On this one:

http://www.stochasticlifestyle.com/multi-node-parallelism-in-julia-on-an-hpc/

it looks like you're running performance-critical code at global scope. If you 
put it in a function (and check that it's type-stable, etc), your single-
threaded code might outperform your parallel version. (Of course, you can 
combine the two to do even better, sometimes.)

See http://docs.julialang.org/en/stable/manual/performance-tips/

--Tim

On Monday, February 22, 2016 11:34:20 PM Chris Rackauckas wrote:
> Hey,
>   I wanted to let everyone know I put a few blog posts out there on Julia.
> You can find them here:
> 
> http://www.stochasticlifestyle.com/
> 
>   I am by no means an expert on Julia, but I put some tutorials for some
> ways I have been using Julia, like how to use Julia on the XSEDE Comet HPC
> (which you can test out too via a startup allocation!), optimizing some FEM
> codes, etc. After I publish a few things I'll get back onto this and
> document some experiments I am doing like interfacing with the Xeon Phi and
> writing stochastic (partial) differential equation solvers. I hope these
> tutorials helps somebody out. And feel free to comment. I still have lots
> to learn about Julia as well.

Reply via email to