Better performance and 080 support?

Xbytez
Site Admin
Posts: 469
Joined: Wed May 20, 2020 3:36 pm
Has thanked: 214 times
Been thanked: 787 times

Re: Better performance and 080 support?

Unread post by Xbytez »

I understand everyone has an opinion and is very much entitled to share this with others, a healthy discussion is fine. However personal attacks and insults to other users is unacceptable behaviour on this forum.

As such I have deleted a number of posts from this thread, those users involved have been warned.
MiSTer_Kirk
Posts: 210
Joined: Thu Feb 04, 2021 11:42 pm
Has thanked: 18 times
Been thanked: 46 times

Re: Better performance and 080 support?

Unread post by MiSTer_Kirk »

However personal attacks and insults to other users is unacceptable behaviour on this forum.
Understood, and apologies.
QuartexNOR
Posts: 10
Joined: Wed Mar 17, 2021 9:24 pm
Been thanked: 5 times

Re: Better performance and 080 support?

Unread post by QuartexNOR »

No worries, understood.
Xbytez wrote: Wed Mar 31, 2021 9:34 pm I understand everyone has an opinion and is very much entitled to share this with others, a healthy discussion is fine. However personal attacks and insults to other users is unacceptable behaviour on this forum.

As such I have deleted a number of posts from this thread, those users involved have been warned.
QuartexNOR
Posts: 10
Joined: Wed Mar 17, 2021 9:24 pm
Been thanked: 5 times

Re: Better performance and 080 support?

Unread post by QuartexNOR »

Just to underline my point (and thank you to the admin for cleaning up the thread, it got a bit heated there).

The vampire is fun, but the price point is quite steep. With shipping and taxes the sub total is in the $800 range for me up here in Norway.
It is here that the Mister makes for a wonderful alternative. I also feel the Mister deserves more representation, it really is one of the best, if not the best, FPGA implementation out there - and combined with the Linux layer that the DE-10 Nano provides, you have some brilliant avenues of technology that can be explored.

There is a lot of life in the Amiga community still. Thriving shops that provides new motherboards, keyboards, spare parts and amazingly enough, brand new hardware. There is likewise a good community on social media where new software and technology is constantly presented and sold. As well as software development groups covering assembly, C/C++, E-lang, Java, BlitzBasic and Amos basic. It might come as a surprise but Amos basic just got a major overhaul 20 years down the line, adding support for chunky modes, RTG display modes, modern file-formats and quite a bit of fixes and speed improvements.

The Amiga has this exploring spirit over it, where developers and users like to push the limits of what is possible. It's not really about the latest and greatest, but pushing boundaries and doing much with little.

My primary interest is not so much the hardware as it is Amiga OS. You have features in that little thing that Apple only added to OS X in 2015. And it's small size and extremely low memory requirements makes it a perfect candidate for embedded work (if modernized naturally).

Aros, which is a reverse-engineered variation of Amiga OS runs on x86, and it's presently being ported to ARM. It retains the use of 68k instructions by treating the instruction-set as bytecodes, much like Java or .Net operates with. This is converted to ARM code in realtime (PJIT) so that people can target the system using the classical toolchains. Which is pretty amazing, especially since you can pick up an SBC with 4-8 gigabytes of ram for less than $100 these days (Raspberry PI 4, ODroid N2, NVidia - plenty of options).

What I would like to see is some work to make it easier to cross compile, which means a faster cpu implementation first of all. This way its possible to use libraries such as SDL to abstract away some of the underlying differences.
Both GCC and Freepascal have up to date compilers for Amiga (some 10 million object pascal developers in the world, so FPC is a great tool to cover both Classics, Vampire, Morphos and Aros). The bottleneck is with the classics, in that some modules are best tested on real hardware, emulation struggles with clock-cycle distribution (e.g cpu takes all the cpu power, causing issues with chipset emulation).

There is also embedded to consider. A system like Aros is perfect for single application systems, typically used in medical and kiosk devices. Even linux has gained some weight over the years, and its now ancient ring system makes relatively simple tasks overly complicated. A simple, reliable and minimalistic OS is very attractive in such settings.

