3 ms·
Would it be that difficult to just queue up those threads on a single 32 bit processor instead of sending them off to separate cores? It wouldn't be optimal, b
by Zikes 13y ago
Would it be that difficult to just queue up those threads on a single 32 bit processor instead of sending them off to separate cores?
It wouldn't be optimal, but at least they wouldn't have to completely write off older or lower-powered devices.
- lttlrck 13y agoDistributing threads and processes across available cores is the job of the OS, not the application.
- Eiwatah4 13y agoIn Rust, (right now) you can choose between a 1:1 or a N:M mapping between OS threads and Rust tasks. With N:M threading, the runtime necessarily does (some of) that internally.
- pcwalton 13y agoIncidentally, Servo uses both: we use 1:1 for the parallel layout worker threads and M:N for the script and layout tasks.