hrydgard
Gangs Of London hangup and graphical problem
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
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]
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 https://github.com/hrydgard/ppsspp/issues/5335
Look at the people and the shadows
psp
ppsspp
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]
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 ?
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...
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]
Hm, I wonder if each async fd gets its own async worker thread, or if they share one if there are multiple..
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]
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...): https://github.com/jpcsp/jpcsp/commit/8a1dad749dd0b04de986a506e04a552c26895a6e
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]
There's a graphics glitch on the ground in vulkan backend if hardware tessellation is ON
(VULKAN)
hw tessellation on
vk_hw_tessellation_on.ppdmp.zip
hw tessellation off
vk_hw_tessellation_off.ppdmp.zip
(OPENGL)
(SOFTWARE)
There are no pedestrians in the game，this leads to some tasks that cannot be completed
What's your device?
您的设备是什么？
I'm on the computer
What's the GPU/CPU and graphics backend are you using while playing this game?
Gameplay bugs like that will probably not be affected very much by specific hardware..
您的设备是什么？
我在电脑上
您在玩此游戏时使用的 GPU/CPU 和图形后端是什么？
CPU：AMD Ryzen 7 4800U with Radeon Graphics 1.80 GHz GPU：AMD Radeon(TM) Graphics ppsspp ver. v1.14.40/ Direct3D 11 backend
I re-test this game ,I find that menu crash problem solved by #13368. Pedestrians issue unusual Only load game from softgpu ,pedestrians appear UCES00113-softgpu.zip UCES00113-opengl.zip
appear Pedestrians crossing the wall
The characters in the plot task will go through the wal，this is in ppsspp ver. v1.14.40 Bugs appear
