openmp for nn

[chtchelkatchev]

I admire your program and began to actively use it to process Lammps data. Unfortunately, some operations are slow, especially the nearest neighbor search. Is it possible to introduce at least OpenMP parallelization into the code?

[jorisparet]

Hello,

Thank you for your message and your kind words. 🙂

When computing nearest neighbors, do you use a specific method (i.e. "fixed", "sann", or "voronoi")? This part of the code is managed either by the Fortran backend (for "fixed" and "sann") or pyvoro (for "voronoi"), so it should be relatively fast compared to other parts of the code that are 100% Python.

I do not have much time to keep maintaining this code unfortunately, but I will try to see what I can do to parallelize this computation one of these days.

[chtchelkatchev]

Hello,

Thank you very much for quick answer. I noticed very slow performance of sann and auto algorithms for nn-search. Voronoi is much faster, surprisingly. Fixed algo i did not use yet, but plan to use.

Best regards, Nikolay

ср, 12 июн. 2024 г., 16:46 Joris Paret @.***>:

Hello,

Thank you for your message and your kind words. 🙂

When computing nearest neighbors, do you use a specific method (i.e. "fixed", "sann", or "voronoi")? This part of the code is managed either by the Fortran backend (for "fixed" and "sann") or pyvoro https://github.com/joe-jordan/pyvoro (for "voronoi"), so it should be relatively fast compared to other parts of the code that are 100% Python.

I do not have much time to keep maintaining this code unfortunately, but I will try to see what I can do to parallelize this computation one of these days.

— Reply to this email directly, view it on GitHub https://github.com/jorisparet/partycls/issues/12#issuecomment-2163055859, or unsubscribe https://github.com/notifications/unsubscribe-auth/ANQOWASUCKNDDLQMUOOCSBLZHBGNBAVCNFSM6AAAAABJEUO5O2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCNRTGA2TKOBVHE . You are receiving this because you authored the thread.Message ID: @.***>