FPGA based A500 accelerator

[Lemaru]

Regarding any Gayle emulation. Have you found a way to do this without breaking PCMCIA support in the A600? Or possibly the option to disable it

[alexh]

It would be interesting to know if the level shifters could easily be replaced with zero ohm equivalents? Do they make them? I only ask because the Atari ST remake PCB H5 moves the level shifters to the motherboard adding a 3.3v CPU socket so addon's don't need to bare the cost of level shifters in the future. Not that important as those boards have a 5V CPU socket too IIRC.

[indigolemon]

Originally Posted by alexh

It would be interesting to know if the level shifters could easily be replaced with zero ohm equivalents? Do they make them? I only ask because the Atari ST remake PCB H5 moves the level shifters to the motherboard adding a 3.3v CPU socket so addon's don't need to bare the cost of level shifters in the future. Not that important as those boards have a 5V CPU socket too IIRC.

Based on my understanding of the component sizes involved to make this fit neatly, I'd suggest you'd be better swapping them out on the remake mother board PCB if you knew you were going to use this accelerator.

[robinsonb5]

Originally Posted by Promilus

Let's see how many commits has TG68 source code since it was published...

Good idea: https://github.com/TobiFlex/TG68K.C/commits/master


Quite a bit of activity recently, actually, with contributions from several people other than the author.

And I'm fairly confident many ppl asking for sources are just leechers ... to feed on original work without adding anything to it.

Yes, that's certainly a problem. There's also the problem of what LGPL even means with regard to hardware.

In the software world when a component is LGPLed then it's fine for closed source components to be linked with it, but it must either be dynamically linked, or object code must be provided so the end user can re-link the application with a different version of the LGPLed code. It's hard to meet that requirement of the LGPL in an FPGA project without supplying the entire source. (The easiest option is usually to negotiate a separate license with the author - but that gets trickier once other people have contributed.)

[mkstr]

Originally Posted by Lemaru

Regarding any Gayle emulation. Have you found a way to do this without breaking PCMCIA support in the A600? Or possibly the option to disable it

Haven't really thought about it yet, although there are option links on the board that could be used to turn it off and on. Any proprietary DMA-capable interface I came up with would not have to be turned off at the same time so that could theoretically coexist with Gayle running PCMCIA.

[alexh]

My understanding of LGPL WRT TG68k means Mike can negotiate a commercial license with Tobias for exclusive use of the code in a commercial product without having to publish his source code changes.

[alexh]

Originally Posted by indigolemon

Based on my understanding of the component sizes involved to make this fit neatly, I'd suggest you'd be better swapping them out on the remake mother board PCB if you knew you were going to use this accelerator.

Erm, you'd just use the 5v 680x0 socket instead of the 3.3v socket.

[mkstr]

Originally Posted by alexh

It would be interesting to know if the level shifters could easily be replaced with zero ohm equivalents? Do they make them? I only ask because the Atari ST remake PCB H5 moves the level shifters to the motherboard adding a 3.3v CPU socket so addon's don't need to bare the cost of level shifters in the future. Not that important as those boards have a 5V CPU socket too IIRC.

The level shifters are just 5V tolerant buffers - they output 3.3V LVTTL anyway (which is compatible with 5V TTL logic levels), so it _should_ just work. Even powering it from 3.3V will probably still work, although it would depend on how much drop-out occurred in the on-board 3.3V buck. Bypassing that might be needed but that would just be a wire link between two test points.

[Promilus]

@mkstr - it's 68000 DIP form factor, insanely impressive design. So it should be pretty straightforward to replace A1000, A500, CDTV and A2000 CPUs with that. What about A600? Not that I care much about it. I'm just wandering.

About that serial differential interface - is softcore fast enough to handle LiteSATA? I know, system driver would be a b.tch but since IDE is basically obsolete and you have to use CF2IDE or SATA2IDE adapters anyway for new disks it makes perfect sense to implement native controller - even if full bandwidth can't be achieved. With CFAST you can even keep it pretty small.

[mkstr]

Originally Posted by Promilus

@mkstr - it's 68000 DIP form factor, insanely impressive design. So it should be pretty straightforward to replace A1000, A500, CDTV and A2000 CPUs with that. What about A600? Not that I care much about it. I'm just wandering.

