Cull particles by camera?

[locvio]

Any maya solutions for deleting off-camera particles (have them added to memory
once they are in view of camera and removed once they
are out of view)? (they dont have to be killed as long
as they don't contribute to memory footprint)

thanks!

[joojaa]

In what context? Rendering, simulation, instancing. What memory? Is disk and gpu memory considered memory footprint?

See, many of the actual operations allready do cull the particles. For example maya viewports cull instanced particles. This does not mean they do not have a memory footprint they do (the vector for particle info is still stored) however a particle that is removed from memory footrint completely WILL cease to exist. That means it can not come back, because the comuter can not update things that do not exist. For computer to be able to do thsi it has to exist in memory somewhere. Also instanced items do not generate much footprint unless you raytrace, but then you can not cull them.

However its prefectly feasible to elimintate the particles before rendering, or during disk caching.  Thing is tough mostly this is not really something you should worry about since rendering systems allready cull if it makes sense to do so,

[locvio]

thanks!!

1)  V represents camera frustum

.........................................................................
    V   --> 

2)

  ......
    V   --> 

1) above shows camera pan through rings of saturn (rocks). in direction of
arrow.. camera is quite far from the rings, so trying to optimize
to only have as many rocks as necessary for camera..

yeah, in houdini would write the points out to disk and cull by
frustum to get 2) .. for any frame..  

Do you mean that there wouldn't be much overhead if Im using
maya instancing and not culling -- 1) vs if I do cull  -- 2)

Thinking about it,  a better approach might be to try create
particles just in front of moving camera, and delete them
when they have passed..   This would save simulation/caching/render time?

ie wouldnt expect viewport/render optimizations to do equivalent job?

[joojaa]

Are you sure your not prematurely optimizing? I mean do you KNOW this in fact will be a problem or do you just think you know. The only way you could know is if you already did this shot and had problems and know your problem is unrelated to other things.

Do you mean that there wouldn't be much overhead if Im using
maya instancing and not culling -- 1) vs if I do cull

Depends on the situation. The graphics card will not be drawing the instances outside of you viewport (At least using viewport 2.0 you can sometimes see this with primitives that are very large) The particle instances do also do not use up memory they are after all reading the same primitive in memory over and over. The software renderer might or might not cull them depending on circumstances renderer and implementation, and situation.

The particles themselves use very little memory, processing them on the other hand is another thing. But basically easch particle uses up about 6-48 bytes of memory. In houdini you usually do suff that maks the memory consumption go over the roof (because you expand the instances) in maya you really can not go there. That means that having a million particles, the particles themselves use about 48 megabytes of memory when simulating. This is not a lot of memory, but can possibly be a problem for viewports. This can off course expand when exporting the stuff to a software renderer if it doesn't use instancing sanely. Simulating that list is another thing entirely especially if your program is slow, but then removing items from simulation might break the simulation entirely.

Thinking about it,  a better approach might be to try create
particles just in front of moving camera, and delete them
when they have passed..   This would save simulation/caching/render time?

It might, or it might not. Anyway you need to account in the time to develop the tools and making sure it performs better then what you already have. So at the end of the day considering the effort you put into the simulation may outweigh the gain, especially if you can not spread this time over many productions.

In a way you already lost 24 hours of rendering time discussing this with me

[locvio]

thanks!

 drawing instances  made tweaking very hard, if displayed as bounding boxes performance was much
better..  still..

I agree about premature optimization, but I think the setup I initally had (having a huge field of rocks when
only a small section was visible on any given frame) was just hugely inefficient for this case.. As you say all those extra particles
were still getting processed.

Wound up having an emitter animated just outside of camera, and had them die off using a reasonable
lifespan, once they are no longer in-camera.

in houdini you usually do suff that maks the memory consumption go over the roof (because you expand the instances)
curious what you are referring to here? the way copy sop handles instancing, perchance?

In a way you already lost 24 hours of rendering time discussing this with me
lol.. fair enough

[joojaa]

I agree about premature optimization, but I think the setup I initally had (having a huge field of rocks
when only a small section was visible on any given frame) was just hugely inefficient for this case..

Yes this is something that pops up a lot. Experts, in general, spend s lot of time thinking on things that don't necessarily affect so much on the outcome. I'm not saying in never happens but quite often it does not. Sometimes you err on side of caution sometimes that's just stupid.

However our brain will nearly always justify even the most superfluous thought. So being honest about the need for something is really hard or next to impossible, because we carry a reality distortion field with us. (yeah, mine is gigantic. still nothing compared to some of the biggest agitators in history)

the way copy sop handles instancing, perchance?

Well the thing is in Houdini they can (can does not mean they have to) be copies in Maya not really. So if you do copies in Maya your very, very deliberately shooting yourself in the foot in Houdini you might not be. I'm not saying Houdini users in general do this. Just that quite many users work in magic land. They follow procedures and processes that have served them right. But that's not the same thing as doing right. Again our reality distortion at play.

I teach at a technical university, and quite often people who don't know anything seem to think they do. Knowing your limitations is the 3 rd stage of learning. People with less than 5-10 years of experience in a subject often fall into stage 2 (delusions of grandeur).

Its only after you realize your limitations that you get to a stage where you can objectively look at your own work. The alternative is to become a pessimist.