README_Rob

This package contains tools to make use of very large matrices of floating point numbers in an efficient manner.

Before using the package, run make to compile the C utilities.

This package contains 5 tools:

API.R, API.py:  These are api's that are used to extract elements or rows of the matrices efficiently.  

transpose2:  This utility transposes a matrix.  The arguments are:

transpose2 <infile> <outfile> <chunk> <start>

where:
  infile: the file to transpose
  outfile: the new, transposed file
  chunk: the number of columns to transpose in one pass.  This is a tuning parameter.  You can try various small integers, or
          just use 1.
  start: This is the char position in the output file to begin at.  When starting, use 0, but if the job is interrupted, you can
         restart at the current size of the output file to pick up where you stopped.

The last two utilities pertain to the row and column counts, which should be the last 8 bytes of the binary matrix file, and 
which are used for sanity checking.  GetCounts can be used to query the current counts.  FixCounts sets the counts.  In both
cases you need to specify the byte location where the counts should begin.

getCounts:
  getCounts <file> <pos>

  This prints the counts found beginning at <pos>.  Normally you would run this giving pos as the size of the file-8.

fixCounts:
  fixCounts <file> <pos> <rows> <cols>
  This overwrites the 8 bytes beginning at <pos> with the provided row and col counts.  

Note that transpose2 expects to find counts at the end of the file, and will swap the counts appropriately in the new file,
so you don't need to worry about them.