Originally posted by bachchain
View Post
Announcement
Collapse
No announcement yet.
Cemu Emulator Plans For 2022 With Going Open-Source, Aiming For Linux Support
Collapse
X
-
Not to mention the 1000 messages I received along the lines of "This emulator is doomed when it doesn't go open-source". It makes me want to prove them otherwise.
The adult in me just want to make sure their work doesn't go to waste like various items of closed-source freeware that withered when maintainership dried up.
- Likes 2
Comment
-
Originally posted by tildearrow View PostLet's respond to the roadmap:
LLVM as a CPU JIT backend
This wouldn't be a good idea, even if LLVM outputs faster code than the custom recompiler.
Reason is because LLVM is not optimized for compilation times which means that unless ahead of time compilation is employed, this may lead to increased stutter and lag spikes.
ACO vs. LLVM benchmarks prove this.
Also, an emulator can just interpret the binary while waiting for the compiler in a background thread, while for shaders the execution of the game blocks until the shader has been compiled, and any new code (except for after starting a new game) is likely to take a small enough proportion of execution time that it doesn't cause stuttering.
- Likes 2
Comment
-
Originally posted by archsway View Post
One of the big reasons for switching to ACO was so that compiler fixes didn't have to wait for a new LLVM version but could just be implemented in ACO immediately. This isn't such a problem for an emulator, because LLVM's CPU backends are more mature and don't have so many bugs.
Also, an emulator can just interpret the binary while waiting for the compiler in a background thread, while for shaders the execution of the game blocks until the shader has been compiled, and any new code (except for after starting a new game) is likely to take a small enough proportion of execution time that it doesn't cause stuttering.
Comment
Comment