*the options "remove faces", "merge" points were removed
*made shrinkwrap work with CVs (curves and nurbs surfaces)
*cleanup shrinkwrap code.. (removed bruteforces and raytree methods)
*Made nearest surface also use "quad" bvh tree (instead of splitting quads in 2 bvh nodes).
Again that leaded to improvements in build and query time.
*BLI_bvhtree_find_nearest api is now following the same concept as BLI_bvhtree_ray_cast
removed code relative to bvhtree_from_mesh_tris.
Somehow it didnt finished and didnt added some files under the svn control
(found thanks to lguillaume that reported some files were missing)
Last merge fixed and also merged modifications up to revision 15584.
I checked the diff
svn diff --new . --old https://svn.blender.org/svnroot/bf-blender/trunk/blender
And everything seems to be right now
Its now faster than raytree (both on build and query)
Things tryed:
X=>Y=>Z=>X split (reduces build time.. but increases query time)
bucket sorts
(initial sorts for fast usage of bucket take a long time)
(nth is linear.. so its quite fast already)
Best times archieve with:
*usage of 4-ary trees.. reduces build time and tree size but didnt decreased query time
*quads are on the same node instead of splitting in 2 tris..
(this actually turned on speedup on query time.. since tree size is reduced by a factor of 2)
*test ray-bb before ray-primitive gives better times on both tris and quads
Notes:
measures where made projecting a sphere from inside the head of suzanne.
small bvh fixes:
*allow to create any tree type >= 2
*save split axis
changed shrinkwrap to perform normal cast with raytree and bvh tree and print both times:
Shrinkwrap (OBCube)24578 over (OBSuzanne)504482
target = raytree_create_from_mesh(calc->target): 1260.000000ms
shrinkwrap_calc_normal_projection_raytree(&calc): 1850.000000ms
tree = bvhtree_from_mesh_tri(calc->target): 3330.000000ms
shrinkwrap_calc_normal_projection(&calc): 3780.000000ms
On general query time is bit smaller on bvh tree..
but the build time of bvh is pretty big.
(build time can be removed from both if a cache system is added)
But I am still trying to see how fast I can make the bvh build
*Fixed a few UI things
*Make SimpleDeform and Shrinkwrap to use vertexgroup_get_vertex_weight, a similar function
"static float vert_weight(MDeformVert *dvert, int group)" existed on modifier.c, changed
it a bit and moved into BKE_deform.h
Genscher made me do it the same way that the other 10 modifiers do.
The reason why I initially choosed memcpy was so that if anyone added
parameters on DNA there would be no need to update _copyData, unless
if it was a feature related to memory (eg.: point cache)
Changed code to remove faces (now quad faces that got one vertice projected are turned on tri)
Merge option is still not very usefull since shrinkwrap does not yet moves unprojected vertices
This adds redcode (the file format of RED one, R3D) support to blender.
Seems to work fine with the footage I found on the web, but keep in
mind, that because of the unoptimized nature of libopenjpeg, frame
decoding isn't that fast.
It is also a rather challenging task, to make 4k-float-footage realtime :)
Some more Bmesh custom data functions and
structures. This still does not do anything
yet because the various conversion functions
don't bother making use of the new custom data
functions. Hooking them up should be fairly
simple though.
Also note that the custom data code is mostly
copy/pasted from the existing custom data
functions for editmode with a few modifications.
Duplicating code like this isn't nice, but I
felt it was better to keep things for Bmesh
'standalone' for the moment and take only what is
immediatly needed instead of creating
a tangle of interdependant code.
Notes:
- edgehash.c still has some weirdo code causing warnings on lines 80 and 117
i.e. if (v1<v0) v0 ^= v1 ^= v0 ^= v1;
- material.c (in pyapi) apparently doesn't seem to be making use of some functions for glossy stuff
Now that new allocator is in place, Custom Data
can be effeciently added to BMesh. The plan is to
make all data not directly related to topology
Custom Data and allow callers to decide precisely
what information a mesh should have in order to
make the best tradeoff between memory usage/speed.
Right now not much to look at, just some structure
definitions and commented out code. More to come
soon...
Added a new pooling allocator for Bmesh based upon
the pool allocator availible in the Boost C++ library
as described here:
http://www.boost.org/doc/libs/1_34_0/libs/pool/doc/concepts.html
Each pool allocates elements of a fixed size, so every
element type in a mesh gets its own pool. For instance
verts occupy a different pool than edges. Each pool
is comprised of multiple arrays of a fixed size and allocating
/freeing elements is simple as removing or adding a head
to a linked list. Since the list of free elements is interleaved
throughout the unused space in the arrays, the overhead
for storing the free list is only 1 pointer total per pool.
This makes building/destroying bmesh structures much faster
and saves quite a bit of memory as well.