So there are avenues and potential in this work that might not be immediately apparent.
throAU
Posts: 181
Joined: Fri Sep 11, 2020 1:06 am
Has thanked: 229 times
Been thanked: 27 times

Re: Better performance and 080 support?

Unread post by throAU »

If you're really looking for a future amiga platform.... wouldn't you be better off with AROS?

68k is 25-35-45 years ago. 68080 is still way, way slower than modern hardware. AROS has integrated 68k emulation but otherwise runs on your native hardware at native speeds. It is supposedly source compatible with AmigaOS.

Port your software to AROS, run old software via built in UAE, be happy with FAR, FAR better performance than you'll get out of any FPGA based 68k family processor?

Rather than continually re-inventing hardware that is like 20 years behind the curve.... run a platform that is x86 (OR ARM) native and evolves with whatever new hardware comes out for the mass market? Effort is IMHO better thrown behind that direction?
lordoftime79
Posts: 97
Joined: Sun Feb 14, 2021 6:29 pm
Has thanked: 1 time
Been thanked: 2 times

Re: Better performance and 080 support?

Unread post by lordoftime79 »

throAU wrote: Sun Apr 18, 2021 1:44 pm If you're really looking for a future amiga platform.... wouldn't you be better off with AROS?

68k is 25-35-45 years ago. 68080 is still way, way slower than modern hardware. AROS has integrated 68k emulation but otherwise runs on your native hardware at native speeds. It is supposedly source compatible with AmigaOS.

Port your software to AROS, run old software via built in UAE, be happy with FAR, FAR better performance than you'll get out of any FPGA based 68k family processor?

Rather than continually re-inventing hardware that is like 20 years behind the curve.... run a platform that is x86 (OR ARM) native and evolves with whatever new hardware comes out for the mass market? Effort is IMHO better thrown behind that direction?
ha ha the only version of Aros I like is ApolloOS that runs on 68k hardware... silly man!
foft
Posts: 334
Joined: Thu Dec 03, 2020 11:05 am
Has thanked: 29 times
Been thanked: 120 times

Re: Better performance and 080 support?

Unread post by foft »

Just watching this great interview with Jim Keller. One good quote is that 80% of the execution is ~6 instructions:load, store, add, subtract, compare and branch.

Perhaps we can clock TG68K and the cache say 4x faster then add a 4 cycle pipeline for all the other instructions.

https://www.youtube.com/watch?v=AFVDZeg4RVY
robinsonb5
Posts: 129
Joined: Fri Jun 19, 2020 8:54 pm
Has thanked: 13 times
Been thanked: 57 times

Re: Better performance and 080 support?

Unread post by robinsonb5 »

foft wrote: Fri Jun 25, 2021 12:03 pm Just watching this great interview with Jim Keller. One good quote is that 80% of the execution is ~6 instructions:load, store, add, subtract, compare and branch.
Unfortunately load and store aren't separate operations on 68k, but are folded into most of the other instructions in the form of all those fancy addressing modes which make the 68k such a joy to code for (and such a pain to code compiler backends for!)
foft
Posts: 334
Joined: Thu Dec 03, 2020 11:05 am
Has thanked: 29 times
Been thanked: 120 times

Re: Better performance and 080 support?

Unread post by foft »

Oh well, then back to plan A :lol:
* increase clock
* pipeline failing paths
* repeat!

Guess that quote was more suitable for a https://en.wikipedia.org/wiki/Load–store_architecture. Though it is still a good interview.
robinsonb5
Posts: 129
Joined: Fri Jun 19, 2020 8:54 pm
Has thanked: 13 times
Been thanked: 57 times

Re: Better performance and 080 support?

Unread post by robinsonb5 »

foft wrote: Fri Jun 25, 2021 2:46 pmGuess that quote was more suitable for a https://en.wikipedia.org/wiki/Load–store_architecture. Though it is still a good interview.
It is indeed (I just watched it, or at least the first half until a mid-roll annoyed me enough to quit the YouTube session).

I daresay the observation applies to 68k as much as it to other architectures (in fact you're probably well-placed to test it - can either Musashi or qemu produce a histogram of executed instructions?) - it's just difficult to make effective use of that information thanks to the addressing modes.
Post Reply