Hi!
I am running in trouble (and into the limits of my cpu ressources) with a scene that contains simply too many light sources. It is a huge room lid by groups of small light (oil lights), all in all there are thousands... Rendering is getting really slow now, and I wonder how to get the most of the scene without loosing to much visual accuracy.
So far, I use a radiance sphere with a light modifier and xform these spheres to all the positions in my room. I guess, as the sphere is a basic and native radiance primitive, it should not make a big difference if I use a sphere, a polygon or any other primitive here? I remember that this question had been on this list some years ago, but I was not able to find any post in my archives. I also believe to remember that there has been someone doing renderings of whole star fields with thousands or millions of sources, am I dreaming or are there some projects out there like that?
When I know how to load the light sources, the next is which settings to use rendering them. They are all extremely small compared to the room. I was also thinking if it might be worth to wrap them into an illum "box", so that for at least the ambient calculation, the groups of "point" (in fact sphere) sources would appear as uniform area light sources.
Thank you for sharing ideas and experiences, at the moment I am not able to get any result in a reasonable time.
CU Lars.
Hi Lars,
If you lights are clustered together, you might be able to put them in an enclosing sphere and use mkillum to determine their collective output distribution, using only as many spherical illum's to light your space. (Spheres are in general faster to use as lights because there is only one sphere per source.)
What is the exact configuration of your oil lamps? Are they all the same? Do you have any close-ups you can post somewhere?
-Greg
···
From: "Lars O. Grobe" <[email protected]>
Date: March 12, 2006 4:11:29 AM PST
Hi!
I am running in trouble (and into the limits of my cpu ressources) with a scene that contains simply too many light sources. It is a huge room lid by groups of small light (oil lights), all in all there are thousands... Rendering is getting really slow now, and I wonder how to get the most of the scene without loosing to much visual accuracy.
So far, I use a radiance sphere with a light modifier and xform these spheres to all the positions in my room. I guess, as the sphere is a basic and native radiance primitive, it should not make a big difference if I use a sphere, a polygon or any other primitive here? I remember that this question had been on this list some years ago, but I was not able to find any post in my archives. I also believe to remember that there has been someone doing renderings of whole star fields with thousands or millions of sources, am I dreaming or are there some projects out there like that?
When I know how to load the light sources, the next is which settings to use rendering them. They are all extremely small compared to the room. I was also thinking if it might be worth to wrap them into an illum "box", so that for at least the ambient calculation, the groups of "point" (in fact sphere) sources would appear as uniform area light sources.
Thank you for sharing ideas and experiences, at the moment I am not able to get any result in a reasonable time.
CU Lars.
Hi Greg!
If you lights are clustered together, you might be able to put them
in an enclosing sphere and use mkillum to determine their collective
output distribution, using only as many spherical illum's to light
your space. (Spheres are in general faster to use as lights because
there is only one sphere per source.)
Good point, so I save 3 faces per cluser.
What is the exact configuration of your oil lamps? Are they all the
same? Do you have any close-ups you can post somewhere?
At the moment, one lamp is represented just as one sphere with modifier
light, and all are the same. Maybe I will add some basic geometry later
(using instances), but as the lamps are so small compared to the room, that
is not urgent.
I do the following at the moment:
1) I have a file conataining one lamp (that is only one modifier and one
sphere description at the moment.
2) I build my clusters by !xform statements in one file per cluster. E.g. I
have conical arrangement (a bit like christmas trees), so the corresponding
file has some !xform-arrays to create circular groups of lights in differemt
heigths and radius. I would put the mkillum-sphere in here, too.
3) The clusters are placed by markers and the replmarks tools.
As I wrote this, I found how stupid it is to have the light modifier
definition IN the lamp file (1), as this means that I have some thousand
definitions of my lamp modifier now (after xforming them to arrays)... all
the same. The first I will change tonight!
I think a close-up of a lght sphere is not interesting? 
But there have
been nice flames for oil lamps modeled before with radiance. I will post an
url tonight to a paper describing it.
CU Lars.
I think a close-up of a lght sphere is not interesting? But there have
been nice flames for oil lamps modeled before with radiance. I will post an
url tonight to a paper describing it.
For those interested in the topic (modeling flame light sources, the
promised url as follows:
http://www.cs.bris.ac.uk/Research/Graphics/archaeology/unesco_paper.pdf
I am not sure if these works are already known here or if the authors
(Devlin, Chalmers, Brown) are even subscribers.
CU Lars.
hi Lars,
Lars O. Grobe wrote:
For those interested in the topic (modeling flame light sources, the
promised url as follows:
http://www.cs.bris.ac.uk/Research/Graphics/archaeology/unesco_paper.pdf
I am not sure if these works are already known here or if the authors
(Devlin, Chalmers, Brown) are even subscribers.
The work that went into that paper was presented at the very first Radiance Workshop in 2002. As I recall, Mr. Chalmers could not attend due to a last-minute conflict, and a very nervous Duncan Brown delivered the lecture instead, and the material was so engrossing that he blew us all away. He played an animation of the oil lamp flame at the end, and then asked if there were any questions. The first question shouted out by an audience member was: "could you play that again?" =8-) Cool stuff.
Can't believe that was 3.5 years ago; man time flies!
- Rob Guglielmetti
www.rumblestrip.org
Actually, I think it was Patrick Ledda who presented this work at the conference. It doesn't appear to be in the notes, unfortunately.
-Greg
···
From: Rob Guglielmetti <[email protected]>
Date: March 15, 2006 12:41:51 PM PST
The work that went into that paper was presented at the very first Radiance Workshop in 2002. As I recall, Mr. Chalmers could not attend due to a last-minute conflict, and a very nervous Duncan Brown delivered the lecture instead, and the material was so engrossing that he blew us all away. He played an animation of the oil lamp flame at the end, and then asked if there were any questions. The first question shouted out by an audience member was: "could you play that again?" =8-) Cool stuff.
Gregory J. Ward wrote:
Actually, I think it was Patrick Ledda who presented this work at the conference. It doesn't appear to be in the notes, unfortunately.
Patrick made a different presentation that year (perception of HDR scenes). It may not have been Duncan Brown who gave the presentation I'm thinking of, but it wasn't Patrick either; it was another Bristol University student. Sorry, I can't remember his name! Cool animation though. Presenter, identify yourself!
(OK, this is a bit on the obsessive-compulsive side, but I just looked at the workshop photos. Our mystery presenter is in http://www.radiance-online.org/radiance-workshop1/radworkshop-images/PA011181.JPG, he's the guy behind and to Patrick Ledda's right.)
- Rob
That would be Peter Longhurst, who is now (incidentally) working for BrightSide Technologies, in Vancouver.
-G
···
From: Rob Guglielmetti <[email protected]>
Date: March 15, 2006 1:19:42 PM PST
Gregory J. Ward wrote:
Actually, I think it was Patrick Ledda who presented this work at the conference. It doesn't appear to be in the notes, unfortunately.
Patrick made a different presentation that year (perception of HDR scenes). It may not have been Duncan Brown who gave the presentation I'm thinking of, but it wasn't Patrick either; it was another Bristol University student. Sorry, I can't remember his name! Cool animation though. Presenter, identify yourself!
(OK, this is a bit on the obsessive-compulsive side, but I just looked at the workshop photos. Our mystery presenter is in http://www.radiance-online.org/radiance-workshop1/radworkshop-images/PA011181.JPG, he's the guy behind and to Patrick Ledda's right.)
- Rob
But there have