@Karlos

https://libre-soc.org/

Source code and documentation available

_________________
retired

@PhantomInterrogative

So a Tabor that is even more expensive, would run normal PPC code even slower and needs super special care in SW to make the 2 CPUs cooperate.

Brilliant idea.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@geen_naam

I appreciate the near autistic dedication to technicality, but for the avoidance of doubt and to save you further embarrassment, by "PPC being too expensive" I mean the PPC/NG platform as a whole, not individual PowerPC SKUs.

That expense is made worse by what a fundamental waste of money it is when you consider how grossly underutilised your N-core 64-bit PPC is. It's just completely pointless. Your condescending reaction to the very suggestion of an alternative that might be cheaper (even if less peformant) underscores everything that's wrong and toxic with the whole enterprise. Pay over the odds for the hardware, get an OS that is incomplete, pay again for some third party enhancements to use the hardware you install and congratulate yourself on only being able to run it at a fraction of it's potential.

I'm sorry but I'm just not wired they way. I'd rather take the 32-bit non SMP OS and run it on a 32-bit single core CPU solution that it can make full use of, especially if that was an affordable solution.

But no. How very dare anyone suggest such heresy. You want something affordable? You may not be part of our club.

_________________
Doing stupid things for fun...

@geen_naam

It is simple, people don't change.

They may change their outside behavior, but thats all.

There was 0 reason for Hyperion to go into 68k but they (he) did fully aware of the consequences.
So he did kick that can of s##t into the fan just like he has done before.

Wether AEON has "run out of licenses" is irrelevant as Trevor might have just enough dirt on him to make him think twice.

Sure little is know for sure but the facts clearly points into the direction that it is all about a bruised ego and little but actual business or the community.

He holds some essential parts of OS4.
AEON hold other parts.
Unpaid developers could revoke licenses anytime.

So he might have the power to grant an "OS4 license" but that is worthless without getting one for the other pieces even before writing drivers.
AEON kinda has all the ducks in a row but I doubt they would just hand it over to some 3rd party. AEON itself seems under the notion that there is nos such thing as sunk cost and all there past failures and mistakes need to payed by costumers, hence the absurd price for the Tabor. A tax they levy on any future PPC project as well.

But feel free to believe that you could do a cheap PPC board and bring to the OS4 market without getting burned.....

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@Karlos
Well, it would be nonetheless an interesting exercise, you seem to like that kind of stuff...
And in the end there is always the hope (practically NULL I guess....) that you could make an ASIC in the future.

_________________

José

@PhantomInterrogative

Without taking the condescending tone, that wouldn't work. Ignoring for a moment the now established position that it's basically not fast enough, there's no obvious way to route the instructions to it for execution and any software trap you need to handle that is going to end up slower than just handling it locally in software anyway.

The problem with the Tabor is simple. The wrong CPU was chosen for it, but you know, I'm the moron here because as green_nam says, the main reason you are paying the high price for such hardware in the first place is that it's designed and validated by professionals.

_________________
Doing stupid things for fun...

@Jose

Quite. It would be interesting but likely not practical.

_________________
Doing stupid things for fun...

@geen_naam

If you think you can create an affordable PPC solution, do it, because frankly Tabor isn't it. Just make sure you use a SKU with a properly compatible FPU and you're already adding value over it.

Last edited by Karlos on 10-Feb-2024 at 08:45 PM.

_________________
Doing stupid things for fun...

@Karlos

Thank you for not taking the route of condescension. I would rather someone tell me I am wrong with reason rather than sarcasm. My idea was bad. Of course so was the idea to use the P1022.

_________________
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

@geen_naam

I didn't suggest a vampire. I asked a very simple question as to whether or or not a softcore PPC was a viable proposition for a more affordable alternative to current offerings and the answer is a resounding "no". I see no reason to labour the point, despite your goading - which seems gauche considering I'd already mapped out the likely outcomes in the question:

Quote:

Is it feasible at all, if so, what would be the sort of upper limits on performance, etc. Is there a middle ground somewhere between the 60x class and G3/G4 that os within reach, or is synthesized hardware just too slow?

For whatever reason it's clear you dislike me and that's your prerogative but I do want to ask, where do you get the impression I'm a PPC hater from?

Last edited by Karlos on 10-Feb-2024 at 11:34 PM.

_________________
Doing stupid things for fun...

PhantomInterrogative Quote:

I suggested a while ago, prior to the A1222 respin, that A-Eon place a FPGA on the Tabor that would be configured to be a standard PPC FPU. Code wouldn't necessarily have to be emulated, just redirected from the SPE to the FPGU.

There are CPUs that have an interface for external coprocessor units but they are slower than integrated units. There are a couple of potential applications for Amiga related hardware.

