is not big enough, of course bigger than the NG platforms)
the library is interesting... I only see the problem that NG platforms based on graphic cards and what you call "classic" is very different
In my view it is difficult to develop something that is both running fast on NG and classic
Most (who want to develop something for amiga) will not use such a library and prefer to bang the hardware directly Last edited by OlafS25 on 07-Dec-2019 at 10:40 AM.
|
|
Status: Offline |
|
|
NutsAboutAmiga
| |
Re: [POLL] What's stopping FPGA-based PowerPC accelerators? Posted on 7-Dec-2019 11:04:18
| | [ #20 ] |
|
|
|
Elite Member |
Joined: 9-Jun-2004 Posts: 12817
From: Norway | | |
|
| @OlafS25
So the idea, is that different version of the library, that is optimized for hardware or the OS it runs on, so you are using the same library, but the API is the same, so that any game that use see the same, Of course it packs all the stuff you need as game developer.
For sound we have OpenAL and AHI, so we don't need to worry about audio.
The bigger issue is something like this should have been made in 1994, it might have had an impact, but now in 2019, and no games that use it, so somehow you have convince people that this is better way to do it.
If we imagine the best developers, started working on it optimized it like crazy, so that it be easier to use it, provided better performance then if you wrote the routines yourself. And everyone know about it, then I guess, something like this for graphics can be nice way to remove the hardware dependencies. And I know there is need for this kind of thing, because lot of libraries that do exists (C2P libs) are not worked on have bugs are not open source, so no one will fix any of the bugs. and often they are statically linked so you can replace the library for different OS or Hardware configurations. without having the source code of the game (or program)... so you recompile it. we see that problem with SDL games ported to AmigaOS3.x AGA. Yes SDL is cross platform. but does make game cross platform on Amiga, myabe on Linux where they use .so objects, but there they don't need it. I think PowerSDL does use amiga .libraries but it runs only on MorphOS. Last edited by NutsAboutAmiga on 07-Dec-2019 at 11:44 AM. Last edited by NutsAboutAmiga on 07-Dec-2019 at 11:12 AM. Last edited by NutsAboutAmiga on 07-Dec-2019 at 11:09 AM. Last edited by NutsAboutAmiga on 07-Dec-2019 at 11:08 AM. Last edited by NutsAboutAmiga on 07-Dec-2019 at 11:07 AM.
_________________ http://lifeofliveforit.blogspot.no/ Facebook::LiveForIt Software for AmigaOS |
|
Status: Offline |
|
|