README

This specializer implements the matrix powers kernel, used in communication-avoiding Krylov subspace methods: given a sparse matrix A and vector x, compute A*x, A^2*x, ..., A^k*x (or a different basis spanning the same subspace).

Files included:
* akx.py - the specializer module itself
  * akx-static.c, akx-powers.tpl, akx.h - templates for C code
* test.py - test driver for matrix powers kernel
* cacg.py - communication-avoiding CG solver
* cgtest.py - test driver for CA-CG solver

Libraries required:
* asp (https://github.com/shoaibkamil/asp)
* numpy and scipy (http://www.scipy.org/)
* PaToH (http://bmi.osu.edu/~umit/software.html#patoh)
* Intel MKL (not needed for the matrix powers kernel itself, but used for other operations in CA-CG solver)