Re: Access pattern optimizations

May I suggest going back through some recent SIGPLAN conference proceedings?
There were a number of related papers in the most recent conference, for
example. Note that loop unrolling helps this case a fair amount by revealing
the reuse, and dependency analysis (determination of which iterations produce
or overwrite values needed by other iterations, important if y was being
updated rather than x) become relevant. Guy Steele's paper "FORTRAN at 10
Gigaflops" tells how Thinking Machines used loop unrolling and good register
allocation to achieve tremendous speed (though the entire processor is highly
parallel, the unrolling and allocation happen within the individual floating
point units, so have relevance to sequential machines). There was also at
least one paper about how to avoid taking a cache miss on every iteration, by
reorganizing the iterations of nested loops. Hope this gets you started!
--


J. Eliot B. Moss, Assistant Professor
Department of Computer Science
Lederle Graduate Research Center
University of Massachusetts
Amherst, MA 01003
(413) 545-4206, 545-1249 (fax); Moss@cs.umass.edu
--

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