Re: JavaMemoryModel: A memory model for the masses

From: Doug Lea (
Date: Wed Nov 10 1999 - 09:12:15 EST

> > Concurrent Java programs routinely outperform
> > concurrent POSIX/C programs these days.
> i would like to see a head-to-head comparison where this is true. i seriously
> doubt it, unless the C version was horribly programmed.

These are just demos and benchmarks, but I invite you to try out the
taskDemo programs in my util.concurrent, and compare them to versions
in Cilk and Hood (C and C++ respectively -- see the URLs in
util.concurerrent to find them). Unfortunately, I believe that Cilk
and Hood are only available for sparcs and linux-intel. Last I checked
on multiway sparcs using exactvm, the ones that do not do floating
point are a little faster in Java, and the ones that do floating
point, especially on matrices, are (sometimes significantly) slower,
than the benchmark programs in C or C++ that they are based on. In
other words, the concurrency/parallelism aspects are pretty even.

I only have crummy, one-shot, non-definitive comparisons of other
stuff in util.concurrent vs C versions, but they have generally been
very encouraging. (Again, tested only on multiway sparcs.)

I don't have any application-level or system-level benchmarks
comparing the same or similar versions in Java and C. Does anyone?


JavaMemoryModel mailing list -

This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:22 EDT