Maths Coder

I started to fork your repo, it takes me some month to understand in deep your great work. Best regards.

Hello @Nucs A. Ok a devel branch is more collaborative than a fork B. Lazy Loading can separate the SharpNum library onto frontend with simple to write specification which will...

@Oceania2018 Yes it is a struct that can be defined like this during internal computation of the backend ``` NDproxy a = new NDproxy(()=>ndArray0) ``` Its contains only one field...

@Nucs you are right! Lazy Loading adds some stacks operations because its a kind of managing expression pattern. Managing code have advantage for simplifying coding and maintenance, but with loss...

@Nucs The idea behind the lazy loading is to introduce step by step some symbolic neural network algorithms with the numerical algorithms. For beginning, a precompilation (or may be caching)...

I forget to comment an important thing about some counters. ``` var a = ND.Uniform(1, 10, 4, 4); // Counters. DataAccess:16 / MethCall:16 var b = 3 * ND.Sq(a -...

I discovered in this article https://techdecoded.intel.io/resources/parallelism-in-python-directing-vectorization-with-numexpr/#gs.63gvwl Numexpr : it is very usefull and it already use lazy-loading and a lot of symbolics optimisations to enhance performance.