Re:Cinder: Meta's internal performance-oriented production v...
by gpderetta on Thursday September 22, 2022 @08:17AM
(#32937819)
As someone who uses python in their day job, "IO bound" is a lie that is perpetuated to cover for the failing of the language. IO is very fast these days and python has the capacity of choking the cpu even on genuinely io bound problems .
For example, do you think that python is an appropriate language to write pylint?
Re:Cinder: Meta's internal performance-oriented production v...
by ehutch79 on Thursday September 22, 2022 @01:01PM
(#32941648)
Network DB access is what I'm thinking about, not reading less than a K from on disk file.
Also, any external API is an example.
To answer your question; "it depends". Is pylints speed an issue? If not, then it's fine, if it is, the sure write it in go or rust or whatever.
Javascript has a great example of this in esbuild.
Re:Cinder: Meta's internal performance-oriented production v...
by dekhn on Thursday September 22, 2022 @11:13AM
(#32939993)
By definition, an IO bound problem does not "choke the CPU", because at that point, it would be CPU-bound.
When you spend more time io bound, it’s not an issue. If your python code is choking on cpu time, a lower level language is probably more appropriate.