New LLE RSP Plugin (accurate, fast but untested?)

[HatCat]

In all honesty I see nothing more which could ever be done to this plugin.
If you think there could be more, you might just continue to enjoy the speeds from HLE'ing everything.
Nobody has reported any bugs for months, and I've tried since the beginning to make it perfect.

Maybe somebody can make a C++ experiment out of it using Bjarne Stroustrup's vector class and other object-oriented template designs, like what SGI started writing about on their site using C++ templates, but it really makes no difference because the C source is already the best at implementing the computer science of the RSP according even to their own older models.

In some ways, it is already like a re-compiler, but it uses the basic interpreter models because it's the most flexible, pure and portable to the vendor extensibility of adding in cycle-accuracy outside the limitations of this plugin spec. No such combination of advantages is feasible with an optimized re-compiler.

[the_randomizer]

So is rsp_pr5 the most recent version or will we need to compile it from the source?

[HatCat]

Usually if bin is released open-source then the src tends to be released with it if that's what you're asking.

[the_randomizer]

Originally Posted by BatCat

Usually if bin is released open-source then the src tends to be released with it if that's what you're asking.

I guess what I was asking if that fifth revision in the downloads of the first page of this thread was your most recent plugin version.

[oddMLan]

I found a bug in Conker's Bad Fur Day using the latest release. Some SFX and voices sound "muffy" (as if there were no interpolation applied). Easily noticeable in the intro with Conker cutting the "N" logo with a chainsaw.
Jabo's RSP and your RSP Public Release 4 don't suffer from this bug.
I could upload some audio comparing the affected audio and the desired result if it is necessary. FWIW, this game compresses voices and SFX using some custom MP3 codec.

Using Project64 2.1 and Jabo's DirectSound 1.7.

[HatCat]

I will check it. In a way it doesn't surprise me, because there was no direct, standard way to debug all outputs in sync with older RSP interpreter code bases in this new SSE-based execution loop.

In fact I had already fixed a bug before releasing with muffled MusyX pre-processed audio, in Star Wars and even Boss Games, but I guess it survives in those crazy Rareware custom microcodes. I'm sorry to say there was no physical way to guarantee the impossibility of a bug with so many rewrites, though I tried.

In the meantime, how are we doing on speed? I can't compare the SSSE3 build to the SSE2 one, or the previous builds posted to this thread. How much faster is this for people? How far behind it is it still from Jabo's RSP recompiler for something as basic as Mario64?

Originally Posted by nintendo1889

I guess what I was asking if that fifth revision in the downloads of the first page of this thread was your most recent plugin version.

What I'm not understanding is why it wouldn't be.

Am I posting updated releases to this thread falling behind what the latest binary really is, which I am secretly keeping to myself?

If you've been following those uh, what was it..."emucr" updates off my repo, pay no mind to that.

It is the latest stable release, like I said.
Also like I said, I see no point in ever continuing with this project, as of yet.

[ExtremeDude2]

Originally Posted by BatCat

It is the latest stable release, like I said.
Also like I said, I see no point in ever continuing with this project, as of yet.

He says as he pushes 4 updates

[HatCat]

eh.......


You weren't supposed to see that.
STFU!

Heh, but actually all that was, were commits I did to the source code I posted to the bin in this thread, which were not pushed to the GitHub first.
So actually in this case this thread was more up-to-date than the GitHub repository.

I honestly never meant for my GitHub account to be so widely known to everyone. It was just meant to stash my dru-I mean organize my shit as I was coding. I don't really think anyone would benefit by paying such attention to it. Just caring what I post here in this thread is fine.

Still got to do another release fixing oddMLan's report though. I thought I'd taken care of the last of those vector-component rewrite bugs but it looks like there's one more thing I have to attend to.

[the_randomizer]

Oy vey I feel silly

[HatCat]

Then you could at least explain where you were confused, because I am not seeing it. :/