Order related query about mlinterp HOT 3 CLOSED

The two should be comparable in performance (see https://github.com/parsiad/mlinterp/blob/master/mlinterp/mlinterp.hpp#L94-L122), but I have not checked. It would be interesting to look at the assembly and/or benchmark, if you are up to it.

from mlinterp.

If I will test it, will let you know. Not very sure. Only point was about doing contiguous memory access for large arrays so that swaping doesn't affect performance. The good part is you have kept it configurable. And one more point are the interp methods MT safe?

from mlinterp.

Consider the following example:

interp(
	nd, ni,        // Number of points
	zd, zi,        // Output axis (z)
	xd, xi, yd, yi // Input axes (x and y)
);

The only buffer that is written to is zi. The other buffers (in this case, zd, xd, xi, yd, yi) are read-only and can be shared among threads.

from mlinterp.