After impl. of Create_graph, I tried to test it in a clustering_separate placeholder, by trying to create graph for all clusters. This showed a very slow speed. (~300 sec vs. less than 3sec for the full clustering_separate function in WCP)
This was fixed and more validations were effectively done:
use C array instead vector, this is useful for extreme conditions, which was the case with the bugs below.
I was creating graphs for all clusters which was not the case in WCP, for ref. event, only 3 large clusters are used for creating graph
two actual bugs:
max_wire_interval, etc were not passed out from sampler, this is fixed by adding an "aux" DS for the sampler
nticks per slice diff. were not considered when coping WCP code, this is also fixed.
Introduction
After impl. of Create_graph, I tried to test it in a clustering_separate placeholder, by trying to create graph for all clusters. This showed a very slow speed. (~300 sec vs. less than 3sec for the full clustering_separate function in WCP)
This was fixed and more validations were effectively done:
After the above changes, the WCT
Create_graphruns pretty fast now. Also the edges are largely consistent with WCP.PS: cheat sheet for WCP: https://hackmd.io/@HaiwangYu/HkOI9XZv0