This is not a completely crazy idea. We could take the building blocks from OpenBLAS or BLIS and do the multi-threading in Julia. The benefits will be that we will not have the OS threads vs. Julia threads issue, and improve on the current state of threading in OpenBLAS.
Of course, this will take a bit of time. -viral On Saturday, March 14, 2015 at 11:20:06 PM UTC+5:30, Jameson wrote: > > in that case, one usage might be that it's necessary for porting a > competitive BLAS into julia. > > On Sat, Mar 14, 2015 at 1:26 PM Erik Schnetter <schnet...@gmail.com> > wrote: > >> As long as linear algebra is handled by BLAS, using multi-threading in >> Julia isn't really necessary. If the matrices are large enough to warrant >> multithreading, then using a multi-threaded BLAS is usually more >> efficient... >> >> -erik >> >> On Mar 13, 2015, at 16:54 , Tobias Knopp <tobias.kn...@googlemail.com> >> wrote: >> > >> > you can start with a simple matrix-vector or matrix-matrix product. >> imfilter (Images.jl) is also a function that can benefit from >> multithreading. >> > >> > Am Freitag, 13. März 2015 04:52:37 UTC+1 schrieb Viral Shah: >> > I am looking to put together a set of use cases for our multi-threading >> capabilities - mainly to push forward as well as a showcase. I am thinking >> of starting with stuff in the microbenchmarks and the shootout >> implementations that are already in test/perf. >> > >> > I am looking for other ideas that would be of interest. If there is >> real interest, we can collect all of these in a repo in JuliaParallel. >> >> -- >> Erik Schnetter <schnet...@gmail.com> >> http://www.perimeterinstitute.ca/personal/eschnetter/ >> >> My email is as private as my paper mail. I therefore support encrypting >> and signing email messages. Get my PGP key from >> https://sks-keyservers.net. >> >>
