file: README.txt
author: Donnie Newell - den4gr@virginia.edu
last modified: 18Jan2012
The optimized version of particlefilter has significant changes from
the original particle filter kernel prior to December 2011. A brief list
is included below:
- float data types changed to double because of overflow issues
- increased synchronization to correctly handle shared memory and partial blocks
- additional memcopies to print intermediate results for each frame
These additions result in different results and runtimes than previous versions.
Single: single precision
Double: double precision