cache radiosity question

Post your tutorials and help questions in this forum
Locked
geo_n
Posts: 363
Joined: Tue May 20, 2008 3:01 pm
Location: jpn

cache radiosity question

Post by geo_n »

So erwin swart and I have a cache radiosity request but before that I'd like to save some time in caching for now. :mrgreen:
A bit long but hopefully very clear explanation.
When I press render sequence in kray to make an animation cache, timeline goes from 0 to 1, 0 to 2, 0 to 3 and so on until it cycles the end of the timeline. After that it then starts to render from frame 0. Once that 0 frame is done rendering and a tga,png file is autosaved in the hd, a cache is also created in the hd and kray continues to render frame 1,2, and so on.
Now my question is,
That cache created from after rendering frame 0, is that the cache for the whole sequence or only frame 0? Can I stop rendering frame 1,2,etc so that I can render all sequences in a network using that 0 frame cache file?
Right now I wait it out until kray renders all frames with no AA and it still takes time. Atleast 1 min per frame with no AA just to get the cache file at the end of the process. So 600 frames is 600mins. Since there's no way to just calculate cache radiosity I have done that way. Any suggestions? :mrgreen:
jure
Posts: 2142
Joined: Thu Jun 02, 2005 6:53 pm

Re: cache radiosity question

Post by jure »

geo_n wrote:So erwin swart and I have a cache radiosity request but before that I'd like to save some time in caching for now. :mrgreen:
A bit long but hopefully very clear explanation.
When I press render sequence in kray to make an animation cache, timeline goes from 0 to 1, 0 to 2, 0 to 3 and so on until it cycles the end of the timeline. After that it then starts to render from frame 0. Once that 0 frame is done rendering and a tga,png file is autosaved in the hd, a cache is also created in the hd and kray continues to render frame 1,2, and so on.
Now my question is,
That cache created from after rendering frame 0, is that the cache for the whole sequence or only frame 0? Can I stop rendering frame 1,2,etc so that I can render all sequences in a network using that 0 frame cache file?
Right now I wait it out until kray renders all frames with no AA and it still takes time. Atleast 1 min per frame with no AA just to get the cache file at the end of the process. So 600 frames is 600mins. Since there's no way to just calculate cache radiosity I have done that way. Any suggestions? :mrgreen:
This is actually something i wanted to try myself but haven't gotten around to actually do it yet.
Now the first pas you were describing is when Kray shoots photons along camera path (this only happens in lightmap mode). Second pass is actual FG rendering pass. You don't need to render every frame for this. It's enough you render ever 5th or 10th frame depending on camera speed. You just need to get FG samples everywhere.

Now what I and Erwin wanted to try is weather the animation would look without flickering even if we skip second (FG) pass from prerendering. So you would just prerender photon map and then start the final animation immediately. However there are potential problems with this method. Because FG will not be precalculated it means that each node will calculate each own FG solution. So we need to check if that would flicker or not. The other question is do we actually gain any time in the end...

Sorry I don't have any exact answers for this yet but it shouldn't take you much time to test it out though. And if you do I'd be interested to hear your results...
- Jure
geo_n
Posts: 363
Joined: Tue May 20, 2008 3:01 pm
Location: jpn

Re: cache radiosity question

Post by geo_n »

So the first pass is the timeline moving and there's no rendering? Have to clarify before I try this.
When kray window actually opens to render frame 0 is second pass or FG pass for cache. Correct?
So in theory I could let the first pass run till the end of the timeline then when it starts to render frame 0 I stop it and switch to render step frame 5-10,etc depending on camera speed. But what setting should the cache be for that pass? load, save, update? I'm guessing update? Then when that second pass finishes I setup AA and all and use load cache settings to start final quality render.
It will save time if there was only calculate cache with no renderin. A minute is not so big but if you count it as 300,600,1500 frames, that minute per frame counts as opposed to calculate cache only that takes seconds per frame. :mrgreen:
jure
Posts: 2142
Joined: Thu Jun 02, 2005 6:53 pm

Re: cache radiosity question

Post by jure »

geo_n wrote:So the first pass is the timeline moving and there's no rendering? Have to clarify before I try this.
When kray window actually opens to render frame 0 is second pass or FG pass for cache. Correct?
So in theory I could let the first pass run till the end of the timeline then when it starts to render frame 0 I stop it and switch to render step frame 5-10,etc depending on camera speed. But what setting should the cache be for that pass? load, save, update? I'm guessing update? Then when that second pass finishes I setup AA and all and use load cache settings to start final quality render.
It will save time if there was only calculate cache with no renderin. A minute is not so big but if you count it as 300,600,1500 frames, that minute per frame counts as opposed to calculate cache only that takes seconds per frame. :mrgreen:

eem not exactly what I meant but yes you could that too (and yes update mode is the one you use). You don't need to cancel render and set every 5th frame or so though. You can do that from the begining.
But what I was going at in my post is to skip the "prerendering" (or saving to file) the FG pass completely.

