I suspect the current implementations of parmapfold and parfold are far from optimal
UnixJunkie opened this issue · comments
One example: parallel folding over a list will first convert it into an array.
I think this is unnecessary.
I may send some propositions (as code) soon.
hmmm... Not sure I will send code, parmap's code is quite complex in fact.
Or, I can send some standalone code, as a proof of concept.
can this issue be assigned to me? I may look into it when my poster (due for next week...) is ready
In theory, I think parfold should be even faster than parmap, because less data needs to be serialized back to the father process.
Not grounded on facts. Just a suspicion.