Possible accelerations to colorize_by_geodesic_distance_from_a_given_point

[zoyt]

Hello,

I'm working on a project where I call colorize_by_geodesic_distance_from_a_given_point several thousand times, varying depending on the mesh. I've been looking into multiprocessing approaches given that I have quite a bit of CPU headroom, which I still think may have some merit, but I have been encountering issues with pickling that I lack the background to understand well enough to troubleshoot. If anyone has any experience or advice with this, I'd love to talk further.

However, in the meantime, I have to believe that there are ways to more efficiently apply the filter within meshlab. This is my current code:

#find faces from verts
    facesFromVertDic = {}
    faceArray = ms.mesh(currentMeshNum).face_matrix()
    for face_index, face in tqdm(enumerate(faceArray)):
        for vertex_index in face:
            facesFromVertDic.setdefault(vertex_index, []).append(face_index)

def multiSortedClosestFacesToPoint(coords):
    """
    returns: [0]-faces  ,  [1]-ordered verts
    """
    ms.apply_filter("colorize_by_geodesic_distance_from_a_given_point",
                    startpoint=[coords[0], coords[1], coords[2]],
                    maxdistance=maxGeodesicDist)  # stores the verts within the search cutoff in quality array
    pointsWithinDist = where(ms.mesh(currentMeshNum).vertex_quality_array() > 0)[0]

    sortingInds = ms.mesh(currentMeshNum).vertex_quality_array()[pointsWithinDist].argsort()
    sortedpoints = pointsWithinDist[sortingInds]

# find all faces
    allfaces = []
    for point in pointsWithinDist:
        for face in facesFromVertDic[point]:
            allfaces.append(face)
    unOrderedFaces = unique(asarray(allfaces))

# order Faces
    sortingRefList = []
    faceMatrix = ms.mesh(currentMeshNum).face_matrix()

    for face in unOrderedFaces:
        verts = faceMatrix[face]  # problem child

        vertsListIdx = [idx for idx, element in enumerate(sortedpoints) if element in verts]
        sortingRefList.append(vertsListIdx[0])
    sortingInds = asarray(sortingRefList).argsort()

    orderedFaces =  unOrderedFaces[sortingInds]

    print("complete")
    return orderedFaces, sortedpoints

I've done quite a bit of profiling and found the ms.apply_filter call to be the most time intensive, taking about .25 seconds per iteration all told. Any and all advice or thoughts would be much appreciated.

Thank you!

Owen