Yes, I have not yet optimized this to minimize memory usage (I've been
focusing mostly on speed).  Each photon currently has two vectors (for 
position and direction), and each of those vectors has three floats.
Then add color, some pointers for the kd-tree struct, and you've got
a fairly large memory structure.  I'm hoping to re-code some of the
kd-tree so I don't need pointers, and cram the direction vector into
two characters (phi & theta angles... but that might slow things down
because I'll have to decode those into another vector using sin & cos).

-Nathan

Bob Hughes wrote:
> 
> Wow... and wow, look at the memory used. I would have expected a measely
> 50K not 5M! Those photons aren't your everyday variety of quanta
> particles, are they?
> Render time looks good though. Seeing caustics as they ought to be is a
> ray of sunshine in a (media) cloudy world.
>

Post a reply to this message