Point in time view based not running on Parallel

Hey @mostapha @mikkel ,

I’m trying to run a daylight point-in-time view analysis with LB 1.8, but it’s not running in parallel. The “view_count” input for parallelization seems to be removed in the newer versions (though it’s still mentioned in the description—maybe it should be updated?). I noticed there’s a default “cpu_count : 12” in the recipe inputs source code, but it still runs on a single core for me.

Any advice on how to fix this?

Thanks!

Hi @HamidrezaSh,

Use HB Recipe Settings to change the number of workers. This determines the number of smaller images your view will be split into. However, it is already set to one less than the number of CPUs currently available on the machine, so the only way not to run in parallel is if your machine only has two CPUs.

If you see more than one “RayTracingLoop”, then it is running in parallel.

1 Like

Hi @mikkel

Thanks for you reply. My machine has 32 cores and I was specifying 12 workers with Recipe Setting for my simulation.
After checking the Log file indeed the RayTracingLoop happened in later stages of the simulation, however what made me think that it was running on single core was the “SplitView” command.
It took almost 30 minutes for this command to run on my machine.
I was doing the simulation for a resultion of 1000 pixels and my radiance parameters were:
-aa 0.2 -ab 5 -ad 2048 -ar 128 -as 2048 -dc 0.5 -dj 1.0 -dp 256 -dr 1 -ds 0.05 -dt 0.15 -lr 6 -lw 0.01 -pj 0.9 -ps 2 -pt 0.05 -ss 1.0 -st 0.25

Hi @HamidrezaSh,

“SplitView” will run an overture calculation (slide 70-80) on a single core, however, when we split the view and run the ray tracing for the smaller pieces of the view it will run faster since they will use an ambient file generated in the overture calculation.

I have never experienced that it takes 30 minutes with those parameters that you listed. You can turn the overture calculation off by setting skip_overture_ to FalseTrue, but I would like to figure out why it takes 30 minutes if you don’t mind sharing you model here or in a private message.

1 Like

Thank you @mikkel,

I tested the skip_overture_ set to false and I did not see much improvement in splitView time reducing.

Hi @HamidrezaSh,

My bad! I meant to say True. The default behaviour is False so that’s why you don’t see any difference.

1 Like

Nice, Thank you. With Set to True it works much faster, splitview only took a minute now!!