Brute force and bucket size

[fAEkE]

I've read about brute force depending on bucket size a couple of times, but never had it explained to me. It says nothing about bucket size in the online manual nor the complete guide. So, what does bucket size have to do with brute force, and is it any difference when using the method as primary or secondary bounce?

[vlado]

There isn't a specific relationship between the bucket size and brute force GI. Like it says in the help index, using smaller buckets might increase the render time a bit because of the additional border around each bucket that needs to be recomputed for proper filtering. On the other hand, smaller buckets help to distribute the calculations more efficiently among many cores or machines.

Best regards,
Vlado

[fAEkE]

Ah, thanks

[InTercepto]

Hey Vlado, one question related to this... how is filtering done in DMC sampler, since you are baking every new sample over preexisting one for pixel? I know Adaptive Subdivision sampler keeps every single image sample in memory and does filtering after rendering but I suppose for DMC you are doing that on per pixel basis?

[vlado]

With the DMC sampler, once an image sample is computed, its effect on the current and neighboring pixels is immediately updated. After this is done, it is no longer necessary to keep the sample.

Best regards,
Vlado

[InTercepto]

Got it. Thanks for the info.