A FPGA 6888x FPU replacement for the 68020-68030 should be possible although it likely wouldn't be much higher performance. It likely could support FSop and FDop instructions allowing it to execute code compiled for the 68040 and 68060 FPU.

The Xilinx Virtex-4 FX FPGA has built in PPC405 cores with an Auxiliary Processor Unit (APU) controller interface. It interfaces with the FPGA likely reducing the interface overhead to a minimum. The PPC cores are "450 MHz, 700+ DMIPS" although they may use hard blocks rather than FPGA resources to reduce routing inefficiency.

https://www.xilinx.com/products/silicon-devices/fpga/virtex-4q/modals/powerpc-32-bit-risc-processor.html

Kronos Quote:

There was 0 reason for Hyperion to go into 68k but they (he) did fully aware of the consequences. So he did kick that can of s##t into the fan just like he has done before.

There is a very good "reason for Hyperion to go into 68k". The retro 68k Amiga is where 99% of the Amiga market is. Ben has highly questionable morality but he is intelligent. PPC was a sinking ship with tiny and declining sales (production down to 100s of units). Nobody else is going to enter the PPC AmigaNOne market. Retro Games Ltd entered the 68k Amiga market and has likely sold hundreds of thousands of units. The AC/Vamp has likely sold over ten thousand units. Tens of thousands of 68k Amiga users are likely using universal retro FPGA hardware. There are likely tens of thousands of emulated 68k Amiga users. There are likely thousands of original 68k Amiga hardware users. Emulation and FPGAs can't compete with ASICs, as this thread discusses, yet there are still hundreds of thousands of units of retro 68k Amiga demand and there would be more with more competitive hardware and lower prices. Maybe there still isn't enough demand for a mass produced 68k ASIC and hardware for the Amiga alone but there is potential for other 68k retro devices and embedded use to improve economies of scale.

Karlos Quote:

If you think you can create an affordable PPC solution, do it, because frankly Tabor isn't it. Just make sure you use a SKU with a properly compatible FPU and you're already adding value over it.

It is easier said than done without backing from powers that have joined forces, whether it is 68k or PPC hardware. There is too much risk that time or money is spent only to have one of the necessary entities back out or refuse to do business. The current conditions may be road blocks to keep competition out, at least from some of the necessary entities who are not aligned. We don't need an AmigaOne but rather one Amiga Corporation.

@matthey

Well, they are going to sell SIN and Heretic II to all markets.

68K market huge disadvantage in 3D, and speed.
At least when it comes to native PPC programs vs native 68K programs.

You can ignore the emulation market, if can run the games on Windows or MacOSX natively no point using UAE.

Last edited by NutsAboutAmiga on 10-Feb-2024 at 11:04 PM.
Last edited by NutsAboutAmiga on 10-Feb-2024 at 11:03 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@NutsAboutAmiga

Quote:

68K market huge disadvantage in 3D, and speed.

3D, perhaps, but it's not as if Sin and Heretic II are that demanding. As for speed, PiStorm will crush it. Here is what it was capable of 1 year ago on an A500 running Quake 2 fluidly enough with the pure software renderer:

https://www.youtube.com/watch?v=iGJAI00AYjQ

Now I don't claim that the 320x240 SW mode is going to be a great experience for either title, but at some point, one of two things will happen:

1. People will get it working in their Frankenstein rigs with existing PCI 3D cards.

2. The 3D capabilities of the SoC will be exposed and drivers will be written for that.

_________________
Doing stupid things for fun...

@matthey

Quote:

matthey wrote: There is a very good "reason for Hyperion to go into 68k".

Greed isn't a "good" reason.
Boosting his ego is neither.

Sure some money to be made (mostly because the work is done by unpaid fo^H^H volunteers but surely nothing worth calling a fortune.

He knew that the understanding was that Cloanto would do the retro/68k part, but when he realized that he could twist the words of the settlement agreement he just couldn't resist. Well "realized" might be a strong word as he might have placed it there for just that reason. Just like the bankruptcy and buyback clauses were added to the Amino-Hyperion contract by each side to f### the other over.

He ended up with big cost over that one, got pyrrhic victory and went for round 2 with similar results.
When the fat lady sings he would most likely be better off if he had spend that time moonlighting behind a McDonals counter.

-> no (sensible/good) reason

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@Karlos

so karlos you start the same shit again.
switch to pc. it will be cheaper.
for this shit you get obvious answer.
get lost. for arm I have android and ios.
want people to switch to arm give them something as good as android and ios.

@matthey

my amiga one work like my amiga 1200 only better bacause 1000 times faster.
my amiga one is better amiga than these made by commodore sfter amiga 3000.
68k gives nothing and there are no reason to swithc to it.

back to topic.
my ideal amiga will be fpga based with 68k and ppc core.
with ocs for old games and better graphics for rest.

