More polished

I've updated a lot of documentation, which is now written to compile nicely in doxygen.   Libpng is no longer assumed to be installed, and will compile without.   Furthermore, issues the old API have become more apparent, with tests finding more useless input.

I've written out revised schedule for the project at this point.   Making sure everything works as bug free as possible, with a understandable interface is top priority.  

July 20 - - an implementation over GF(3), GF(5) and GF(7) of this API (or a variant
thereof)

------------------------------------------------------------------------------------------------------------
m3d_t *m3d_create(rci_t m, rci_t n);

void m3d_free(m3d_t *A);

m3d_t *m3d_concat(m3d_t *C, const m3d_t *A, const m3d_t *B);

m3d_t *m3d_stack(m3d_t *C, const m3d_t *A, const m3d_t *B);

m3d_t *m3d_submatrix(m3d_t *S, const m3d_t *M, const rci_t lowr, const rci_t lowc, const rci_t highr, const rci_t highc);

m3d_t *m3d_init_window(const m3d_t *A, const rci_t lowr, const rci_t lowc, const rci_t highr, const rci_t highc);

void m3d_free_window(m3d_t *A);

m3d_t *m3d_add(m3d_t *C, const m3d_t *A, const m3d_t *B);
m3d_t *m3d_sub(m3d_t *C, const m3d_t *A, const m3d_t *B);

m3d_t *m3d_addmul_naive(m3d_t *C, const m3d_t *A, const m3d_t *B);
m3d_t *m3d_mul_naive(m3d_t *C, const m3d_t *A, const m3d_t *B);
m3d_t *m3d_mul_scalar(m3d_t *C, const long a, const m3d_t *B);

void m3d_rand(m3d_t *A);

m3d_t * m3d_copy(m3d_t *B, const m3d_t *A);

void m3d_set_ui(m3d_t *A, long value);

long m3d_read_elem(const m3d_t *A, const rci_t row, const rci_t col);

void m3d_add_elem(m3d_t *A, const rci_t row, const rci_t col, const long elem);

void m3d_write_elem(m3d_t *A, const rci_t row, const rci_t col, const long elem);


int m3d_cmp(m3d_t *A, m3d_t *B);

int m3d_is_zero(const m3d_t *A);

void m3d_add_multiple_of_row(m3d_t *A, rci_t ar, const m3d_t *B, rci_t br, long x, rci_t start_col);

void m3d_add_row(m3d_t *A, rci_t ar, const m3d_t *B, rci_t br, rci_t start_col);

void m3d_rescale_row(m3d_t *A, rci_t r, rci_t start_col, const long x);

void m3d_row_swap(m3d_t *M, const rci_t rowa, const rci_t rowb);

void m3d_copy_row(m3d_t* B, rci_t i, const m3d_t* A, rci_t j);

void m3d_row_add(m3d_t *M, const rci_t sourcerow, const rci_t destrow);

void m3d_print(const m3d_t *M);
------------------------------------------------------------------------------------------------------
    I have most of these written out, though a few, such at concat don't have m5d and m7d versions at the moment.

July 27 - test cases covering these functions with multiple inputs, at least more complicated ones like arithmetic.



August 3 - clear documentation of what each of these functions explaining what these functions do (with their inputs) and what conditions they assume.

Have a lot of these written at this point.

August 17 - an implementation of Strassen for at least m3d , with documentation, test cases and timing.    Worked on a lot of m3d_strassen but need to test input and accuracy.