% File: README

This is the README file for the test directory, where I
have included some sparse matrices for you to play around
with. The first simple example is:

  test2

For this data, we have the following files:

-rw-r--r-- 1 suvrit    58 2005-05-21 23:44 test2_col_ccs
-rw-r--r-- 1 suvrit     8 2005-05-21 23:44 test2_dim
-rw-r--r-- 1 suvrit   200 2005-05-21 23:44 test2_row_ccs
-rw-r--r-- 1 suvrit   900 2005-05-21 23:44 test2_txx_nz
-rw-r--r-- 1 suvrit    87 2007-12-04 12:31 test2.vals
-rw-r--r-- 1 suvrit  1221 2007-12-04 12:31 test2.vecs

That is, the data is in the 3 file version CCS (compressed
column storage) data.

The next important file is: options.las1 who contents are:

$ cat options.las1
test2 M_CCS3TXX TXT 10 5 -1e-30 1e-30 1e-3 VECS ASCVECS VERBOSE

Which means that 

$ ../src/svd 1 options.las1 

will run the LAS1 algorithm on the matrix in test2 and will 
deposit the requested singular values in test2.vals and the
(in this case) requested singular vectors in test2.vecs


II. The next example is the classic3 data, for which we have
$ cat options.las2
classic3 M_CCS3TXX TXT 50 20 -1e-6 1e-6 1e-3 NOVECS ASCVECS VERBOSE

$ ../src/svd options.las2

will now compute the top 20 (need to experiment with the value of the 
number of lanczos steps (here == 50) to ensure 20 singular values are 
computed to within the requested accuracy.