English Amiga Board

English Amiga Board (http://eab.abime.net/index.php)
-   Hardware mods (http://eab.abime.net/forumdisplay.php?f=105)
-   -   FPGA based A500 accelerator (http://eab.abime.net/showthread.php?t=100729)

mkstr 03 February 2020 15:20

FPGA based A500 accelerator
 
I know some people are hoping for the 32-bit Romulator, but I'm afraid I have been tinkering with something cooler: I have grafted a Terasic DE0 Nano into the 68000 socket of an A500+ and have TG68 up and running.

This is currently a work in progress but I thought it was time to share it. You can read about it here: https://www.mike-stirling.com/2020/0...the-amiga-500/

Mike

kipper2k 03 February 2020 15:22

this is really awesome, amazing work. there is a lot more planned for this and progress has been fantastic. The numbers are great on the actual Amiga 500 for the basic TG68 core. The video is very smooth :)

awesome job!

kipper2k 03 February 2020 16:06

here is a pic of the sysinfo screen...

http://kipper2k.com/accel/8444.jpg

Edit - dhrystones has now been increased to 15,116

utri007 03 February 2020 16:32

Would this be little more flexible than actual CPU board?

Would it be possible to make it CDTV compatiable without "big pain"?

https://github.com/terriblefire/cdtv.device

kipper2k 03 February 2020 16:41

Quote:

Originally Posted by utri007 (Post 1376216)
Would this be little more flexible than actual CPU board?

Would it be possible to make it CDTV compatiable without "big pain"?

https://github.com/terriblefire/cdtv.device

I can't go into specific details yet of the features, can you be more specific as to what your issues are ?

mkstr 03 February 2020 16:47

Quote:

Originally Posted by utri007 (Post 1376216)
Would this be little more flexible than actual CPU board?

Would it be possible to make it CDTV compatiable without "big pain"?

https://github.com/terriblefire/cdtv.device

The FPGA approach does allow for other modules to be built in beyond just the basic CPU functionality, but currently it just looks like a fast 68000 with some autoconfig fastmem.

Are you thinking about using it in a CDTV or turning an A500 into a CDTV?

Mike

Niklas 03 February 2020 16:56

Quote:

Originally Posted by mkstr (Post 1376190)
I know some people are hoping for the 32-bit Romulator, but I'm afraid I have been tinkering with something cooler: I have grafted a Terasic DE0 Nano into the 68000 socket of an A500+ and have TG68 up and running.

That's very nice!

If you make improvements to the CPU core e.g. to support more instructions, or to optimize the internal architecture to get higher IPC, will you release the result as open source?

It would be amazing to have a free and open source CPU core that can eventually compete with the closed source core used by Vampire.

mkstr 03 February 2020 16:57

Quote:

Originally Posted by Niklas (Post 1376226)
That's very nice!

If you make improvements to the CPU core e.g. to support more instructions, or to optimize the internal architecture to get higher IPC, will you release the result as open source?

It would be amazing to have a free and open source CPU core that can eventually compete with the closed source core used by Vampire.

Absolutely. The core is LGPL so any changes there will be published in accordance with that licence.

Niklas 03 February 2020 17:25

Quote:

Originally Posted by mkstr (Post 1376227)
Absolutely. The core is LGPL so any changes there will be published in accordance with that licence.

Splendid. If you also make the hardware open source then I very much look forward to building one and contributing to the CPU core in any way I can.

kipper2k 03 February 2020 17:35

The board will be BGA FPGA and also the RAM will prob also be BGA, unless you have the know how and reflow oven etc it will be a very tough build, and expensive if you get a bad solder

Niklas 03 February 2020 17:55

Quote:

Originally Posted by kipper2k (Post 1376236)
The board will be BGA FPGA and also the RAM will prob also be BGA, unless you have the know how and reflow oven etc it will be a very tough build, and expensive if you get a bad solder

That makes sense. As long as the design is open source I'm happy to pay someone else to build it for me.

