This software is a prototype implementation of the on-line fault tolerant
matrix multiplication(FTGEMM) described in the paper

Fault Tolerant Matrix-Matrix Multiplication: Correcting Soft Errors On-Line
(under review)

This project is in its very early stage; it's only for testing ideas now.

Before compiling you should modify CMakeLists.txt properly.  See the instructions
in CMakeLists.txt.  You should have GNU Scientifc Library, a fast BLAS
implementation and pthread for dgemm_test.

To compile:
    cmake .
    make

To test ft_dgemm:
    ./test/dgemm_test matrix_size errors_per_op check_numbers



TODOs:
- implement BLAS compatitble interface