Multi pass Light Cache?

[ngrava]

Hey Vlado and all others,

I've been experimenting with DR today. I noticed that there is a new multi pass option for the light cache. Does this depend on how many machines/CPUs you have? If I'm using 5 slaves with 2 HT CPUs each (four threads total), should I set this to use 20 passes? Or is this a setting per node? As in, each should use 4 threads?

Thanks,

-=GB=-

[percydaman]

this is a setting per computer. dual xeon with HT will have a value of 4, one for each bucket.

[ngrava]

Ahh... So that means I should leave it at 4? Actually, the only way I've got it to work is setting it at 20 for all the threads.

-=GB=-

[percydaman]

good point. i guess it would be once again per bucket, so if your using DR, it would be total # of buckets.

[plastic_]

what are you talking about? lightmap works with DR now?

[ngrava]

Well, that's what it says in the release notes. As far as whether or not it's working, I have no idea. It doesn't seem like anything different is happening.

[plastic_]

* Added a parameter Number of passes for the light cache; this allows to achieve the same light cache on single- and multi-processor machines while keeping the advantage of multithreading.

if you talk about this note...
i thought the reason for that was to ensure that the lightmap looks the same way on single and multiprocessor machines. previously this wasn't the case.

[vlado]

The number of passes should be equal to the maximum number of threads that you expect to work on the light cache. The default value of 4 is good enough, since you can have at most 4 threads on a dual HT machine.

Since the light cache is not distributed at this time, there is no point in changing this setting for a DR rendering.

Best regards,
Vlado