Source
- URL: https://x.com/preshing/status/1977058692693655623
- Author: Jeff Preshing (@preshing)
- Posted: 2025-10-11 17:08:18
Branch
1/ @preshing
@rfleury
That’s a very cool way to exploit parallelism in branchy CPU code! Never seen it done that way before. I imagine it could also be applied to a specific subsystem (like a physics engine) while keeping the rest of the app single-threaded if desired… Thanks for posting!
2/ @rfleury
Definitely! I recently converted a few big parts of the debugger codebase over to this style. As you say, I kept the UI / control threads (historically single core), but moved asynchronous background processing (as well as the PDB -> RDI conversion) to being completely within this style.
Ultimately what I hope to do is convert the UI and control threads to this “wide” style of code, so that they are simply more flexible, in that they allow me to make them as wide as I want. Once the code contains that extra information, both single and multi-core are on the table, whereas right now, only single core is on the table.