[Bf-cycles] ray depth

[David Fenner]

I just did the same tests in arnold, and although the render time decreases
noticeable when deactivating bounces per lights, is not comparable to
limiting the global bounces, even when the result is the same. For example,
in arnold a simple scene with

1 bounce globally and two lights:     11 secs,
0 bounce globally:                             3 secs,
0 bounce per light but 1 globally (visually the same as above):    7 secs.
0 bounce light A, 1 bounce light B, 1 bounce globally:    9,2 secs.

It seems that making the lights not bounce even when GI is globally enabled
is optimized and gives quite a nice speedup, which would be very nice in
cycles. Right now the speed gain of deactivating GI in lights even when it
is enabled globally is almost unnoticeable. I'm talking of simple scene
with plain diffuse shader, and only taking diffuse bounces into account.




2014-10-06 15:57 GMT-04:00 David Fenner <d4vidfenner at gmail.com>:

> Well that was the question, if it was possible to skip bounces of some
> lights even when other lights have GI. So it isn't possible? if for example
> a scene with 10 lights all with bounces renders in 5 minutes, and the same
> but globally no bounces renders in 2 minutes, then what you are saying is
> that it is inevitable that if only 1 of the 10 lights have bounces the
> scene will always render in the same 5 minutes if GI is used globally (Even
> when 9 of 10 lights don't bounce)?? That is the way it is now, I was hoping
> that even when GI is globally activated, as long as we started reducing
> bounces on the lights that don't need them render times would decrease.
>
>
>
>
> 2014-10-06 15:47 GMT-04:00 Brecht Van Lommel <brechtvanlommel at pandora.be>:
>
> I understand the use case, GI with some auxiliary lights that do not
>> use GI. But the tests being done here are not representative of that,
>> the main speedup you are seeing in that test is because you're
>> skipping bounces that you can't skip if you need GI for other light
>> sources.
>>
>> On Mon, Oct 6, 2014 at 8:50 PM, David Fenner <d4vidfenner at gmail.com>
>> wrote:
>> > Hmm I don't understand what you mean brecht... I never said anything
>> about
>> > wanting no global bounces at all, in the first mail I explained that
>> usually
>> > some lights (like environment) need GI but many auxiliary lights used to
>> > "paint" polygons don't. For example, think of a scene that has 2 lights
>> with
>> > GI, and 8 lights (rims, face aux, eye aux, some torch, whatever) don't
>> need
>> > GI, and in fact GI works worse visually. In this scene there would be no
>> > gain at all by making these 8 lights not bounce, but since they don't
>> > bounce, can't these rays not be calculated to make it faster?
>> >
>> > Please check first mail, the idea is to make it easier to make some
>> lights
>> > not bounce for artistic control and that this also helps to optimize
>> speed,
>> > since they don't bounce... I don't even know if its possible, but I do
>> > believe that a scene with 8 lights bouncing should render slower than
>> one
>> > with 1 light bouncing (of 8), or this is not possible with pathtracers??
>> >
>> >
>> > 2014-10-06 13:49 GMT-04:00 Brecht Van Lommel <
>> brechtvanlommel at pandora.be>:
>> >
>> >> Well yes, it would be nice if Cycles could automatically detect when it
>> >> can set global bounces to 0, but you're not gaining any new
>> capabilities
>> >> this way. I imagine that if you want no light bounces at all, then you
>> would
>> >> have probably just set global bounces to 0, rather than going through
>> the
>> >> trouble of changing all lamps, emissive surfaces, volumes, ambient
>> >> occlusion, the world shader, and whatever else in the scene might be
>> >> emitting light.
>> >>
>> >> There are various optimizations possible here, but this specific case
>> >> doesn't seem all that helpful for a production scene?
>> >>
>> >> On Mon, Oct 6, 2014 at 1:52 PM, Greg Zaal <gregzzmail at gmail.com>
>> wrote:
>> >> > Done some tests too and David is right.
>> >> >
>> >> > 128 bounces, 5:14 - http://i.imgur.com/XFdhp9l.png
>> >> > 128 bounces with Ray Depth trick, 5:11 -
>> http://i.imgur.com/p8kL5ki.png
>> >> > 0 bounces, 4:10 - http://i.imgur.com/xQdrSQD.png
>> >> >
>> >> > Nodes used for one of the lamps: http://i.imgur.com/MBsi7hh.png
>> (also
>> >> > tried
>> >> > http://i.imgur.com/htdCsxi.png - made no difference)
>> >> > Blend file: http://pasteall.org/blend/31928
>> >> >
>> >> > Cheers,
>> >> > Greg
>> >> >
>> >> > _______________________________________________
>> >> > Bf-cycles mailing list
>> >> > Bf-cycles at blender.org
>> >> > http://lists.blender.org/mailman/listinfo/bf-cycles
>> >> >
>> >>
>> >>
>> >> _______________________________________________
>> >> Bf-cycles mailing list
>> >> Bf-cycles at blender.org
>> >> http://lists.blender.org/mailman/listinfo/bf-cycles
>> >>
>> >
>> >
>> > _______________________________________________
>> > Bf-cycles mailing list
>> > Bf-cycles at blender.org
>> > http://lists.blender.org/mailman/listinfo/bf-cycles
>> >
>> _______________________________________________
>> Bf-cycles mailing list
>> Bf-cycles at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-cycles/attachments/20141006/ad59a262/attachment.htm