Using a PC as an accelerator? Had an idea last night....

[BillHarrison]

I mean, that could be said of this entire forum Bloodline.  Why mess with a slower / fragile / expensive hardware period?  Why not just ALL run winuae and forget the hardware completely?  

Nostalgia, the technical challenge, the feel of real amiga hardware, etc.  Why not stay at home and watch tv instead of climbing a mountain?  Its certainly easier!

[bloodline]

BillHarrison wrote:
I mean, that could be said of this entire forum Bloodline.  Why mess with a slower / fragile / expensive hardware period?  Why not just ALL run winuae and forget the hardware completely?  

Nostalgia, the technical challenge, the feel of real amiga hardware, etc.  Why not stay at home and watch tv instead of climbing a mountain?  Its certainly easier!

This is a better idea, you you are going down this route:

http://www.amiga.org/forums/showthread.php?t=37763

[sim085]

Although the idea is very nice and interesting it still more hard and expensive to achieve then BillHarrison idea (IMHO).

The idea on the other thread requires hardware to be developed on bought only for the Amiga machine. While the idea on this thread make the Amiga owner able to use the resource of the PC he/she (most probably) already has.

Also this idea is much easier to implement for the non technical people on this forum; it is ibviously easier to build some form of connector between the PC and Amiga then building a whole card - since obviously if a card is ever developed for the Amiga it will only be for a small range of Amiga's (such as the A1200) and not mass produced for reasons discussed on other posts.

[ollygd]

This is an interesting thread - its stuff I've thought about quite a lot. I think the die-hard fans (myself included!) like to keep using their old hardware for the real "experience", and ideas like this are aimed at preserving this. A certain amount of the romance is lost when using UAE, no matter how good it is!

We've recently started using Windows Server 2008 at work, and one of the nice features is the "Remote Application" function via Terminal Services. In previous versions, terminal services allowed access to a remote machine's entire desktop, but with this, individual applications can be sent to a client, from a TS server, as if it were installed locally.

I wondered what the odds of developing a compatibile RDC client for a suitably specced Amiga would be? You could open anything on a Server 2008 box, in Workbench. All the processing goes on in the server, and the presented application is little more than an image.......

Just a thought!

[amigaksi]

>How does this external converter know what size element is being accessed in order to do that?

>The CPU core will perform byte/word/long/quad reads and writes on its own caches but the CPU as a whole will perform cache line transfers to/from the main memory. I can't see how any interface in between can have any way of knowing what the data being transferred is in order to to do any swapping. All it will see are long bursts of fixed width accesses.

The data/address lines (including A0) must be coming outside of the processor since the I/O ports are not cached, VGA memory is not cached, and some other areas aren't cached or configurable.  In worst case if some processor is only accessing memory via cache, you can disable the cache; at least on my Toshiba laptop, it has the option for write-through so writes are not cached.

Interesting though that people have declared in this thread:
(1) not do-able in hardware,
(2) not do-able in software (me),
(3) do-able in software, and
(4) some imagining that the entire amiga is do-able in software.  

[Linde]

bloodline wrote:
If you are going to use the PC's RAM, why not use the PC's GFX and Sound cards? Why not use it's Keyboard and Mouse... in fact why waste time with the Amiga at all, and JUST RUN UAE...

Had you not ignored my last message in this thread maybe you would have taken a hint. The original idea was to use a PC as a classic styled CPU accelerator with FastRAM. While support for RTG graphics and PC sound is excellent in WinUAE, AGA emulation is still not quite there, so there's definitely a hole to be filled by a new accelerator for classic amigas. Now it's just a matter of how to do it, and I think this idea is excellent.

If you don't agree there's still no use in coming here and calling the thread a "wankfest". We're toying and dreaming about an idea, and there's no harm in that, so there's no need to be rude.

[bloodline]

Well if the AGA emulation isn't very good in UAE, name me a program I can run that would fail in UAE please :-)

This is a wankfest because of the initial assumption that UAE has problems.

[Linde]

Well, I tried booting from the CF card I use in my Amiga in the latest version of WinUAE (based on the A1200 quckstart configuration with "best compatability"), emulating my setup (stock 1200 with 8mb expansion RAM).

First game I tried was Lionheart (since it uses a lot of visual effects based on graphics timing) and I randomly get graphical artifacts in the background.

The second game I tried was The Addams Family (less hardware wank), which worked very well except that the sound from when you jump on an enemy would randomly be reduced to just a "click" noise. This could be because of my sound settings, though, but I don't usually experience the problem of sounds disappearing in other games.

