Il giorno gio, 13/06/2013 alle 18.06 +0100, Peter Cock ha scritto:

On Thu, Jun 13, 2013 at 6:02 PM, Nicola Soranzo <soranzo@crs4.it> wrote:

...

...

There are two issues here, one that we would like to be able to handle multi-threading better via Galaxy configuration on a per-tool level, but for now it must often be hard coded or done via tool-specific environment variables.

Regarding the configuration of multi-threading for BLAST+ tools, instead of hardcoding "-num_threads 8" in the XML files, I'd like to do implement something like this:

http://toolshed.g2.bx.psu.edu/repos/jjohnson/cdhit/rev/cca0838c1597

What do you think? Should I prepare a pull request to discuss at GCC2013?

That's a sensible idea, something like $BLASTTHREADS as an environment variable?

I was thinking something like BLAST_SITE_OPTIONS='-num_threads 8', which could also include other parameters (e.g. memory limits) in the future.

I'd still prefer something built into Galaxy like $THREADS or $GALAXYTHREADS or whatever which can be sets with a default value and adjusted in the per-tool job runner setup (e.g. send BLAST jobs to this cluster queue with 16 threads).

This is definitely a good general topic for the tool authors and/or BLAST wrapping BoF sessions at the conference: http://wiki.galaxyproject.org/Events/GCC2013/BoF/ToolDevelopers http://wiki.galaxyproject.org/Events/GCC2013/BoF/GalaxyBlast

Peter