|
Couriersud is a net list machine.....
#321457 - 02/05/14 03:14 AM
|
|
|
I've been keeping track (as best as I can understand it). Are we going to see DICE merge with MAME?
#ifndef NL_DICE_COMPAT_H_
#define NL_DICE_COMPAT_H_
I think we are getting sound now.
http://git.redump.net/mame/commit/?id=6ba06b3eec4a952b5780ac49d818afbc75a9cf7c
The man (or maybe he's a well designed android made by Mr. Belmont) is on fire.
I just wanted to thank him for his work. Very impressive.
 
|
|
|
|
Re: Couriersud is a net list machine.....
[Re: Antny]
#321466 - 02/05/14 10:05 AM
|
|
|
He also improved sound in 1942 in another recent last commit. It's disabled by default but if you compile a test build it's very impressive....
|
|
|
| gregf |
|
Ramtek's Trivia promoter
|
|
|
|
|
Reged: 09/21/03
|
|
Posts: 8604
|
|
Loc: southern CA, US
|
|
|
|
Send PM
|
|
|
Re: Couriersud is a net list machine.....
[Re: Antny]
#321467 - 02/05/14 12:04 PM
|
|
|
>The man is on fire. I just wanted to thank him for his work. Very impressive.
* applauds that a full working version of Pong is running while MAME is now at its 17th anniversary (Feb 05 1997) *
|
|
|
| R. Belmont |
|
Cuckoo for IGAvania
|
|
|
|
|
|
Reged: 09/21/03
|
|
Posts: 9716
|
|
Loc: ECV-197 The Orville
|
|
|
|
Send PM
|
|
|
Re: Couriersud is a net list machine.....
[Re: Antny]
#321491 - 02/05/14 09:43 PM
|
|
|
> The man (or maybe he's a well designed android made by Mr. Belmont) is on fire.
If I could build androids that well I wouldn't be working on drivers myself ;-)
But yes, Couriersud is awesome.
|
|
|
|
Re: Couriersud is a net list machine.....
[Re: Antny]
#321494 - 02/05/14 10:06 PM
|
|
|
> I've been keeping track (as best as I can understand it). Are we going to see DICE
> merge with MAME?
My understanding is that MAME works differently to DICE, but the netlists already created for DICE only need 10% changes to work in MAME. Whether DICE continues is up to Adam.
|
|
|
|
Re: Couriersud is a net list machine.....
[Re: ]
#321532 - 02/06/14 09:33 PM
|
|
|
> > I've been keeping track (as best as I can understand it). Are we going to see DICE
> > merge with MAME?
>
> My understanding is that MAME works differently to DICE, but the netlists already
> created for DICE only need 10% changes to work in MAME. Whether DICE continues is up
> to Adam.
Adam can do optimizations which MAME netlists can't support. Therefore DICE will have better performance.
The design goals of netlist were different:
- Replace the old discrete core at some time.
- Only very limited understanding of electronics and physics needed to convert schematics to netlist format.
- Have a tool to replace the various attempts in MAME to do discrete analog emulation.
- Be able to test netlists outside of MAME.
Discrete games were only a minor part of the plan. Pong just happened to be a perfect test bed since it had some challenging features like pulling TTL outputs to ground. When I started, I had no idea whether I'd end up with 1 FPS or 60FPS.
I hope that the netlist format encourages more people to contribute. Nearly all pre-90 game had some sort of output filtering with audible impact. Most of this yet is not emulated in MAME.
|
|
|
|
Re: Couriersud is a net list machine.....
[Re: couriersud]
#321539 - 02/06/14 11:35 PM
|
|
|
Thanks Couriersud for your HARD work involved to improve the discrete emulation in Mame!
|
|
|
|
Re: Couriersud is a net list machine.....
[Re: couriersud]
#321546 - 02/07/14 02:35 AM
|
|
|
> Adam can do optimizations which MAME netlists can't support. Therefore DICE will have
> better performance.
This is a pretty common issue no matter what emulator you compare MAME to.
Hopefully you'll figure out some way of running the calculations on a gpu 
|
|
|
|
Re: Couriersud is a net list machine.....
[Re: ]
#321603 - 02/07/14 08:06 PM
|
|
|
> Hopefully you'll figure out some way of running the calculations on a gpu
We had some discussions on solving linear equations recently. The typical matrix size observed for sound circuits (relatively small) will make the overhead most likely more expensive than what you could gain.
In addition, for diodes and other non-linear circuits like BJTs the matrix has to be recomputed on every step of the Newton-Raphson algo.
The logic subsystem is build around a timed queue and thus not really suited for GPU parallel solving.
FPGAs could be an option - read netlist, compile FPGA code and run. But depending on the communication rate between cpus and the discrete circuit the overhead again may be more expensive than what you gain.
Edited by couriersud (02/07/14 08:09 PM)
|
|
|
|
Re: Couriersud is a net list machine.....
[Re: couriersud]
#321627 - 02/08/14 12:05 AM
|
|
|
Some more beef on matrix (or problem) sizes and performance:
Citing from
http://vtb.engr.sc.edu/vtbwebsite/downlo...a%20Methods.pdf
"We have used Graphics Processing Units (GPUs) to acceler-
ate the solution of the types of equations typically encoun-
tered in dynamic system simulators. Compared to commer-
cial matrix solvers that run on a CPU, we realized speedups
ranging from 5 (for system size ≈ 700) to 460 (for system size ≈ 5800)."
I estimate MAME netlists to be well below a system size of 100. Thus overhead would kill performance. We are calculating realtime sound, i.e. at a rate of 48000 Hz. Thus you need to pass a couple of matrices 48000 times a second to the GPU. I assume, that the transfer CPU-GPU is designed to cope with Screen Refresh frequency (order 50 Hz).
But if interest (or need) arises, the netlist model abstraction can easily be extended to use a GPU solver instead of the Gauss-Seidel and Gaussian Elimination solvers currently used.
|
|
|
Index 
Threaded 
Couriersud is a net list machine.....
Reply
[Re: 
