vTune Profile

[ahorek]

here's a profile from https://software.intel.com/content/www/us/en/develop/tools/vtune-profiler.html

measured on an intel platform (avx2 build, gcc 9.2.0)

hopefully, it might be useful for someone to investigate where the bottlenecks are

paq8px.exe -8 data.dat // default CM
overview

paq8px.exe -8l data.dat // default CM
overview (ltsm)

paq8px.exe -8 data.dat // default CM
tree

paq8px.exe -8 data.dat // default CM
tree (ltsm)

[moisespr123]

Hi,

Please suggest this at the official paq8px thread here: https://encode.su/threads/342-paq8px/page70

[micsthepick]

I looked at the callgrind graphs for some simple files, and a lot of time is spent on APM classes initialization, and when I added some custom file caching code to prevent recalculating each set of coefficients for the given APM sizes, it significantly sped up