Gangs Of London hangup and graphical problem

[daniel229]

it may get invalid addresses in the title sometime.walking in the street also may get invalid addresses and the cars and people are drifting in the sky.

also get these errores after intro video.
52:57:445 user_main E[KERNEL]: HLE\sceKernelThread.cpp:2945 sceKernelResumeThread(290): thread not suspended

same thing happen in the demo
demo download
http://1drv.ms/1xuFlYi

a savestate for demo that driving in the street.it will see the cars drifts in the sky and invalid addresses when driving.
http://1drv.ms/1xuFvit

[unknownbrackets]

Well, a bunch of memory errors don't happen when jit is disabled, so I suppose it's jit related.

There's a function at 088640E0 that always has the memory errors, in the demo.

Oh, darn, hit a memory error just now with jit off, but less... hmm. But, I'm not actually seeing any cars in the sky.

-[Unknown]

[daniel229]

It's a little higher than the road,not just the cars,people also and softwareGPU seems fine, but the graphics are terrible,not sure, like #5335

Look at the people and the shadows
psp

ppsspp

[unknownbrackets]

Oh, I see, I was looking for a bigger difference. Hmm, it probably is the same cause then... grr... but there's also the invalid address thing.

-[Unknown]

[sum2012]

I use Eur version UCES00113 ,get invalid address before menu or in the menu press "create profile"
v1.0-25-ga8a7adc debug log
https://drive.google.com/file/d/0B3OaSdeV0L8kMXN0TWU3OE9mWDA/view?usp=sharing
@unknownbrackets
JPCSP fix Eur version in http://code.google.com/p/jpcsp/source/detail?r=2072
Does PPSSPP need the same change ?

[hrydgard]

That change can't be applied directly as we currently don't even create an "async thread" in the emulated PSP, seems that maybe we should...

[unknownbrackets]

So, it's probably memory allocation related... we indeed should have io async HLE threads. We could try cheating it just by allocating stacks for fds to see if it actually helps the issue.

-[Unknown]

[hrydgard]

Hm, I wonder if each async fd gets its own async worker thread, or if they share one if there are multiple..

[unknownbrackets]

I think you can independently set the thread priority per fd, so I assume each one has its own thread. I have not done deep testing in that area yet...

-[Unknown]

[unknownbrackets]

This is still happening in the latest git build, even after implementing async IO more correctly.

Just a note, the referenced JPCSP change is here, and it might still be relevant (I had trouble detecting any memory allocation in my tests, so thought it was using kernel RAM...):
jpcsp/jpcsp@8a1dad7

That said, increasing memory allocation for async IO threads only makes this crash in a different way. I suspect the issue JPCSP fixed was a different issue.

That said, it doesn't look like things are floating anymore, at least...

-[Unknown]

Game also have a performance issue on create profile part.

recording.ppdmp.zip