@Karlos


Quote:

how practical is a softcore PPC, one designed specifically for the fundamental limitations of 32-bit addressing and single core?

How much work is it?

Making an FPGA RISC softcare is a lot easier and a many times less work
than making a good CISC core.
Making a PPC FPGA would be a magnitude less work than the development of the 68080 needed.


How fast would this be?

FPGA does significantly limit the clockrate compared to using an ASIC.

For example at IBM at the CELL development we put the PPC core of the CELL
for testing and debugging in an high end very fast $10.000 FPGA.
In the FPGA the PPC core reached 50Mhz ... while in the ASIC it reached over 4000 MHz.
Sony sold the PS3 with 3.2 Gigaherz - which was a low value actually for the design.
We at IBM ran the CELL ASIC at 4.8 Gigaherz with proper cooling.

As you can see there is a big difference between the clockrate the same design gets in an FPGA versus an ASIC.


The 68080 reaches about 90-100MHz in consumer FPGA - while in an ASIC the 68080 could easily reach several Gigaherz

Lets say your PPC will reach around 100MHz




How much performance would this be?

As you know, a RISC chip is limited in what it can do in comparison to a CISC chip.
What does this mean?
This means that very often a RISC chip does need several instructions to do the same work that a CISC chip does with one instruction.

One simple example:
lets say you run a game and that wants to adds some points to your score.

on 68K you could write this in 1 instruction

add.l #100000,myscore

The PowerPC will need more instructions todo the same work.
In this case it will need 6 instructions!

This one instruction needs 1 cycle on 68080.
The PowerPC will need many cycles for doing the same work.


Lets sum up

1) making a PPC in an FPGA is absolutely possible.
At IBM we did put any PowerPC in FPGA for development.
We put the lower 4xx that you have in SAMs in FPGA, and also higher chips like A2,CELL, POWER6,7,8
Making an PPC software is a magnitude less work than the development of the 68080 was.


2) The FPGA will limit the clockrate.
100Mhz is a reachable target for a consumer FPGA


3) The RISC needs often more instructions for doing the same amount of work than a CISC.
Often a PPC needs 3 or 5 or even 6 or 7 instructions to do what a single 68K instructions does,
This means the "net-performance" of a PPC running at 100Mhz
would in reality often be lot less than what the 68080 reaches today in the Vampire.

@ppcamiga1

Quote:

my ideal amiga will be fpga based with 68k and ppc core. with ocs for old games and better graphics for rest.

68K is nice to code.
Very nice to code in fact.

I believe that a big factor of so many people writing nice games and demos on Amiga - was that the 68K is so nice to program.
The PPC is very "unnice" to program. Its by far not as nice to code as the 68k..

A PPC in FPGA will most likely at same clock as a good 68K
Which in return means it will deliver less performance.
As you know the RISC chips need more instruction for doing the same "operation" as a CISC chip.


Developing your own PPC core might nevertheless make sense.
If you really believe that PPC is what you want.
Then design an FPGA PPC core is doable in reasonable time.
Limited by the FPGA clockrate limit this would be not impressive in performance.
You will be slower than Vampires....

But you will be independent!

This means you can make it to have much better compatible than some PPC not having a normal FPU

And if you get the pocket-money to make a real ASIC - you could get it 10 times higher clockrate ....
And you own the design....
So you are not depending on other companies anymore....

IMHO an FPGA softppc core implemented as we know in a Vampire stand alone machine could be a great solution and very funny to program.
Instead to waste transistors to emulate an old non sense chipset maybe a full functional powerpc core could be implemented in the same space, probably with altivec and/or some spe units.
The perfect thing could be add a low cost and low power gfx chip and make a new entry level amigang.

The speed will be a fraction of current ng machines for sure, but if it will have similar cost of current Vampire stand alone, it could brick the walls and sell many hundreds units.

About PowerPC architecture, I know most members of this old great forum are graduate in "MC68k architecture", but is quite evident when discussion goes on powerpc no one knows about what he is writing.

To add a 16 bit const in a powerpc register you need one instruction, for a 32 bit value you need two.
Load a 32 bit constant is not a so frequent case and it's not a real problem about powerpc arch.. The advantages of PowerPC versus MC68k and other old ISA architectures are BIG.
You can look at current Power 9 and Power 10 performances and capabilities.. once again the difference among Power architecture versus others relies inside fab processes and not in cpu design.

If someone is really interested about NG development should stop write here and swap to Amigans.

_________________
Pegasos II G4@1GHz 2GB Radeon 9250 256MB
AmigaOS4.1 fe - MorphOS - Debian 9 Jessie