cache a sequence per processor

[Omegaroth]

Hello, I am currently caching 240 frames of 2 mil points a frame and my cpus are only using 10% to cache this thing, but it is taking several hours to cache the full frame range.

I was wondering if there are any workarounds or scripts where I can specify each cpu (i have 20 cpus 40 threads) to batch a certain range per cpu (maybe 12 frames a core over 20 cores)

Any tips or resources references would be greatly appreciated!

Thanks all

[Guest tar]

whats the setup? you can't cache a serial sequence any quicker but the Render in Background uses all threads if appropriate. 20 cores means quite a slow GHz usually, so a faster single core speed would help in general.

[Omegaroth]

I brought in 2 million points a frame from a third party fractal generating program. the program writes the point data out as .csv and im reading in color and position with table import

Directly after a few attrib namings I am filecaching as bgeo.sc

Since this process isn't taxing on the cpu's at all, and I'm loading/writing via ssd I assumed I could potentially speed up the process a lot more rather than using the traditional vanilla filecache. 

Each core has 2.8ghz, 3.5 turbo boost (if that ever actually works)

Thanks for your response!

Edited August 13, 2017 by Omegaroth

[Guest tar]

a simple way would be to launch multiple hbatchs from the command line with different frame ranges:

render -f 1-20 /obj/rop_geometry1

render -f 21-40 /obj/rop_geometry1

etc

EDIT: you should also try the -j1 option to limit hbatch to a single core - i.e. hbatch -j1 /pathtohipfile/

Edited August 13, 2017 by tar

[Omegaroth]

Thanks for your response! I will definitely look into command line bactching per core. Maybe it would be a good exercise to practice writing a python script for it.