[Bf-committers] Does Cycles waste half it's possible performance or am i wrong?

Tobias Oelgarte tobias.oelgarte at googlemail.com
Sun Oct 28 01:49:38 CEST 2012


Hi,

Sorry that I used the wrong list. But I hope you have nothing against 
continuing this discussion on this list.

I don't know every detail of the implementation. But i assume that you 
use something like spreading rays (fitted to the camera and film) for 
every pixel on a fixed grid, where the gauss function defines the fine 
spread of the rays from every pixel center. I understand that this is 
most likely faster as to start rays from anywhere or to let them have 
different angles.

But if you have samples with errors, then this error will only 
contribute to one pixel. Especially caustics will take a long time to 
get clean if they are a seldom occurrence, because a neighbouring pixel 
might just miss nearly any time. If you would send in rays which "start 
at subpixels" of the film, then it is much more likely to get a better 
average (lower error), because most samples contribute to multiple 
pixels. A firefly found at the film-grid-intersection would evenly 
affect at least four pixels, giving that ray four times the probability 
to contribute the same result to all those pixels. Rays close to the 
center of a pixel will loose that effect. That way it would be about 2 
times less noise in average.

So i would assume that this sampling method would give better results 
(in same rendering time) as long you don't double the overall 
computation time for every sample. Especially more complex scenes (lot 
of bounces, comparably less important ray creation and weighting on the 
film) should benefit from it.

Greetings from
Tobias Oelgarte

Am 27.10.2012 23:45, schrieb Brecht Van Lommel:
> Hi,
>
> This is a topic to discuss on the bf-cycles mailing list. We use a
> different pixel filtering method which fits better on the GPU. If you
> set the filter width to 10 then yes, you might be wasting half the
> samples. But for a typical filter width the number of wasted samples
> would be quite small in my opinion. Still it would be good to improve
> this, but I'm not sure where the half the samples number comes from.
>
> Brecht.
>
> On Sat, Oct 27, 2012 at 11:16 PM, Tobias Oelgarte
> <tobias.oelgarte at googlemail.com> wrote:
>> Hello,
>>
>> I'm curios why i never see fireflies in cycles that affect more then one
>> pixel. I have chosen to speak about fireflies, because they are the most
>> unique sample that one could observe. It doesn't matter if set the
>> gaussian filter to 1.5 or 10. The fireflies, no matter how strong they
>> are, only cover one pixel. I see no distribution to other pixels of the
>> film. If I'm not mistaken then i should see this sample contributing to
>> at least four pixels (except the sample is perfectly centred and
>> gaussian is 1.0).
>>
>> So I'm wondering why this samples do not contribute to neighbour pixels.
>> If they would, then it would expect a significant noise reduction.
>> (about half the samples for same result)
>>
>> Greetings from
>> Tobias Oelgarte
>> _______________________________________________
>> Bf-committers mailing list
>> Bf-committers at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-committers
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> http://lists.blender.org/mailman/listinfo/bf-committers
>



More information about the Bf-committers mailing list