SPC7110 Test Build

Post by **Nach** » Sat Jul 26, 2008 11:46 pm

adventure_of_link wrote: That cache feature you guys were planning

Planning? Excuse me?

adventure_of_link wrote: would it help the S-DD1 games speed up as well?

It's completely unrelated. I also wasn't aware of S-DD1 being slow enough that it warranted any kind of cache development for it.

denzilla · Post by **denzilla** » Sun Jul 27, 2008 12:38 am

OT, but did I read right that zsknight worked on this as well? Thats cool to see he's sorta back after all these years. I haven't been here in awhile, is Pagefault still active?

I.S.T. · Post by **I.S.T.** » Sun Jul 27, 2008 1:17 am

Yes, though he hasn't posted in the forum in several months.

adventure_of_link · Post by **adventure_of_link** » Sun Jul 27, 2008 5:56 am

Nach wrote:
adventure_of_link wrote: would it help the S-DD1 games speed up as well?
It's completely unrelated. I also wasn't aware of S-DD1 being slow enough that it warranted any kind of cache development for it.

Well I thought the cache would help boost the SDD-1 games so.. -shrug-

byuu · Post by **byuu** » Sun Jul 27, 2008 6:05 am

denzilla wrote:OT, but did I read right that zsknight worked on this as well? Thats cool to see he's sorta back after all these years. I haven't been here in awhile, is Pagefault still active?

He wrote the initial implementation for ZSNES, a few years back.

denzilla · Post by **denzilla** » Sun Jul 27, 2008 12:46 pm

byuu wrote:
denzilla wrote:OT, but did I read right that zsknight worked on this as well? Thats cool to see he's sorta back after all these years. I haven't been here in awhile, is Pagefault still active?
He wrote the initial implementation for ZSNES, a few years back.

Ah, ok. Thought he was back

mnk · Post by **mnk** » Thu Jul 31, 2008 6:42 pm

Nach, I don't want to sound too pushy,
but would you mind posting a source tarball
with your changes,
so that some people can, like, build it ?
It seems to work, when run through wine,
but I'd rather without it.

Cause, AFAIK, since bountysource went down,
there's no public svn.

Post by **grinvader** » Thu Jul 31, 2008 7:42 pm

He'll probably make an update once spc7110 code is polished.
It's out when it's out, yak yak yak and so on.

Post by **Nach** » Thu Jul 31, 2008 11:50 pm

Okay, anomie and I rewrote pretty much all of neviksti's code and seriously optimized it like nuts. We still have a little bit more that we can optimize (and hopefully will), but we're MUCH faster now.

http://zsnes.sourceforge.net/zsnesw-spc7110-3.zip

This build offers over the previous build even more optimized decompression, and a persistent cache (which at this point I don't think we really need anymore).

Let me know what kind of FPS you get now, and how much CPU usage it takes (and list which CPU and MHz you have).

Also, please test save states and all that.

We're almost done here. Once I am, I'll figure out exactly what it is that I changed, so I can package a new source release.

If you run into any issues, let me know. If the persistent cache for whatever reason is causing issues, delete the *.idx and *.gfx files, and let me know about it.

Thanks.

Post by **odditude** » Thu Jul 31, 2008 11:56 pm

Nach wrote:anomie and I blah blah blah win win win

Is it a safe guess that Snes9x will be sharing in the love sometime in the near future?

Also, is the source release likely to be accompanied by a Windows binary, or is compiling the Windows version as painless as the Linux version?

franpa · Post by **franpa** » Fri Aug 01, 2008 12:10 am

you can use zget to compile it (most likely) hence the windows build will be very simple to compile.

I.S.T. · Post by **I.S.T.** » Fri Aug 01, 2008 3:12 am

Ok, running the intro of FEoFZ on fast forward makes the FPS go high enough to break the counter. o_O It doesn't properly display the current FPS most of the time. I get anywhere from 48-58% CPU usage on that. Without fast forward, I get 11-26% CPU usage.

In gameplay, it's 10% to 30% CPU usage without fast forward, 46-56% with.

My CPU is a Pentium Dual-Core 2160, which is basically a Core 2 Duo Allendale core CPU with half the cache. Its' default speed is 1.8 ghz with an FSB of 800 mhz. I have it overclocked to 2.4 ghz with a 1066 mhz FSB. I have speed step turned off so it doesn't interfere.