About that serial differential interface - is softcore fast enough to handle LiteSATA? I know, system driver would be a b.tch but since IDE is basically obsolete and you have to use CF2IDE or SATA2IDE adapters anyway for new disks it makes perfect sense to implement native controller - even if full bandwidth can't be achieved. With CFAST you can even keep it pretty small.

kipper2k has done an adapter that will work in the A600.

I was assuming most people would just use the SD card for storage. Even SATA is going obsolete now so PCIe would seem like the better option for a "modern" interface. I did consider adding that but it needs a more expensive FPGA with gigabit transceivers, and the feature creep line had to be drawn somewhere.

[alexh]

My CDTV is set up for TF536 testing and I've been actively working with others to try to patch the ROM. If you want some testing on a CDTV then I'm happy to buy and help debug.

[Promilus]

I was assuming most people would just use the SD card for storage

That's a valid assumption. I was just thinking about slightly more reliable storage. But new SDHC and SDXC are pretty solid anyway and if UHS is implemented in the future it would still give incredible boost.

[Lemaru]

Originally Posted by mkstr

kipper2k has done an adapter that will work in the A600.

I have done an adaptor for the A600 too, works great with the TF536 so should be good for this too. Going to revise it a little and do a different form factor for use with smaller cards such as this.

I don't have any other A500 cards to test it with though but a few others I have sent it to have had it working with other boards (just not the TF534).

Can't wait to get my hands on one of these to give it a go.

[d4rk3lf]

Why I waste time on politics and elections, when this is so much more interesting and cool?
Amazing work.

I hardly understand whatever you guys talk, and very glad that I see you exchange knowledge, and kudos to Mike sharing his view on all questions.
From my poor understanding, this will make my Amiga 500, something like OCS/040?
Price is yet to be established, but on the good way to be "higher then average 68000 accelerators, but lower then some others"?
Sound amazing.

For us, poor understanding souls, any chance of some new video testing the card?
Something along:
Gloom (OCS)
Doom (OCS)
Wolf... Wing Commander, Dune 2, Nemac IV, Stunt Car Racer, Legends of Valor... etc

Also, how you guys think rendering in 3D apps would improve?
I know (somewhat) Lightwave, and a littkle bit of Real3D.
It would be cool seeing if there's improvement, and how much.

Again - great job!

[alexh]

TF has said he'll share his knowledge of adapting the TF536 to the Atari ST including the bus arbiter.

https://www.exxoshost.co.uk/forum/vi...=3587&start=10

Just make sure that all the 68000 pins that the Amiga doesn't use are wired to the FPGA

[kipper2k]

Originally Posted by alexh

TF has said he'll share his knowledge of adapting the TF536 to the Atari ST including the bus arbiter.

https://www.exxoshost.co.uk/forum/vi...=3587&start=10

Just make sure that all the 68000 pins that the Amiga doesn't use are wired to the FPGA

Nice,
i do have an Atari that i removed the soldered 68k and socketed it ready to some some testing once we get going. The size of this is so snall that i cant think of any comp that it will not fit inside without causing issues with the case closing etc. even when used with the a600 adapter which i was able to design with 2 securing screws that can allow you to swing it around your head if you want to lol

[mkstr]

Originally Posted by alexh

TF has said he'll share his knowledge of adapting the TF536 to the Atari ST including the bus arbiter.

https://www.exxoshost.co.uk/forum/vi...=3587&start=10

Just make sure that all the 68000 pins that the Amiga doesn't use are wired to the FPGA

All the pins are connected to the FPGA, including the bus grant ones. I think the key differences between the Amiga and the Atari are that the latter uses the bus grant and also non-autovectored interrupts. The bus grant I think should be able to be handled in the bus cycle logic, which is separate from TG68, and the interrupt handling might just work anyway because (IIRC) the Amiga effectively uses that mechanism too but relies on the hardcoded jump table at the end of the ROM to make it look like autovectoring.

[Higgy]

Will this allow the Megadrive to have true Blast Processing?
Great work on the device, looks smart and very micro.

[lesta_smsc]

Originally Posted by alexh

TF has said he'll share his knowledge of adapting the TF536 to the Atari ST including the bus arbiter.

https://www.exxoshost.co.uk/forum/vi...=3587&start=10

Just make sure that all the 68000 pins that the Amiga doesn't use are wired to the FPGA

I thought TF was leaving the scene after some people ripped off his work? Glad to see he's still about doing what he does best

[alexh]

Originally Posted by lesta_smsc

I thought TF was leaving the scene after some people ripped off his work? Glad to see he's still about doing what he does best

He has stopped working on things but the work is already done.