The third game I tried was Agony, which works great until you move the player to the bottom of the playfield and fire. Strange looking artifacts appear on the top of the playfield.

None of these are even AGA games, and these bugs don't appear on the real machine.

As I pointed out earlier there's no need to try every piece of software when you can just have a look at the WinUAE changelog and see what bugs were fixed in chipset emulation over the last few versions. Do you really think that the AGA bug that was fixed in 1.5.0 ("AGA mode sprite garbage may have appeared in some cases.") would be the final one?

Honestly I think that you should just get over your besserwisser attitude until you actually get a clue.

[InTheSand]

ollygd wrote:
We've recently started using Windows Server 2008 at work, and one of the nice features is the "Remote Application" function via Terminal Services. In previous versions, terminal services allowed access to a remote machine's entire desktop, but with this, individual applications can be sent to a client, from a TS server, as if it were installed locally.

I wondered what the odds of developing a compatibile RDC client for a suitably specced Amiga would be? You could open anything on a Server 2008 box, in Workbench. All the processing goes on in the server, and the presented application is little more than an image.......

Published applications are great, provided the client is up to the job of receiving data and updating the display quickly enough. However, even the fastest 68K CPU could have problems with this, especially with a lot of audio or graphical data.

Even if, hypothetically, I guess there could be a chunky-to-planar graphics data conversion done at the server end, then the planar data fed in compressed format to the Amiga... And similar for any audio, with it being reduced to low bitrate possibly mono sound, then for a local Amiga client to not have to do too much, this still might be too taxing to get a really usable experience.

 - Ali

[freqmax]

I found this thread today
And would like to say it's definitely feasable. The paths to accomplish it may be:
0) Buy an 68000 in-circuit hardware emulator for x86-PC. And write the PC end of it.
1) Buy an PCI-FPGA card ~100 EUR. And wire the I/O to an 68000 cpu socket. Keep lengths short though!
2) Software program that runs on an unmodified Amiga, but makes the Amiga a slave via SCSI. Ie read instruction+data, amiga->pc, execute, pc->amiga, write/read, repeat..

Latency IS an issue. This makes USB/Ethernet/Firewire unsuitable. This is especially true for usb. Usb were force feed by putting those {bleep} interface chips on the motherboard without asking. Something along firewire would been a better performer.

As for the amount of data, just for 7,14 MHz it will be at least >500 Mbps.

Gains: Cycle exact sound & graphics. Though CPU cycle lengths may may vary
Something all software based emulators will struggle with is latency!

[Fransexy_]

I was thinking in something similar since years ago.An Amiga accelerator using an x86 cpu as emulated 68k.The x86 is used 100% only to emulate 68k instructions and  you used the rest of the logic of the accelerator to wire to CPU slot.The Amiga only see a super fast 68k cpu and the x86 are always emulating a 68k and only see and returns 68k code
You can even use the newer ultra low power consumption  neetbooks x86 cpus for using in the trapdoor slot of original amiga cases

Said this, I would add that the cpu could be of any family, we are talking about x86 but an accelerator with the same concept could be made with ARM, PowerPC,MIPS, SPARC or with a FPGA

[ajlwalker]

Had to lol at Linde's put down to bloodline.  :lol:

[KimmoK]

I vote for these:

Alternative1:
- 68k built in FPGA, on a CPU slot compatible card (natamiguys can deliver their future 070 chip code for that)
- can be changed to a PPC accelerator by programming the FPGA again
- no need for that stooped x86 dongle-kludge-box beside the Amiga

Alternative2:
- AGA or Minimi or Natami GFX chips on a 3.3volt PCI card, then the 68k can be emulated by the host CPU.
- would be interesting & usable also in SAM440


... I might buy both  :-P

[freqmax]

Ofcourse the 100% FPGA route is the prefered way. But if you going to do the x86-accelerator path. Then my listed alternatives might be valid.

The only catch with FPGA is that it requires a whole set of transistors to "fake" one real ASIC one. Which translates mainly to speed and functionality per chip.

I think the most interesting board right now is the fpgaarcade one. It has the potential for 020/AGA.

[joekster]

All of this discussion sounds almost exactly like AmiJoe. Anyone know whatever happened to this project. It seemed pretty far along. Maybe someone would want to continue working on it...AmiJoe on bboah

[sim085]

How much is it possible that something like this is developed for the A500? Maybe even through some 'do it yourself' project like the GB1000 project!