Open Parvfect opened 5 months ago
Profiling for one decoding iteration for the large code - 5 min per iteration
Precomputed symbol poss - could possibly make it even faster by getting the 2 or 3 lookups. Permuter is the major overhead currently. Not completely sure how I make it faster, but it must exist
Roman's idea which will work - adaptive searcher for the least number of symbol possibilities, use a data structure that can deal with sorting updation
Create tree
Need to run a large code for the cluster - max time let's say 4 hours or something. Need robust testing and parallaization