Project64 source code

[HatCat]

Smasher's right about GoldenEye 007. It fluctuates a hell of a lot in GE all the time just during the intro...with Limit FPS turned off it keeps wildly changing from like 7 to 200 VI/s rapidly on my build of the EXE.

On Project64 1.7.0.50 alpha 23, however, the VI/s remains fairly steady (with some minor exceptions). Point is, it is comparatively worse by far in the current Project64 pre-2.2 EXE than it was in 1.7.0.50.

However, whether I build the EXE using VS2008 or VS2013 has no effect. The result is the same--this is not a fault of the 2013 edition of Microsoft Visual C. It's something that has been changed recently about the Project64 core.

[HatCat]

Changing CPU core style from Recompiler to Interpreter appears much better in consistent performance with GE 007, so it is a regression in the recompiler core.

RSP LLE/complete HLE appears to affect nothing about said recompiler speed issues.

[HatCat]

A combination of Interpreter CPU + CounterFactor=6 (rather than 1) seems to pretty much dissolve the issue entirely with the VI/s fluctuations. Recompiler CPU + CF=6 does not fix it.

VI Refresh Rate / AI count per byte have no effect whatsoever on the fluctuating. Neither a very high nor a very low VI refresh rate matter at all.

[=X= Smasherx74 =X=]

yo wtf why does changing the recompiler to incompiler actually make the game "playable" (not playable), it starts out at full 60fps then goes around 50 - 55fps, then when i get into the menu screen it slows down and once In a level it goes to unplayable 15-30 fps then drops to 1-5fps randomly

This still doesn't work though...

Im mad

and its even slower in my builds which is also a larger file than cathat

[HatCat]

Turning off Register Caching and turning on Advanced Block Linking appear to "help" the issue, but still sounds nowhere near as good as interpreter and still has more FPS nosedives than the interpreter. (Protip: Recompilers are worse than interpreters. Don't give a fuck how they're coded; if they exist, then they're worse.)

There is one fix I have found to get the recompiler to sound as good as the interpreter and not have all the shitty dropouts of the FPS: Virtual Lookup Table.

First you have to disable ALL of the self-mod code methods.
Then switch from Physical to Virtual lookup table.

Upon doing the second change, it sounds and feels at least as good as the interpreter does.

Screenshot of my new RDB settings:


So it looks like your FPS sinking was caused by the recompiler's Physical Lookup Table getting even bulkier and wasting even more time compiling x86 code than normal, than what Project64 1.7 did, hence it sounding worse in pj64 2.0 ... but if you use virtual LUT it sounds much better, would you agree?

[=X= Smasherx74 =X=]

all you are doing here is troubleshooting a faulty emulator build. Your build and my build have problems for me. With yours and your settings the game runs seemingly better, but still the same problem.


Also what are you doing differently than me that is causing your build not to immediately drop FPS like mine does, yours starts off seemingly fine.

[=X= Smasherx74 =X=]

Okay so it's not just building from the github, it's the github it's self being faulty. I've tried EmuCR's build and it has the Same exact problem.

F zero,
Goldeneye
SSB64

all drop FPS and are unplayable.

[RPGMaster]

Originally Posted by =X= Smasherx74 =X=

Okay so it's not just building from the github, it's the github it's self being faulty. I've tried EmuCR's build and it has the Same exact problem.

F zero,
Goldeneye
SSB64

all drop FPS and are unplayable.

I tried HatCat's build and Fzero & SSB64 ran fine. Couldn't tell a difference between original 2.1 and Hat's build.

[=X= Smasherx74 =X=]

Originally Posted by RPGMaster

I tried HatCat's build and Fzero & SSB64 ran fine. Couldn't tell a difference between original 2.1 and Hat's build.

Lol, it's hanging, obviously. I mean if you don't emulate the games you might not tell but they are hanging more than they should... More than the 1.4 build anyways.

There is something wrong with the github entirely. But the orignal 2.1 seems to be fine.

[HatCat]

Originally Posted by =X= Smasherx74 =X=

all you are doing here is troubleshooting a faulty emulator build.

Originally Posted by =X= Smasherx74 =X=

Okay so it's not just building from the github, it's the github it's self being faulty. I've tried EmuCR's build and it has the Same exact problem.

Can't you get it through your skull yet?

Neither GitHub website nor the build environment (official one the main releases have been using) have anything to do with this. It's a regression in the SOURCE CODE ITSELF.

Seriously...you think that just because both EmuCR and pj64 forums build the source off of GitHub, that it means your GoldenEye issues are caused by GitHub itself?

I'm not arguing about this bullshit.