Before working on the shim, I added optimizations to IBM's CLP() compiler, as part of my doctoral research. Since the original code is not freely releasable, and I have not had time to clean up and organize the parts I wrote for public release, the best pointer I can give you to this work is my dissertation itself, Optimizing Threads of Computation in Constraint Logic Programs. The conversion to pdf for this copy is much better quality, and so much easier to read, than the one posted at OhioLINK.
Outside of rebuilding from the latex and postscript sources to take advantage of modern dvips and ps2pdf systems, I've resisted the temptation to ``improve'' it, and left the text strictly alone. The introduction, in the first chapter, and the results and conclusions, in the last two chapters, are the most accessible parts. Chapter 3, Affine Transform-Based Analysis of Loop Bodies, is the heart of the work.