These are chat archives for libmir/public

6th
Feb 2017
Ilya
@9il
Feb 06 2017 07:22

what are you doing to ensure safety of the reference to the engine?

Nothing can be done. The docs should be noted, that it hold a reference and it is unsafe to pass it outside from a function.

@anton-dutov Yes ndslice is slower for very small matrixes like 3x3 and 4x4. See also https://github.com/gecko0307/dlib . The new ndslice https://github.com/libmir/mir-algorithm is faster then the old ndslice because it optionally may not have strides. But the length are not known at compile time, so fixed sized types are faster