Bug release of v3.0

Bug release of v3.0. Missed calc.c

Sparse implementation in C (under Cython) and Scala

Sparse implementation in C speeds up Cython quite a bit. Same sparse algorithm implemented in Scala offers C-like speed (using MTJ and EJML for matrix storage). These are very impressive benchmarks. Large sample takes less than a minute to finish.

Cython sparse implementation

This release added a version of sparse implementation in Cython, suggested by Will Meyer. Its performance deteriorated somewhat, compared to dense matrix, due to more complicated security key lookup. But the memory footprint improved as you would expect from sparse matrix.

Cython and Scala Implementations

Both Cython and Scala implementations offer much higher performance than Python. Cython provides a baseline benchmark for C-like performance.

Large sample is added for better large-scale benchmark

Large position size sample is added. This is 7 times of the small sample in v1.0. This will take 30 minutes to an hour to process on a typical desktop; therefore, serves as a better performance benchmark. Python benchmark is updated for the large sample in this release.

Python implementation

This first release contains the sample data and the python sparse matrix implementation.