I need to replace some Matlab code based around quadprog from the Optimization toolbox with C/C++, minimizing a quadratic function in an N-dimensional convex region.  I cannot pass Matlab code to the programming droids, lest their heads explode, so I need to convert it to something that they understand.  A recommendation for a reference book on quadratic programming is appreciated.

This is only a guess, but there used to be a book (probably out of print now) called "Numerical Recipes in C" which might have covered that territory.

http://en.wikipedia.org/wiki/Numerical_Recipes

ETA: Maybe there's something here:

http://www.nr.com/

And see the discussion at:

http://www.nr.com/forum/showthread.php?t=1379

I have that particular book, but unfortunately, solutions for Quadratic Programming are excluded.

I figured it out, between searches on the web, and a few insights from Barazaa's book:

Nonlinear Programming

I'm using a Cutting Plane algorithm to train a Linear Support Vector Machine which requires the solution of a sequence of quadratic programs.  The sub-problems are convex and apparently the state-of-the-art solution for these are Internal Point Methods.  The biggest difficulty I found, and in the end there is a very simple solution for, is how to keep the variables all non-negative.  If you attempt to put barriers on all of the variables the equations become very ill-conditioned, but just putting them on the slack variables apparently does the trick.  Another surprising thing that I found is that as the feasible region is convex you do not need a full blown line search, a simple ratio test that keeps the variables feasible along the Newton direction is sufficient.