Hi everyone, I just asked the IRC list this, but I guess no one was there. I'm trying to grid a few million particles onto the yt grids, but checking if the left and right edges of my particles are inside every cell is slow: #find particles inside cell idx = na.logical_and(na.all(le < pos,axis=1), na.all(re > pos,axis=1)) #assign them to a grid ... #remove particles from future searches pos = pos[~idx] I start with the finest grids at the highest levels, and once the particles are gridded they're removed -- so the list of particles is always decreasing, and the coarsest mesh tends to have only a few particles. I'm guessing that somewhere this has been done before - I'd rather not reinvent the wheel. Has anyone used any relevant kdtree (or cython) code in the yt utilities that can select a subvolume of particles quickly? Thanks ahead of time! chris
I don't have anything better at the moment, but there's a fast algorithm in FLASH for doing this that could be ported over and made part of yt that would probably be pretty useful. Sent from John ZuHone's iPhone On Apr 15, 2012, at 10:40 AM, Christopher Moody <cemoody@ucsc.edu> wrote:
Hi everyone, I just asked the IRC list this, but I guess no one was there. I'm trying to grid a few million particles onto the yt grids, but checking if the left and right edges of my particles are inside every cell is slow:
#find particles inside cell idx = na.logical_and(na.all(le < pos,axis=1), na.all(re > pos,axis=1)) #assign them to a grid ... #remove particles from future searches pos = pos[~idx]
I start with the finest grids at the highest levels, and once the particles are gridded they're removed -- so the list of particles is always decreasing, and the coarsest mesh tends to have only a few particles. I'm guessing that somewhere this has been done before - I'd rather not reinvent the wheel. Has anyone used any relevant kdtree (or cython) code in the yt utilities that can select a subvolume of particles quickly?
Thanks ahead of time! chris _______________________________________________ yt-dev mailing list yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
Hi Chris, Britton ported CIC routines to yt a while back. They might work for what you're looking at; you can see them used to deposit particles in the universal fields, and the source is in the CIC Cython routine. -Matt On Sun, Apr 15, 2012 at 1:40 PM, Christopher Moody <cemoody@ucsc.edu> wrote:
Hi everyone, I just asked the IRC list this, but I guess no one was there. I'm trying to grid a few million particles onto the yt grids, but checking if the left and right edges of my particles are inside every cell is slow:
#find particles inside cell idx = na.logical_and(na.all(le < pos,axis=1), na.all(re > pos,axis=1)) #assign them to a grid ... #remove particles from future searches pos = pos[~idx]
I start with the finest grids at the highest levels, and once the particles are gridded they're removed -- so the list of particles is always decreasing, and the coarsest mesh tends to have only a few particles. I'm guessing that somewhere this has been done before - I'd rather not reinvent the wheel. Has anyone used any relevant kdtree (or cython) code in the yt utilities that can select a subvolume of particles quickly?
Thanks ahead of time! chris
_______________________________________________ yt-dev mailing list yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
participants (3)
-
Christopher Moody -
John ZuHone -
Matthew Turk