Re: Intermediate Code Optimizations

"S.Bharadwaj Yadavalli" <sby@cs.usask.ca> writes:


>2. The classical optimizations such as dead code elimination,
> jump elimination, common subexpression evaluation and so on,
> *seem* adhoc ( given my very limited comprehension). Has there
> been any effort to provide some sort of a frame-work to these
> optimizations ?


Hi,


Thanks for giving me an excuse to plug one of my papers: ;-)


@incollection{ms,
          AUTHOR = {S. Novack and A. Nicolau},
          BOOKTITLE = {Languages and Compilers for Parallel Computing},
          PUBLISHER = {Springer-Verlag},
          TITLE = {Mutation Scheduling: A Unified Approach to Compiling
for Fine-Grain Parallelism},
          YEAR = {1994}
}


This paper presents a technique for combining instruction
scheduling, code selection and register allocation into a
unified framework in which trade-offs between the functional,
register, interconnect and memory bandwidth resources of the
target architecture can be made "on the fly" during scheduling
in response to changing resource constraints and availability.


Though not explicitly presented in terms of providing a
framework for performing "classical" optimizations, Mutation
Scheduling does have the effect of performing many classical
optimizations (e.g., scheduling, CSE, strength reduction, etc),
within a unified framework in which trade-offs inherent to
performing these trade-offs can be made. This differs
significantly from the conventional approach in which the the
different types of optimization are ordered and applied
separately, thus severely limiting the ability to make meaningful
trade-offs.


If you are interested in taking a look at it, the Mutation
Scheduling paper is available on-line in Postscript format at:


http://www.ics.uci.edu/~snovack/papers/ms/ms.ps


Cheers,
Steve


snovack@ics.uci.edu
http://www.ics.uci.edu/~snovack
--

Return to the comp.compilers page.
Search the comp.compilers archives again.