Am 12.04.19 um 13:26 schrieb Sven Schreiber:
Am 12.04.2019 um 12:42 schrieb Artur Tarassow:
>
> Yes, that's an example I had in mind. But given that some functions
> already natively seem to support multi-threading [at least for big
> matrices and certain linear algebra operations I frequently enjoy
> watching all cores fully demanded ;-)], there would be no real
> value-added in parallelizing stuff. For more complex task, one can
> (fortunately) always rely on your 'parallel_specs' package ;-)
>
So I guess one would have to identify those functions which do get
native multi-threading via BLAS or whatever. But with respect to
parallel_specs, are you actually using it in practice?
I use parallel_specs for running model estimation in parallel which can
be distinguished by different sets of regressors or so . To be more
concrete, I ran logit/probit as well ARDL type of models in the past. Of
course parallel_specs could be very useful for cross-validation purpose
as well.
Best,
Artur