With Phase 5 not looking likely. I was wondering how hard it would be for you cleaver FPGA people out there to make a FPGA PPC accelerator. I'm guessing very hard.

_________________
A1200 ACA1232 128MB Indivison MkIICr
Elbox empty Power Tower
RPi3 Emulating C64 ZX Atari PS BBC
Wii with Amiga emulation
Vampire v4 SA

@g01df1sh

The real question is: Why??
There are PPC CPUs that are very fast, why HW-emulate??

The reason there are 68k in FPGA is because there are no fast 68k CPUs, but fast FPGAs.

The best way to make a new Classic PPC accelerator would be to use a SoC of some kind. Like the one in Tabor or Sam.

_________________
- Don't get fooled by my avatar, I'm not like that (anymore, mostly... maybe only sometimes)
> Amiga Classic and OS4 developer for OnyxSoft.

I'd second that guess. Apparently, a 604 PPC processor has about 3.6 million transistors, the G5 58 million and a 68030 has about 273,000, so there's a massive difference in recreating something from the 68k family vs PPC, at least ten times the complexity. I think boards like the MiST and Vampire already utilize most of their FPGA chip for their recreations of vintage 68k systems (though they re-create more than only the CPU core).

I'd say software JIT emulation is a much better method, if for some reason you can't use a modern, existing PPC processor that's still in production.

Last edited by AdvancedFollower on 13-Feb-2018 at 02:50 PM.
Last edited by AdvancedFollower on 13-Feb-2018 at 02:49 PM.

@AdvancedFollower

So we have next to no chance of seeing new PPC cards for Classic Amigas

_________________
A1200 ACA1232 128MB Indivison MkIICr
Elbox empty Power Tower
RPi3 Emulating C64 ZX Atari PS BBC
Wii with Amiga emulation
Vampire v4 SA

@g01df1sh

I'll put aside the technical issues that others are explaining and point out something else:

The real problem is that Amiga is living in an abandoned market. The choices are basically between Freescale's/NXP's "underpowered" processors designed primarily for the embedded market, though from what I'm told PPC is gradually dying even there, and IBM's "overpowered" processors that no one can afford.

Each member of the AIM alliance separately made a conscious choice to abandon the desktop market. Apple's choice was made with much fanfare, but it was precipitated by Motorola/Freescale's earlier and quieter choice to focus increasingly on the embedded market, while IBM never really took their eyes off the server, supercomputer, and AI market. IBM dallied with the Cell processor idea briefly, like all large companies, eventually stuck with their golden goose.

For Apple this panned out, but for Freescale and IBM this seems to have been a dumb, dumb mistake: ARM has so completely overwhelmed the embedded market that even Freescale/NXP is abandoning PPC, whereas Intel and AMD have leveraged their economies of scale have into significant inroads into IBM's preferred markets.

This may sum it up best: An embedded engineer I spoke to about this last fall said that unlike ARM, PPC's makers failed to listen to their customers, and that's why PPC is vanishing. Whether he's right is another story, but from my admittedly lay vantage point, even Arduino seems to have a brighter future than PPC in the embedded market, if only because it seems to attract more attention. (This would be a good place for someone like Beans to step in & confirm or correct.)

The funny thing is that 10 years ago it really looked as if PPC had found new life: all the major video game consoles had moved to that platform in one way or another.

A healthy company like Microsoft, Nintendo, or Sony can perform a whiplash switch from one architecture to another because they designed well from the get-go: so, Playstation went from Cell to AMD's Jaguar; XBox went from some PowerPC to Jaguar; and Nintendo went from PowerPC in the Wii's to ARM in the Switch.

Amiga, alas, hasn't been managed by a healthy company for about 3 decades.

_________________
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.

@simplex

There is a new PPC accelerator on the way. It is tied to Sonnet PPC Project.

http://eab.abime.net/showthread.php?t=76633

@simplex

A very good summary, except the last line. Amiga has never been managed by a healthy company. Even when Commodore was making money they were doing stupid things.

_________________
"Unix is supposed to fix that." -- Jay Miner

@g01df1sh

Quote:

So we have next to no chance of seeing new PPC cards for Classic Amigas

http://eab.abime.net/showthread.php?t=76633&page=87

@bison

Quote:

A very good summary, except the last line. ...Even when Commodore was making money they were doing stupid things.

Point well taken (sitting on the OCS, "Amiga Jr."), but they were at least making money, which was more the point..

