Generating analyzers with PAG

Abstract

To produce high qualitiy code, modern compilers use global optimization algorithms based on it abstract interpretation. These algorithms are rather complex; their implementation is therfore a non-trivial task and error-prone. However, since thez are based on a common theory, they have large similar parts. We conclude that analyzer writing better should be replaced with analyzer generation. We present the tool sf PAG that has a high level functional input language to specify data flow analyses. It offers th specifications of even recursive data structures and is therfore not limited to bit vector problems. sf PAG generates efficient analyzers wich can be easily integrated in existing compilers. The analyzers are interprocedural, they can handle recursive procedures with local variables and higher order functions. sf PAG has successfully been tested by generating several analyzers (e.g. alias analysis, constant propagation, inerval analysis) for an industrial quality ANSI-C and Fortran90 compiler. This technical report consits of two parts; the first introduces the generation system and the second evaluates generated analyzers with respect to their space and time consumption. bf Keywords: data flow analysis, specification and generation of analyzers, lattice specification, abstract syntax specification, interprocedural analysis, compiler construction.