DaemonMode multiprocessing without performance increase?

[pomates]

Hi,

I'm using DaemonMode, and it works fine. However, I'm now calling the solver in parallel with the same script. To make use of multiprocessing, I tried starting the server with julia -t auto .... However, using this, the clients run into problems when executing the code (seems like they access the variables of each other).

Using julia -p auto everything works fine, but it seems like the clients are still called sequentielly (at least the total runtime is the same as using no -p flag)

What could be the cause of this?

[dmolina]

Hi, thank you for using it. Have you tested with the async=true in the Daemon? You must be sure to use -t auto in the Daemon, like:

julia -t auto -e 'using DaemonMode; serve(async=true)'

[pomates]

Hi, using -t auto seems to cause some issues on actually executing the tasks.
It causes error writing files to the disk; however, the file name mentioned in the error log is the one of another client task currently being executed.