Re: Multi-threaded compilers

My research group at the University of Toronto did a lot of work on
the design of multi-threaded compilers. The hard problems were
finding enough work to keep the processors busy during the early parts
of a compilation and handling symbol table access when the table was
being used and built concurrently. With a lot of hard work tuning the
structure of the compiler for performance we were able to get quite
good parallel speedup (see the PLDI92 paper or CSRI-235). The
production compiler (DEC SRC Modula-2+) that we used as a basis for
our research prototype did very heavy weight dynamic allocation of
compiler internal data structures. I'd expect that a compiler which
did less in this area would achieve better parallel speedups. We've
published a number of papers about our work.


V. Seshadri and D.B. Wortman, An Evaluation of Concurrent Semantic
Analysis, Software - Practice & Experience, v. 21, n.12, December
1991, pp. 1323-1348


V. Seshadri, I. Small and D.B. Wortman, Concurrent Compilation,
Proceedings of the IFIP Conference on Distributed Computing,
Amsterdam, 5 Oct 1987, pp. 627-641


V. Seshadri, D.B. Wortman, M.D. Junkin, S. Weber, C.P. Yu and
I. Small, Semantic Analysis in a Concurrent Compiler, ACM Sigplan '88
Conference on Programming Language Design and Implementation, Atlanta,
June 1988, pp. 233-240


D.B. Wortman and M.D. Junkin, A Concurrent Compiler for Modula-2+,
Proc. ACM SIGPLAN Symposium on Programming Language Design and
Implementation, San Francisco, June, 1992, pp. 68-81 (* Final
performance and design summary *)


M.D. Junkin, D.B. Wortman, Implementation of a Concurrent Compiler,
Technical report CSRI-235, Computer Systems Research Institute,
University of Toronto, Dec 1990 (* Everything you ever wanted to know,
up to Dec 1990 *)
--

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