I'll check the other games later, except for that Soccer game. Soccer sucks. A lot.

Post by **Nach** » Fri Aug 01, 2008 10:42 am

Jonas Quinn noticed I made a mistake in the previous build, which affects the out of order cache, causing graphics to be lost.

Please re-download the previous build, and delete your *.idx and *.gfx files.

Post by **Nach** » Fri Aug 01, 2008 5:02 pm

Okay, _Demo_ pointed out some relics of neviksti's code still in our optimized version. I rewrote some of the tightest loops to use 64 less bytes of memory, and only loop 25% of the amount of loops it was currently doing.

http://zsnes.sourceforge.net/zsnesw-spc7110-4.zip

Let me know if I broke anything. If you also were getting a slow down on initial load in the previous build, that should be cut down now.

Test with the cache deleted, and let me know how it works.

I.S.T. · Post by **I.S.T.** » Sat Aug 02, 2008 4:49 am

CPU has not been changed. Running the intro on fast forward results in 49/50%-55% CPU usage. Without fast forward, it's about 10%-22%, roughly.

For some reason, sound cut out after I minimized zsnes to start typing here. o_O It didn't do that on the previous version. Turning it off and on fixes it.

bobthebuilder · Post by **bobthebuilder** » Sat Aug 02, 2008 7:09 am

Is the cache enabled by default and can it be disabled if it is unnecesary?
I noticed this line:
"We still have a little bit more that we can optimize (and hopefully will)"
Is this what was taken care of in this newer version. I will do more testing perhaps on older computers to get a better feel for the improvements. Great Job BTW!

Deathlike2 · Post by **Deathlike2** » Sat Aug 02, 2008 7:18 am

Pay attention to what was said relative to the time+build was released.

Hint: Obviously it's optimized even more since then.

neo_bahamut1985 · Post by **neo_bahamut1985** » Sat Aug 02, 2008 2:48 pm

Yep, just tested the recent build and tried out Far of East Eden Zero (J) and I just wanted to tell you (Deathlike2 and Nach) that
the game runs perfectly at 60/60 FPS. I can't believe how much Zsnes has progressed in the past couple of years!

Post by **grinvader** » Sat Aug 02, 2008 6:28 pm

... it hasn't. Not in the past year-and-a-half, at least. -_-
We've just added SPC7110 code on the bleeding stub that is 1.51.

Deathlike2 · Post by **Deathlike2** » Sat Aug 02, 2008 6:30 pm

I'm not responsible for the SPC7110 stuff.

The only bug that I can see is persistant cache corruption when the Rewind key is used.

zerodarkage · Post by **zerodarkage** » Sat Aug 02, 2008 6:30 pm

Nach

There is a missing graphic... only occurs in test build 3 and 4
A young Fairy graphic when she looks for north.

If needed i have the save state and the sram to demonstrate this bug
Ok !?!?

Deathlike2 · Post by **Deathlike2** » Sat Aug 02, 2008 6:35 pm

Make sure to delete the cache to double check if it is still the case.

neo_bahamut1985 · Post by **neo_bahamut1985** » Sat Aug 02, 2008 6:48 pm

grinvader wrote:... it hasn't. Not in the past year-and-a-half, at least. -_-
We've just added SPC7110 code on the bleeding stub that is 1.51.

Wait..I could've sworn they were in the middle of coding 2.0...

zerodarkage · Post by **zerodarkage** » Sat Aug 02, 2008 6:51 pm

Thanks a lot!!!

Deleting the cache files and loading the game via test build 4 solves the problem

Thanks for the tip!!!

Deathlike2 · Post by **Deathlike2** » Sat Aug 02, 2008 6:52 pm

zerodarkage wrote:Thanks a lot!!!

Deleting the cache file an loading the game via test build 4 solves the problem

Thanks for the tip!!!

You do realize that the cache reference was made in both 3rd and 4th builds? Please read more carefully next time.

neo_bahamut1985 wrote:
grinvader wrote:... it hasn't. Not in the past year-and-a-half, at least. -_-
We've just added SPC7110 code on the bleeding stub that is 1.51.
Wait..I could've sworn they were in the middle of coding 2.0...

Pay attention to what grinvader has said, and draw some better conclusions.