Dynarec Completed For Psx4all Project.


X-Code posted on Aug 20 2006 at 06:55 PM said:
I tend to check this topic for PSX4GP2X Specific stuff B) .


Hey... normally it would be WE who should be awaiting your news ;)

Or maybe you want to send me your beta so i could try a few games? :unsure: lol

Have you tried any of the Oddworld games? i have both and despite their graphical problems, they both run at good speed, especially the first

Crash Bandicoot 3 has some graphics missing as far as I remember

I think I remember Silent Hill was unplayable due to graphical problems but I can't tell what it was, unless I rip the game again

Anyway, testing is useless for us unless the new beta comes out ;)

I think i'm going to try R-Type Delta and Wipeout 3, as i haven't ripped them before :)
 
Last edited by a moderator:
Wasn't intended to sound harsh, I test loads of games and I'm coding " T.I.M. - The Imploding Mutant "... And I like to keep some order when possible :lol:


You can't have the beta, it's mine, all mine, it's my treasure B)... I'll repeat, I won't output any stats until I'm told to do so... As for testing on the previous beta... No point, things are simply not the same :lol:.
 
I know its probably a dumb question and it has been asked a thousand times before, but (cover & duck!) couldnt we "just" use the 2nd CPU for the sound emulation?
 
Wolfsclaw posted on Aug 21 2006 at 08:21 PM said:
I know its probably a dumb question and it has been asked a thousand times before, but (cover & duck!) couldnt we "just" use the 2nd CPU for the sound emulation?

Nope, not yet. For more details nag the guys, but I can't grant that you wont get flamed :p
 
Last edited by a moderator:
afaik using the second cpu can slow things down when used 'incorrectly'. the problem seems to be that both cpu's have to share the ram acces and therefore the second cpu can block the other one. and its tricky to code the right way to make profit out of the second cpu so if zodttd will use it sometime it will probably be one of the last things he wil implement cause it takes up much time. i hope this was right.
 
Vollgasasi is pretty much right.
The second cpu might be able to be used for the GPU, but that's Unai's plan.
Having the GPU in there makes more sense to me, and is probably just as hard to do as the SPU. Considering the SPU doesn't just load WAV files, but stream's DMA accessed data from the ISO, it's a lot more complicated than I can handle at the moment. I understander sound is important though. Keep in mind the last beta out had 50% of the time being played allocated to CPU emulation, and the other half going to GPU emulation. With the dynarec in place, that 50% will go down, and if Unai is able to magically get the GPU on the second core, that 50% will go down as well. Then there will be alot more room for SPU (sound emulation) on the first core.

As for the dynarec status...I redid the caching system, it's working on every ARM handheld I have, but the GP2X! I'm still bug hunting. :p

I hope to have this bug figured out soon. It's more stable than the version sent to X-Code, but without this being fixed, it's still not compatible to fully play a game it seems.
 
Hi everyone,

I fixed the bug related to the GP2X version dynarec. I switched toolchains and went moved over to an SDL build, and the bug went away. Considering I went through every bit in the outputted recompiled code and nothing was wrong, it could of very well been the old toolchain I was using.

The move to SDL didn't seem to impact performance much at all, due to paeryn's supurb job.

I won't comment much on the dynarec's performance until I hear back from X-Code, but I expect it to perform similary to the interpreter at this point. This is because I think something's throttling it's performance, as even turning off interlacing doesnt affect performance any, as if there's CPU processing to spare. I will see if Unai can help me with finding out what's holding this code from it's full potential, if thats truly the case.

There's a lot of work to do on the dynarec to improve it's speed. There's some "dumb" things I had to do to get it to work on ARM, and they slow it down. And every instruction that no longer needs to be recompiled, not only makes it faster by not having to be executed, but also faster since it leaves more room in the dynarec memory/cache for other instructions. The version of psx4gp2x sent to X-Code only uses half the maximum memory I could potentially use for the dynarec memory/cache.

I'll see what I can do to improve performance in the meantime. :)
 
Thank you very much for the hard working ZODTTD.And for the daily reporting ;)

My english is bad, i know it :D
 
That's good news, as usual!! We all know you are doing a good job, you are becoming some kind of super-hero, an exemple for any programmers like me, who just starts coding C :lol:

I had an idea : now that the recompiler is working and that you are optimising it, do you plan to release the source of earlier PSX4GP2X? I mean, you may be working on the dynamic recompiler while somebody else would work on the classic emulator (static recompiler, something like that?)
I don't know what performance we can get by optimizing the old emulator though!


And for the question everyone is waiting... Do you have plan for the next release?? :p
 
Great Job Zodttd, we knew you would fix that bug sooner than later. I doubt that after all this hard work he will release a source, because its hell for you! LOL!
 
FQ22 posted on Aug 22 2006 at 04:35 AM said:
Great Job Zodttd, we knew you would fix that bug sooner than later. I doubt that after all this hard work he will release a source, because its hell for you! LOL!


i talked about the old emulator, which is now no more developped, since the dynarec seems to be a complettely rewritten emulator, am i wrong?
 
Last edited by a moderator:
Correct, but another way to improve upon that emulator may require some very hard work. If he does release the source, I hope you do a fantastic job on it just like I hope zodttd continues to improve his psx4gp2x project.
 
Back
Top