About saving time for creating cache without rendering. I'm not 100% sure this will sure whole lot of time... Kray still needs to render the scene to cache the FG sample points and displaying them on the screen as an extra that isn't really taking much time anyways. In any case this will be improved in future for more solid performance and ease of use.
- Jure
geo_n
Posts: 363
Joined: Tue May 20, 2008 3:01 pm
Location: jpn

Re: cache radiosity question

Post by geo_n »

How do I skip prerendering?
I did try the method in my post and stopped kray after the timeline 0-0,0-1,0-2,etc process then switched to step frame for FG second pass. It works. No flicker. But if I can do it as you said from the beginning and use step frame that would be faster.
I'm sure this is case by case though just like any cache method. Fast moving and deforming objects need frame step 1. :mrgreen:
jure
Posts: 2142
Joined: Thu Jun 02, 2005 6:53 pm

Re: cache radiosity question

Post by jure »

geo_n wrote:How do I skip prerendering?
I did try the method in my post and stopped kray after the timeline 0-0,0-1,0-2,etc process then switched to step frame for FG second pass. It works. No flicker. But if I can do it as you said from the beginning and use step frame that would be faster.
I'm sure this is case by case though just like any cache method. Fast moving and deforming objects need frame step 1. :mrgreen:
Ok let me try to describe this more clearly once again.

The 'usual' method is like this:

1st pass: set framestep to every Nth frame, select Photon mapping+cached as diffuse method and "UPDATE". Set AA off, min/max reflection blur rays 1/1, lights quality 1/1 - so the render will go faster (Note: FG passes must be 1).
2nd pass: set framestep back to 1, set original AA and reflection blur/lights setting and render away the whole animation.

Second method (untested):
1st pass: framestep 1, shared file mode as UPDATE, and render away. Kray will scrub through the timeline shooting photons and then start rendering 1st frame. Wait for it to finish and then cancel render. Photon map is now created in GI file.
2nd pass: just start rendering on all your nodes from the created GI file (still in UPDATE mode).
- Jure
geo_n
Posts: 363
Joined: Tue May 20, 2008 3:01 pm
Location: jpn

Re: cache radiosity question

Post by geo_n »

Ok the first method is normal method. :mrgreen:

The second method is the one I was looking for thanks. Because my original question was after kray does the timeline thing and render 0 frame it creates a cachefile and I was asking if that alone can be used so that I can render on the network immediately instead of waiting for kray to finish the FG second pass for all frames in one computer.

"That cache created from after rendering frame 0, is that the cache for the whole sequence or only frame 0? Can I stop rendering frame 1,2,etc so that I can render all sequences in a network using that 0 frame cache file?"

I will try the cache file and see if it flicker on the second method. I hope it works. :mrgreen:
jure
Posts: 2142
Joined: Thu Jun 02, 2005 6:53 pm

Re: cache radiosity question

Post by jure »

geo_n wrote:Ok the first method is normal method. :mrgreen:

The second method is the one I was looking for thanks. Because my original question was after kray does the timeline thing and render 0 frame it creates a cachefile and I was asking if that alone can be used so that I can render on the network immediately instead of waiting for kray to finish the FG second pass for all frames in one computer.

"That cache created from after rendering frame 0, is that the cache for the whole sequence or only frame 0? Can I stop rendering frame 1,2,etc so that I can render all sequences in a network using that 0 frame cache file?"

I will try the cache file and see if it flicker on the second method. I hope it works. :mrgreen:
yes. report back what you find out...
- Jure
geo_n
Posts: 363
Joined: Tue May 20, 2008 3:01 pm
Location: jpn

Re: cache radiosity question

Post by geo_n »

This works jure. Even on deforming objects. :P
Not sure on interior or enclosed space though. WIll try it when I have time to do interior scene.
kamesan
Posts: 2
Joined: Wed May 30, 2012 1:05 pm

Re: cache radiosity question

Post by kamesan »

Hi,
Is this still works with Kray2.5 and is it a good tip anymore?
So, does it accelerate the whole rendertime?
Do you have success with interior and enclosed space?

One more question, i read that it's not necessary to generate the full-quality GI solution on just one machinebut a 10 % resolution because this will greatly speed up the generation of the GI cache, and once you have a 'baseline' to work from, the nodes will generate results that are much more similar to one and other.

Can you confirm this?
i will love to see a bit more explanation in the wiki about netrender with this tips. Maybe video too...

Regards,
Locked