_________________
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.

@g01df1sh

Quote:

g01df1sh wrote: With Phase 5 not looking likely. I was wondering how hard it would be for you cleaver FPGA people out there to make a FPGA PPC accelerator.

If there would have been a 68060 with 400MHz for less than 50$, no one would ever ask for a Vampire4

There are a lot of cheap PPC CPUs around, that could beat most FPGA implementations of the PPC ISA in price and speed.


For example here:
https://www.men.de/products/discontinued-products/ek6n/
https://www.nxp.com/products/processors-and-microcontrollers/power-architecture-processors/mpc5xxx-55xx-32-bit-mcus/mobilegt-51xx-52xx/evaluation-board-for-the-mpc5200b-processor:Lite5200B
http://www.hitex.co.uk/hardware/boards-modules/embedded-modules-by-processor/

Last edited by OneTimer1 on 13-Feb-2018 at 07:39 PM.

@Deniil715

Quote:

There are PPC CPUs that are very fast, why HW-emulate??

They tend to be considered very expensive by contrast.

Quote:

The best way to make a new Classic PPC accelerator would be to use a SoC of some kind. Like the one in Tabor or Sam.

The problem with this, as demonstrated in those machines, is that they aren't exactly compatible with a 603e or whatever was used. Someone thinks, wow my new Classic PowerPCX accelerator, 800 Mhz (or whatever), this whips. Loads up Heretic. Boom! Crash! What the hell! This is a real Amiga!!

So my point is, that it needs to be like a G3 or G4 even or forget it. Embedded stuff won't cut it. It will crap it.

Last edited by Hypex on 15-Feb-2018 at 01:34 PM.

@Hypex

Amiga, WarpOS and 500mhz PPC

https://www.youtube.com/watch?v=rUics1KTwsE

@g01df1sh

Well if you put the VDHL PPC code in the Vampire we would have the best of both worlds :)


Offcourse this is hypothetical ;)


P10xx cpu's are cheaper then a FPGA, so why is the next question.

Last edited by bennymee on 14-Feb-2018 at 08:08 PM.
Last edited by bennymee on 14-Feb-2018 at 07:53 PM.

@Hypex

Why not a soc, they are cheap, use less power, use more modern RAM memory and are available.

Yes they are missing instructions, but Hyperion is working on it as we can read on a lot of places. Besides that, benchmarks of the Tabor show that it can be a lot faster then the Sam460ex soc and if emulation kicks in, it's 2/3 of that. But as a replacement for the Blizzard or CyberstormPPC's every soc is a lot faster :)

Last edited by bennymee on 14-Feb-2018 at 08:07 PM.

@g01df1sh

There uaed to be something called openzcore for an opensorced ppc softcore. I cant find it now, perhaps with more searching it will show up, but I dont recall what ppc it tried to implement.

It will of course be far slower than a "real" ppc chip due to fpga overhead, but could be functional. Im not sure if it could match even csppc or sam4x0 speeds.

Found the softcore code in a yahoo groups file downloads page, uploaded to GitHub at
https://github.com/amigabill/openzcore

Last edited by billt on 16-Feb-2018 at 03:39 AM.
Last edited by billt on 16-Feb-2018 at 03:17 AM.
Last edited by billt on 14-Feb-2018 at 10:46 PM.

_________________
All glory to the Hypnotoad!

@utri007

Quote:

https://www.youtube.com/watch?v=rUics1KTwsE

1:26-1:30 is interesting -- I didn't think sloping floors were possible with static BSP trees. I played a lot of Hexen, but I don't remember this. I may have been too busy running for my life to notice the floors...

_________________
"Unix is supposed to fix that." -- Jay Miner

@bennymee

Because if it is a Book E, it will be a lot less compatible. They need a Book S. User instructions are easy enough to work around if they can be trapped and worked around. But supervisor instrutions are more tricky.

The PPC kernel emulation API layers we have, AFAIK, don't account for incompatibilities with instructions. If they did it wouldn't matter as much.

I've seen Hereric run on a 440EP and soon crash. Fine on my G3. Embedded vs. desktop. Doesn't need AltiVec or anything. But lacking an FPU makes it worse. It would need a lot of emulation layers to work. Okay maybe just three. Soft FPU, user and super emulation.

@bennymee

Quote:

P10xx cpu's are cheaper then a FPGA, so why is the next question

Tabor only uses the P1022 because the low end T series SOCs weren't available during the design phase. If you were building a new board today the T1014 or one of the dual core variant are what you'd want to use. They have the same E5500 core as the X5000 so no need for trapping FPU instructions.