On principle ("The chucky principle"?) I'm not interested in closed source Amiga accessories. If closed source hadn't been an issue for me I suppose I could buy a Vampire today.

kipper2k 03 February 2020 17:58

Quote:

Originally Posted by Niklas (Post 1376243)
That makes sense. As long as the design is open source I'm happy to pay someone else to build it for me.

On principle ("The chucky principle"?) I'm not interested in closed source Amiga accessories. If closed source hadn't been an issue for me I suppose I could buy a Vampire today.



The board will obviously be a LOT cheaper than a vampire and will be fully compatible with the Amiga, the board will be professionally built within the UK and sales will originate there also. I am just a very happy tester :)

robinsonb5 03 February 2020 18:12

Really cool - nice work!
Are you planning on using the same FPGA for production boards, or something newer / larger / smaller / cheaper?

kipper2k 03 February 2020 18:20

Quote:

Originally Posted by robinsonb5 (Post 1376251)
Really cool - nice work!
Are you planning on using the same FPGA for production boards, or something newer / larger / smaller / cheaper?

hmmm, i am allowed to say, newer, larger, smaller dimensions, faster and comparable price :)

Niklas 03 February 2020 18:23

Quote:

Originally Posted by mkstr (Post 1376227)
Absolutely. The core is LGPL so any changes there will be published in accordance with that licence.

There's also this project https://www.a1k.org/forum/index.php?threads/72081/ that builds a 68k replacement in FPGA which is also based on TG68. It will be very interesting to see how these projects can benefit from the advancements developed in the other project.

kipper2k 03 February 2020 18:27

Quote:

Originally Posted by Niklas (Post 1376257)
There's also this project https://www.a1k.org/forum/index.php?threads/72081/ that builds a 68k replacement in FPGA which is also based on TG68. It will be very interesting to see how these projects can benefit from the advancements developed in the other project.

interesting, very similar, :)

meynaf 03 February 2020 18:34

Quote:

Originally Posted by mkstr (Post 1376227)
Absolutely. The core is LGPL so any changes there will be published in accordance with that licence.

Do you plan doing such changes or will you always use unmodified TG68 ?
Enhancing programming flexibility has always interested me as asm coder.

mkstr 03 February 2020 18:40

Quote:

Originally Posted by meynaf (Post 1376266)
Do you plan doing such changes or will you always use unmodified TG68 ?
Enhancing programming flexibility has always interested me as asm coder.

Priority would be to add compatible enhancements in line with the real 680x0 I think. Are you thinking SIMD?

meynaf 03 February 2020 19:36

Quote:

Originally Posted by mkstr (Post 1376268)
Priority would be to add compatible enhancements in line with the real 680x0 I think. Are you thinking SIMD?

No, not SIMD. IMO this is the job of a GPU, not of a CPU.
I'm really thinking about general purpose additions to make progamming more pleasurable and enhance code density. This would in turn make all new programs faster, not only a few selected ones.
This has been a major cause of disagreement with Gunnar when i was in the apollo team, and the reason i left.

kipper2k 03 February 2020 19:41

Quote:

Originally Posted by meynaf (Post 1376282)
No, not SIMD. IMO this is the job of a GPU, not of a CPU.
I'm really thinking about general purpose additions to make progamming more pleasurable and enhance code density. This would in turn make all new programs faster, not only a few selected ones.
This has been a major cause of disagreement with Gunnar when i was in the apollo team, and the reason i left.

From what i can tell there are a lot of instructions in the TG68 core that can be improved similar to the enhanced instruction set that apollo uses. I do not know enough to do this myself but i think there is enough interest to improve common instructions etc to speed up the core. Would need someone with more knowledge to confirm this


All times are GMT +2. The time now is 21:17.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, vBulletin Solutions Inc.

Page generated in 0.06319 seconds with 11 queries