Re: Parallel compiling

We have been looking at parallel techniques for parsing, starting
from the lexing result of the folks at Thinking Machines. We have
shown that, for SIMD-type architectures, languages whose grammars are
in LL can be parsed in O(log n) time. This does NOT extend to LR
(in fact, it's another interesting characterization of the difference).
Of course, this means that parallel parsing can be done on MIMD machines
with an extra log n penalty. We've also shown how to pipeline the
algorithm.


        We've also started looking at parallelizing semantic analysis.
Stay tuned for success/failure.


        We can send you a copy of the paper either electronically if you
have a troff environment and a decent email address; or by snail
mail if you send me an address (to skill@qucis.bitnet). We'd also
like to hear from anyone who's been thinking along the same lines.


                                                                                          David Skillicorn
                                                                                          David Barnard
                                                                                          Computing and Info Science
                                                                                          Queen's University
                                                                                          Kingston Ontario
                                                                                          Canada
--

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