reaction diffusion (again)

I took a old solver from about 6 years ago and finally updated it to opencl, so much faster

I didn’t bother to do a exact speed comparison, but its pretty substantial.

I then moved on to doing this in opencl on a 3D grid, not sure the implementation is correct but it generates some interesting results.

It took me a bit to change the 5 point stencil to a 7 point for the volume grid, I originally tried it with connected point grid but things got really weird trying to reuse the 2D versions code which is based on connected points

link to the hip file

bootleg twee parser for houdini

I wanted to import a network of twee passages into Houdini to view as a connected graph

I used twee2 the command line tool to decompile twine2 gui html files

I the tw2 was a lot easier to strip apart in python.

each tagged passage become a node in the graph, originally each passage became a node, but I wanted some combined.
a bigger node tree

Just drop the code in a python sop and point it to the tw2 file, might work?

box packing 2D maps

I wanted a way to remove overlaps from some random rectangles.

each box getting pushed out, its a bunch of packed primitives with a vector attribute for width and height

Here is the vex code – run once in a detail wrangle it loops a number of times over the points and pushes them outside while checking for overlaps

int relax_iter=chi("iter");

vector pp[];

for(int i=0;i<npoints(0);i++){
    vector p1=point(0,"P",i);

for(int q=0;q<relax_iter;q++){
    vector newP={0,0,0};
    int in=0;
    for(int i=0;i<npoints(0);i++){
        for(int j=0;j<npoints(0);j++){
                vector min1 = pp[i]-point(0,"dimen",i);
                vector max1 = pp[i]+point(0,"dimen",i);
                vector p1 = pp[i];
                float x1min = min1.x;
                float x1max = max1.x;
                float z1min = min1.z;
                float z1max = max1.z;
                vector min2 = pp[j]-point(0,"dimen",j);
                vector max2 = pp[j]+point(0,"dimen",j);
                vector p2 = pp[j];
                float x2min = min2.x;
                float x2max = max2.x;
                float z2min = min2.z;
                float z2max = max2.z;
                newP = p1+rint(normalize(p1-p2)*4);
                in = x1min <= x2max && x2min <= x1max && z1min <= z2max && z2min <= z1max? 1:0;

for(int i=0;i<npoints(0);i++){
    setpointattrib(0, "P", i, pp[i], "set");

I also wanted to build a connected structure from this – so I used prim’s algorithm to connect a tree, I disconnected some of the small boxes to keep more primary paths, since the small islands gets connected any way when everything is converted to right angles