Morel desperately needs a good editor. This paper is replete with errors of grammar, punctuation, and usage; many of the sentences are awkward, vague, or poorly worded. The paper further suffers from Morel’s use of terms he does not define, incomprehensible notation, and a lack of references. (Where, for example, can I find out what is meant by “interval algebra”?) The poor quality of the presentation is unfortunate because some of the content of this paper is quite good.
The most interesting section is that on optimization in an ADA compiler, in particular the information on optimization of constraint checking. Other optimizations discussed are in-line expansion of procedures, constant folding, value propagation, and data overlaying. No measurements are given for performance improvements attributable to the optimizations.
The rest of the paper discusses:
Global data flow analysis--Including a formalization of data flow analysis, interprocedural data flow analysis, and the application of global data analysis to the elimination of redundant computations and to determination of live variables.
Global optimization--In this section, Morel shows that elimination of redundant computations and removal of invariant code from loops can be merged into a single optimization, which he calls the “elimination of partial redundancies.”
Implementation of data flow analysis and global optimizations in the ADA compiler.
I wish I could recommend this paper more enthusiastically. I can only say that it has some good material, but because of the poor quality of the writing and editing, getting through it will require some work.
