Hi all :)
I spend the hole weekend squeezing a bit more optimizations for the
surface reconstruction from particles, I have implemented a marching step
only to the surrounding voxels that contains the surface (and not to
every surrounding voxels as before) also I have implemented corner value
reusing (every new step reuse the half of the previous calculated voxel
corner values) and finally I have rewritten my previous spatial hash code
that use integer operations to reuse the metaballs bitwise hash functions
.... all of this optimizations are pushing the surface generation to its
limits, further speed up I think will be gained only with multithreading
and or OpenCL but that is out of my current scope/skills.
As a comparison note, this implementation has outperformed the metaball
instanced particles by a huge margin.
I probably will spend this week cleaning the code for make a separation
between fluid particle engine and visualization engine and will send
Jahka the first patch for review.
Once integrated the fluid particles engine the efforts could be focused
on renderable meshings :)
Cheers Farsthary
_______________________________________________
Bf-committers mailing list
Bf-committers@blender.org
http://lists.blender.org/mailman/listinfo/bf-committers
