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
Less of an issue and more of an FYI for others about threads and IGL.
I was getting tons of threads (20 per mesh processing step I did) being run which was making slurm and our IT managers grumpy. I tracked it down to identify that the signed distance function in PCU was causing the threads. Looks like there is an environment variable flag that can be set for igl that lets you limit how many threads it uses (https://github.com/libigl/libigl/releases).
Someone can add the following to their bash rc:
export IGL_NUM_THREADS="1"
This will limit IGL to using only 1 thread. I found no performance difference with point clouds of ~250k points using thread = 1 vs the default that was spawning 20+ threads.
Hopefully this helps someone else. Alternatively, there could be a way to add this into the environment variables dynamically within PCU is this is desired as a user modifiable parameter.
The text was updated successfully, but these errors were encountered:
Less of an issue and more of an FYI for others about threads and IGL.
I was getting tons of threads (20 per mesh processing step I did) being run which was making slurm and our IT managers grumpy. I tracked it down to identify that the signed distance function in PCU was causing the threads. Looks like there is an environment variable flag that can be set for
igl
that lets you limit how many threads it uses (https://github.com/libigl/libigl/releases).Someone can add the following to their bash rc:
export IGL_NUM_THREADS="1"
This will limit IGL to using only 1 thread. I found no performance difference with point clouds of ~250k points using thread = 1 vs the default that was spawning 20+ threads.
Hopefully this helps someone else. Alternatively, there could be a way to add this into the environment variables dynamically within PCU is this is desired as a user modifiable parameter.
The text was updated successfully, but these errors were encountered: