sverchok icon indicating copy to clipboard operation
sverchok copied to clipboard

Published 20 hours ago • verified publisher icon nortikin

Vector Interpolation node in Cubic mode

Open Durman opened this issue 2 years ago • 5 comments

Problem statement

It seems that it's possible to do something because having only 9 hundreds of vertices as input and only 8 output vertices for about 25ms seems pretty slow.

image

image

untitled.zip

Durman avatar May 13 '22 04:05 Durman

what happens to performance when the input-vertices are a numpy array? looks like this is the heavy bit.. range, where n is the number of input points.

https://github.com/nortikin/sverchok/blob/ed9b54be807a2cbdc0868290074b55fb26cbd5a7/utils/geom.py#L245-L259

The use-case for this node is more probably a len(input_vectors) < len(output_vectors).

zeffii avatar May 13 '22 09:05 zeffii

what happens to performance when the input-vertices are a numpy array?

No difference

Durman avatar May 13 '22 10:05 Durman

a numba implementation is worth investigating

zeffii avatar May 13 '22 11:05 zeffii

Cubic Spline node shares the same implementation and large input can be quite expected I guess.

image

Durman avatar May 13 '22 11:05 Durman

this is not pretty, but seems quite fast here after the first execution. https://github.com/nortikin/sverchok/pull/4471

zeffii avatar May 13 '22 13:05 zeffii