Code I've written that may be useful. Let me know if you find a bug. Don't blame
me if your computer catches on fire.

<p> <b>Individual files</b>:
<p> <a href="opensrc/FindWindowedPeaks.m"><tt>FindWindowedPeaks.m</tt></a>:
  Extract peaks from a time series given a desired minimum gap between peaks.

<p> <a href="opensrc/ptr.m"><tt>ptr.m</tt></a>: A pointer in Matlab! If <tt>p =
    ptr()</tt> and you've set <tt>p.d.fld = 'hi'</tt>, then dereference it like
    this: <tt>p.d.fld. Think of "<tt>.d.</tt>" as the C language's "<tt>-></tt>"
    operator.

<p> <a href="opensrc/mop.m"><tt>mop.m</tt></a>: Perform (A op B) where one of A
  or B is a vector instead of a matrix that needs to be <tt>repmat</tt>'ed to
  conform. On the command line, this gets tiresome; <tt>mop</tt> figures out
  what needs to be done and does it.
<p> <a href="opensrc/SaveStreamData.m"><tt>SaveStreamData.m</tt></a>: Functions
  to save a stream of data to a file. The data should be column oriented, so
  that one is building a wide matrix by appending columns. The purpose of this
  file is to provide an efficient implementation of a common I/O model in
  Matlab: save a stream of data as a simulation proceeds. For example, this
  routine could be used in the <tt>OutputFcn</tt> for <tt>ode</tt>* when solving
  a large problem.
<p> <a href="opensrc/GetMyCode.m"><tt>GetMyCode.m</tt></a>: It's useful to save
  the code associated with data along with the data in a mat file. That way, in
  the future, you know exactly what version of the code created the data. Use
  this routine to do that.
<p> <a href="opensrc/SlashUnderscore.m"><tt>SlashUnderscore.m</tt></a>: Place a
  backslash before underscores in a (cell array of) string(s) before plotting.
<p> <a href="opensrc/FdApprox.m"><tt>FdApprox.m</tt></a>: Calculate the
  coefficients in an approximation to the <i>d</i>th derivative for a uniformly
  spaced grid.
<p> <a href="opensrc/ssbin.m"><tt>ssbin.m</tt></a>: Matrix-free approximate
  equilibration for a symmetric matrix.
<p> <a href="opensrc/snbin.m"><tt>snbin.m</tt></a>: Matrix-free approximate
  equilibration for a nonsymmetric matrix.
<p> <a href="opensrc/dsbin.m"><tt>dsbin.m</tt></a>: Livne and Golub's
  algorithm for the equilibration of a symmetric matrix.
<p> <a href="opensrc/dsbin_mex.c"><tt>dsbin_mex.c</tt></a>: Livne and Golub's
  algorithm for the equilibration of a symmetric matrix. Mex implementation.


<!--
<p> <a href="opensrc/mywhos.m"><tt>mywhos.m</tt></a>: <tt>whos</tt> with
  human-readable memory size. Depends on <tt>lstrlen</tt> and <tt>hbytes</tt>.
<p> <a href="opensrc/lstrlen.m"><tt>lstrlen.m</tt></a>: Return the longest string
  length in a cell array of strings.
<p> <a href="opensrc/hbytes.m"><tt>hbytes.m</tt></a>: Human-readable memory size.
<p> <a href="opensrc/Plot3Shadows.m"><tt>Plot3Shadows.m</tt></a>: <tt>plot3</tt> +
  2D projections.
<p> <a href="opensrc/CurveArrows2.m"><tt>CurveArrows2.m</tt></a>: Plant arrows
  along a curve.
<p> <a href="opensrc/Plot3ShadowsArrows.m"><tt>Plot3ShadowsArrows.m</tt></a>:
  <tt>Plot3Shadows</tt> + add arrows to the 2D projections. Good for making 3D
  phase plots: e.g., <a href="hot_plot.jpg">this one</a>.
<p> <a href="opensrc/Pretty.m"><tt>Pretty.m</tt></a>: Make a plot pretty. You
  might modify this function to taste.
<p> <a href="opensrc/myaxistight.m"><tt>myaxistight.m</tt></a>: <tt>axis
  tight</tt> with some space around the edges.
<p> <a href="opensrc/pn.m"><tt>pn.m</tt></a>: Print something simple without
  skipping a line like <tt>disp</tt> does.
<p> <a href="opensrc/spdiag.m"><tt>spdiag.m</tt></a>: Sparser syntax
    than <tt>spdiags.m</tt> for sparse diag matrix.
<p> <a href="opensrc/mexlu.c"><tt>mexlu.c</tt></a>: Call <tt>[L U p q R] =
    lu(A,'vector')</tt> from a mex file and solve systems. Optionally change a
    column of A and append a corresponding factor to the factorization. At least
    one reason to perform triangle solves in C rather than call a mex function
    is for thread safety when using OpenMP in a mex file. Another is to avoid
    unnecessary memory movement. [Later: I recently learned that the UMFPACK
    symbols are exported in Matlab, so one can call the factorization and solve
    routines directly from mex files. I have yet to play around with this. Such
    functionality, if it does not have a catch, obviates the need
    for <tt>mexlu.c</tt>.] [Still later: See next entry.]
<p> <a href="opensrc/testumf.c"><tt>testumf.c</tt></a>
  and <a href="opensrc/testumfmp.c"><tt>testumfmp.c</tt></a>: Demonstration of
  calling UMFPACK in a mex file. The code uses the symbols available in Matlab
  rather than relying on a separate installation of UMFPACK. I'm not sure how
  portable this technique is. The second file demonstrates parallelizing
  multiple solves using OpenMP. My experience is that the technique demonstrated
  in <tt>mexlu.c</tt> is 10-30% faster than that in <tt>testumf.c</tt>. (One
  needs to be careful in such tests: for example, the number of iterations of
  iterative refinment should be the same. I'm not sure the other defaults in
  UMFPACK's control array are the same that Matlab uses.) I'm not ready to
  conclude that one is definitely faster than the other, but they are at least
  comparable.


-->