Closed mojaie closed 4 years ago
Here is a benchmark result of execution times of descriptor functions per 20 molecules, n=100, median. At least, SSSR caching seems to be very effective.
Execution times per molecule of descriptor functions , (20 molecules dataset, n=100, median).
Caching valence arrays (lonepair
, valence
and apparentvalence
) may work.
0d84cb7
@cache
macro enables reuse of intermediate results of molecular descriptor calculation for fast computation, but I don't know whether it is really effective.Tentatively I wll remove
@cache
from molecular descriptor functions except for computationally expensive and frequently used functions listed below.mincyclenodes
(for SSSR detection)coords2d