Marcel

Has this become any worse recently? I don't know exactly what updates made it happen, but between February '22 and now, `ld.exe` turned into consuming 5-10 times the time for...

Unlikely, I only have Windows Defender. I just disabled it, and still counted 80+ seconds for something that was 15s, maybe 20s. Performance of v12.1 `gcc.exe` appears unchanged though.

Thanks, it's appears to be better by 10% over multiple some test runs, but that's not the magnitude I was hoping for. I also may need to revert the project's...

Good idea, I just tested that and I'm back at 15s.

Sounds good: Dropping Py2, sticking to a single encoding as far as data is ingested into GemRB, and relying on fmtlib. Should put a lot of things back into maintainability.

Yeah, I just think that implementing [this](https://www.cplusplus.com/reference/cstdio/printf/) with a sufficiently large feature subset, in a safe and performant way saves us from rewriting every formatted line but is that less...

I noticed that an RPi 400, with specs that should outnumber old Pentium platforms, really suffers from framedrops in some situations, like when flying over a map. Profiling shows that...

> I doubt that setting a uniform variable is actually a bottleneck. likely its caused by the fact that we keep switching programs and the call to `SetUniformValue` is just...

> this issue was opened because SDL _is_ adding an API for that. I've tried to find a roadmap or updates but not sure if and when there is something...

I have played around a bit here, it's actually easy I just never had that use case before: ``` 1. glGetIntegerv(GL_CURRENT_PROGRAM, &programID) where safe to use 2. glGetAttachedShaders(programID) 3. glDetachShader()...