Query on histogram algorithm to calculate loudness

[Rohit-Khali]

Hi,

I am facing some issues with resources (cpu & memory) while computing program loudness (EBUR128_MODE_I) for long duration tests. Issue is on the expected lines since there is a need to maintain the history (Default is ULONG_MAX (at least ~50 days)) in order to get more accurate results.
While checking for the alternate solutions within library, I came across 'EBUR128_MODE_HISTOGRAM' in "ebur128.h". As per the comments and the code lookup looks like this mode is defined to use histogram algorithm to compute program loudness and LRA in a resource friendly manner. I also did some benchmarking to compare the results with and without 'EBUR128_MODE_HISTOGRAM' and results look similar with differences in decimal place which is encouraging.
Just wanted to confirm once whether above observations are correct including the term resource friendly in context of 'EBUR128_MODE_HISTOGRAM' and if there are any known limitations?

Thanks,
Rohit Khali