the `cat` thing is probably a better idea, but FYI having n_processes > n_cpus is fine and good if you're IO bound
-
-
-
especially true with network IO, but even disk IO you might benefit from 2 or 3 processes per CPU opening and reading
-
(i guess with 48 cores that's probably not true unless you're reading from many physical hard drives at once)
-
2 cpu workstation, only ("only") 24 physical cores. Lots of my operations are parallelizable, and I have a ton of data to analyze
-
I like tightening code, but my advisor just wants results, so almost anything that cuts down user time --> worth it. And RAM's cheap
-
I did some core use optimization on a test set, gains stopped being linear after . . . I can't remember, but well before 24 procs
-
I think I determined it was an IO issue by process of elimination, since the python profiler doesn't play nice with multiprocess mod
End of conversation
New conversation -
-
-
pass a bill banning home heating with fossil fuels, everyone switches to heating by bitcoin mining
-
I think
@Ctzn5 was thinking about something like that at scalehttps://twitter.com/Ctzn5/status/830174202959450113 …
End of conversation
New conversation -
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.