Am 29.01.2014 19:22, schrieb [GDS|Entropy]:
> clipka <ano### [at] anonymousorg> wrote:
>> A scene with 1000 randomly placed light sources with distance-dependent
>> attenuation (fade_power, fade_distance).
>>
>> First image: POV-Ray 3.7, render time 73s.
>>
>> Second image: experimental patch, render time 9s.
>>
>> For each light source, the patch automatically computes a cutoff
>> distance based on light source brightness, fade parameters, and a global
>> threshold (currently this is the global adc_bailout setting). Beyond
>> this distance, the lights source is simply ignored.
>>
>> (As an alternative, the patch also allows for manually specifying the
>> cutoff distance, but that's not used in this scene.)
>
> Hey, that looks pretty nice there; it reminds me of the MegaPov Glows a bit. ;)
>
> Ian

No, just a bunch of ordinary lights with matching looks_like emissive 
spheres, placed neatly on the floor. Place them somewhere in mid-air and 
most of the visual magic is gone.

Post a reply to this message

Very nice, thank you! It sounds strangely familiar ;)

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> A scene with 1000 randomly placed light sources with distance-dependent
> attenuation (fade_power, fade_distance).
>
> First image: POV-Ray 3.7, render time 73s.
>
> Second image: experimental patch, render time 9s.
>
> For each light source, the patch automatically computes a cutoff
> distance based on light source brightness, fade parameters, and a global
> threshold (currently this is the global adc_bailout setting). Beyond
> this distance, the lights source is simply ignored.
>
> (As an alternative, the patch also allows for manually specifying the
> cutoff distance, but that's not used in this scene.)

Cool work, Christoph!

There's something odd going on between the two images you posted: the spheres in
the UberPOV version are noticeably darker, but the plane is not. Did you change
the spheres' material settings between renders?

Sam

Post a reply to this message

Am 30.01.2014 20:24, schrieb Samuel Benge:

> There's something odd going on between the two images you posted: the spheres in
> the UberPOV version are noticeably darker, but the plane is not. Did you change
> the spheres' material settings between renders?

You mean the diffuse spheres, right?

No, that's not odd at all:

- The darkening of the spheres of course comes from the fact that even 
though the cut-off light sources don't contribute much brightness 
individually, they still do contribute a good deal in total.

- The (apparent) non-darkening of the plane comes from the fact that 
here the contribution of the individual light sources isn't just 
affected by the inverse-square distance-based falloff, but also the 
falloff due to the angle of incidence.

If you'd use a "brilliance 0" finish on the plane (causing diffuse 
brightness to be entirely independent of the angle of incidence), the 
darkening effect would be just as visible on the plane as it is on the 
spheres.

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> A scene with 1000 randomly placed light sources with distance-dependent
> attenuation (fade_power, fade_distance).
>
> First image: POV-Ray 3.7, render time 73s.
>
> Second image: experimental patch, render time 9s.
>
> For each light source, the patch automatically computes a cutoff
> distance based on light source brightness, fade parameters, and a global
> threshold (currently this is the global adc_bailout setting). Beyond
> this distance, the lights source is simply ignored.
>
> (As an alternative, the patch also allows for manually specifying the
> cutoff distance, but that's not used in this scene.)

How does the patch deal with radiosity? does it still provide an improvement?
any order of magnitude for it?

Post a reply to this message

Am 01.02.2014 10:19, schrieb Mr:
> clipka <ano### [at] anonymousorg> wrote:
>> A scene with 1000 randomly placed light sources with distance-dependent
>> attenuation (fade_power, fade_distance).
>>
>> First image: POV-Ray 3.7, render time 73s.
>>
>> Second image: experimental patch, render time 9s.
>>
>> For each light source, the patch automatically computes a cutoff
>> distance based on light source brightness, fade parameters, and a global
>> threshold (currently this is the global adc_bailout setting). Beyond
>> this distance, the lights source is simply ignored.
>>
>> (As an alternative, the patch also allows for manually specifying the
>> cutoff distance, but that's not used in this scene.)
>
> How does the patch deal with radiosity? does it still provide an improvement?
> any order of magnitude for it?

Radiosity scenes will probably pretrace faster, but speedup of the main 
render may not be as noticeable, as in that phase radiosity is primarily 
about looking up already-computed samples (which is independent of the 
number of light sources in the scene), and tends to significantly 
outweigh classic lighting calculations in terms of processing power demands.

But it will actually depend heavily on the scene you are rendering, and 
it's difficult to make any estimations.

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> Am 30.01.2014 20:24, schrieb Samuel Benge:
>
> > There's something odd going on between the two images you posted: the spheres in
> > the UberPOV version are noticeably darker, but the plane is not. Did you change
> > the spheres' material settings between renders?
>
> You mean the diffuse spheres, right?
>
> No, that's not odd at all:
>
> - The darkening of the spheres of course comes from the fact that even
> though the cut-off light sources don't contribute much brightness
> individually, they still do contribute a good deal in total.
>
> - The (apparent) non-darkening of the plane comes from the fact that
> here the contribution of the individual light sources isn't just
> affected by the inverse-square distance-based falloff, but also the
> falloff due to the angle of incidence.
>
> If you'd use a "brilliance 0" finish on the plane (causing diffuse
> brightness to be entirely independent of the angle of incidence), the
> darkening effect would be just as visible on the plane as it is on the
> spheres.

Ah, OK. Are there any visible artifacts when using this patch?

Have you conducted a neon sign test yet? :D

Sam

Post a reply to this message

Am 04.02.2014 22:22, schrieb Samuel Benge:

> Ah, OK. Are there any visible artifacts when using this patch?

Yes: If you set the threshold too high, you can clearly see where the 
light is abruptly cut off.

Post a reply to this message

Am 29.01.2014 03:19, schrieb clipka:
> A scene with 1000 randomly placed light sources with distance-dependent
> attenuation (fade_power, fade_distance).
>
> First image: POV-Ray 3.7, render time 73s.
>
> Second image: experimental patch, render time 9s.
>
> For each light source, the patch automatically computes a cutoff
> distance based on light source brightness, fade parameters, and a global
> threshold (currently this is the global adc_bailout setting). Beyond
> this distance, the lights source is simply ignored.
>
> (As an alternative, the patch also allows for manually specifying the
> cutoff distance, but that's not used in this scene.)

The patch is now available as part of UberPOV v1.37.0.0-beta.5:

https://github.com/UberPOV/UberPOV/releases/tag/v1.37.0.0-beta.5

Post a reply to this message

On 28-5-2014 22:23, clipka wrote:
> The patch is now available as part of UberPOV v1.37.0.0-beta.5:
>
> https://github.com/UberPOV/UberPOV/releases/tag/v1.37.0.0-beta.5
>

Thank you sir.

Thomas

Post a reply to this message