This is the mail archive of the java@gcc.gnu.org mailing list for the Java project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

RE: 100x? (Was: http://gcc.gnu.org/java/done.html)


> From: Jeff Sturm [mailto:jsturm@one-point.com]
> 
> On Wed, 30 Jul 2003, Boehm, Hans wrote:
> > Could you supply any details?  Platform?  GC_dump() output?  GC log
> > output with GC_PRINT_STATS environment variable defined?  A 
> profile?  (I
> > can point you at a relatively painless way to get one under 
> Linux/X86,
> > if you like.)
> 
> It sure would be good to have that information.  Also I wonder if this
> application would be a good test case for incremental mode.
> 
Perhaps.  But based on the posted performance numbers, I strongly suspect
something is clearly broken here.  My impression based on past experiments is
that if generational collection works really well on your app, and you do
nothing but allocate memory, Sun's VM may be a factor of 4 or 5 faster.
If generational collection doesn't work well, and you do nothing but allocate,
gcj may be a factor of 2 or 3 faster.  But I've not seen a factor of 10,
especially on real code.  Nor have I seen an application that spends 95%
of its time in the collector, as this one would have to.

Hans


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]