From: Craig Chambers (chambers@cs.washington.edu)
Date: Wed Dec 18 2002 - 16:01:18 PST
I'm publishing some more.
*) I fixed the bug I discovered with splitting in the Java
benchmarks.
*) I fixed another bug in a different part of Vortex, where
transformations were being applied to N-way branches, but N-way
branches don't support non-trivial graph replacements. (Man, how
could these programs have been compiled any time recently?!)
*) I added a limit to the size of Vortex comparison facts (i.e.,
results of comparison tests, e.g. for array bounds check
optimization), to throttle back on some reallllllly slooooow
compilations (sometimes seeming to take an infinite amount of time).
*) I made some small improvements and fixes to the debugger, yet
again. This round adds printing of any type parameters for methods
and messages and fixes a small problem with doing a continue command
after a next command.
*) I made a small tweak to ^C signal handling in the run-time system
for solaris8 machines, since the second ^C seemed to be killing the
program rather than reenterring the Vortex debugger.
*) To make the Java benchmarks easier to test systematically, and
similarly between Vortex and Whirlwind, I shuffled around some of the
Java benchmark control files.
*) I fixed some small problems in the benchmark copy (in
vortex/Cecil/src/benchmarks) of the Cecil stdlib, to make that code
compatible with the current run-time system.
*) I tweaked the for-release notes on how to make a release.
>From these new sources, I built a new vortex executable, used that to
(re)build the no-debug_support vortex executable (which will become
the release executable), and used that executable to build and run all
the Java benchmarks. I also used the no-debug_support vortex to build
whirlwind, and then used that whirlwind to build and run all the Java
benchmarks. Finally, I also used the no-debug_support vortex to build
and run all the Cecil benchmarks. Since all this worked, this means
that people who download the solaris8 version of the release
executable will be able to compile all the benchmarks, at least.
(Although we don't (currently) give out our Java benchmarks, nor do we
explain anywhere (I think) how to run the Cecil benchmarks. For the
next release, e.g. of Whirlwind, we should probably try to have a set
of canned benchmarks and instructions available, so that outside
people can reproduce our experiments, and have a framework in which to
add their own benchmarks.)
-- Craig
_______________________________________________
Cecil mailing list
Cecil@cs.washington.edu
http://mailman.cs.washington.edu/mailman/listinfo/cecil
This archive was generated by hypermail 2.1.5 : Wed Dec 18 2002 - 16:01:23 PST