You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
could someone please elaborate on @timholy 's concern about thread safety here: #132
i ask, because i've been having a problem writing .mat files from julia inside a pmap. julia's added procs are processes, not threads, correct? and if each process writes to a separate .mat file, shouldn't all be okay? yet doing so, what was eventually getting stored to disk was on occasion not what i wrote to it. strings were getting truncated or pre-pended to. numbers were changing values. all in a non-deterministic way. only single-threading matwrite by passing the data back to the main thread solved the problem.
The text was updated successfully, but these errors were encountered:
Yes, Julia uses processes for pmap. If you see a problem while accessing different files from different processes, then this can also be a problem with the file server -- I've seen systems where even reading (!) 100 files in parallel would return wrong results. That's obviously a system problem, I assume the NFS server was at fault.
From your description it sounds as if HDF5 is not at fault.
could someone please elaborate on @timholy 's concern about thread safety here: #132
i ask, because i've been having a problem writing .mat files from julia inside a
pmap
. julia's added procs are processes, not threads, correct? and if each process writes to a separate .mat file, shouldn't all be okay? yet doing so, what was eventually getting stored to disk was on occasion not what i wrote to it. strings were getting truncated or pre-pended to. numbers were changing values. all in a non-deterministic way. only single-threadingmatwrite
by passing the data back to the main thread solved the problem.The text was updated successfully, but these errors were encountered: