Kemal Afzal
Kemal Afzal
Known issue, very likely ARM9 timing related. See #380
Sorry I have Intel iGPU, I can't debug these things. They probably just need even more work tiles or some other index starts overflowing. I need to come up with...
> Increasing the size of FinalTileMemory didn't do anything either. FinalTileMemory isn't dynamically allocated, it cannot overflow. Not really sure what could be the issue here. It seems to be...
oh no, there still might be undefined behaviour somewhere in the shaders. I'm going to run the shaders with mesa and a debug context eventually to see whether it warns...
thank you for debugging that. Regarding the barriers that is obviously a mistake. Though GL_SHADER_STORAGE_BUFFER is defined as 0x90D2 (which *should* result in a GL error), though if the driver...
The problem is that the amount of work done per variant is not known to the cpu at the time it issues the indirect compute calls (that's the reason they...
apparently you can run it from the console with `melonDS > log` (surprised that actually works).
melonDS 0.9.5 should be completely statically linked. From 0.8.3 to 0.9.5 the frontend was completely replaced. Redirection of the log output worked fine for me with the official Windows binary...
looks like it only works when running from mingw shell.
?