cout/*.cout        Output of python script (usually benchmark or test result)

*.py                Python script, usually benchmark or test, usually requires
                    flint_sage

flint.binding/      Python binding for small piece of FLINT library, requires
                     sage. Installs as flint_sage. Tests in separate files test*
                        
flint.binding/C        fast C code doing integer linear algebra (HNF, PLU, 
                        triangular inversion, matrix multipliction, determinant)

flint.binding/*.README  kinda documentation with installation instruction and FAQ

benchmark.script/       benchmark, test-benchmark and profile programs


A_by_B_mod_P.hacks/     Secret voodoo to calculate a*b mod p for small p on amd64 
                         and GPU, public domain

Changelog               History of bugs fixes and additions

To.do                   Plan of changes

ebuild/                 Useful ebuild scripts


Subroutines I am proud of:
~~~~~~~~~~~~~~~~~~~~~~~~~~

C file name              test program                   comment
-------------------+---------------------------------+---------------------------------------------------
nmod_mat_HNF.c     |  benchmark_nmod_mat_HNF-sage    | for dim=200, my code is 4.30/1.65=2.6 times faster

                   |  benchmark_Sage_hnf_square      | my re-implementation is slightly faster 
                   |                                 |  (by 11% for dim=3000)

tmod_mat/*          |  test_tmod_mat                  | for matrice over ring Z/(2**64*Z), find PLU 
                     |                                 |  decomposition and inverse

fmpz_mat/window_unsh.c | test_benchmark_invert_triU-4ways | my specialized subroutine to invert HNF
                       |                                  |  matrice with small determinant is faster
                       |                                  |  than Sage or FLINT general subroutine

nmod_mat/mul_pk_classical.c | nmod_mat-mul.c               | my subroutine multiplies 4x4 matrix modulo
                            |                              |  3**40 150% (2.5 times) faster than FLINT
                            
fmpz_mat/det_hermitian_deco  | test_hd_det.py              | my subroutine is 2.69 times faster than 
                 mposition.c |                             |  FLINT fmpz_mat_det() for bits=1000 and dim
                             |                             |  =500


             Hint: if          this column                  contains file name with no dot, then it is
                                                             both .py and .cout
