According to perf analysis, 71% of the total time is spent on updating the model. 23% is spent on calculating the entropy.
Some <1% is spent on matching and parsing is 1.1%.
In contrast, the stats for order-0 ranking are:
7.1% of the total time is spent on matching, parsing is 65%
Updates are practically 0% but we do some stuff with occurrence dictionaries that come up to 8.9%, which leaves about 14% to entropy calculation.
According to perf analysis, 71% of the total time is spent on updating the model. 23% is spent on calculating the entropy. Some <1% is spent on matching and parsing is 1.1%.
In contrast, the stats for order-0 ranking are: 7.1% of the total time is spent on matching, parsing is 65% Updates are practically 0% but we do some stuff with occurrence dictionaries that come up to 8.9%, which leaves about 14% to entropy calculation.