search

DESI-UR / VAST

Void Analysis Software Toolkit
https://vast.readthedocs.io/en/latest/
BSD 3-Clause "New" or "Revised" License
9 stars 8 forks source link

VoidFinder crashing based on batch size and/or mask parameters #98

Closed QuiteAFoxtrot closed 8 months ago

QuiteAFoxtrot commented 1 year ago

Running on the BGS_c000_ph000.dat, batch_size=1000 is causing the Cell_ID_Mem to grow huge which is indicative of a "big ol empty area" in the mask, causing VF to search a big area.

However, on batch_size=10, the main hole growing completes in .08 seconds and finds a bunch of holes but then crashes in volume cut - and finds number of holes 17,179,334 which is very close to 17,435,384 aka the number of holes to search when run in larger batch sizes.

Need to investigate behavior of VF when the Ra-Dec-Z style mask has valid regions that touch the border and "wrap" to the other side of the sphere, possibly related since this BGS data has a mask of this type.

QuiteAFoxtrot commented 8 months ago

working on #96 discovered a bug in the Cell ID Generator whereby VoidFinder returned early in Periodic mode due to returning a batch of 0 cells to search before actually going through the whole survey - with large batch sizes returning 0 cells is very unlikely which likely explains why this hasnt been a problem with non-periodic surveys and the like