small objects, exclude for lightcache, wrong??

[plastic_]

hi,

i'm doing a static flythrough, and i noticed that there is a problem with my workflow.

i used to turn off furniture and small objects, while calculating the lightcache, and then activating them for irr-map calculating and rendering.
but the irr-map render time goes up the roof, when i'm doing so.

i was under the impression, that removing small objects from the secondary bounces was a good idea, because that's how it worked with radiosity/photon mapping.
small objects are irrelevant for the secondary bounces and shouldn't be considered, to optimize speed.

so am i wrong?

i don't quite understand why the irr-map calculation becomes dead-slow when i do so...

should i always calc the lightcache with all objects enabled?

[joconnell]

The Irradiance map is to an extent quided by whatever secondary bounce method you use so if you've got objects hidden for the secondary bounces, your light cache file may be pointing the irradiance map in the wrong direction which will lead to render errors. It seems that if theres an error in an irradiance map such as people making changes to a scene and trying to reuse an old irmap calculation times seem to jump quite a bit. The LCmap is pretty quick to go through anyway so you might be causing yourself a lot more problems by trying to save a bit of time calculating the secondaries.

[studioDIM]

The LCmap is pretty quick to go through anyway so you might be causing yourself a lot more problems by trying to save a bit of time calculating the secondaries.

Second that.
if you keep the sample size big enough, it shouldn't have a problem with small object.
If it does, raise the number of averaged samples (if the renderitmes grow too much, prefilter it): you'll lose some fine detail, but gai a lot in smoothness (i tried up to 500 samples in some tests with good results).

Lele

[DaForce]

yeah there is no need to hide things.

And thats whats causing your slow IR map.
Also the LC data is actually saved in the IR map as well.