[Gc] Re: [bigloo] Performance evaluation

Manuel.Serrano at sophia.inria.fr Manuel.Serrano at sophia.inria.fr
Wed Aug 11 20:58:48 PDT 2010


> There are two things that worry me about this:
> 
> 1) The live data size seems to have increased.
> 
> 2) It seems to be collecting earlier, in spite of the larger heap.
> 
> You're using the collector without threads or incremental GC, right?
We never use the collector in incremental mode and in that particular example
we use it without threads (although in general, we do use it with threads).
> 
> How is the live size computed?
The GC is slightly patched for adding some extra profiling information. To 
compute the live size, we modify the function GC_finish_collection by adding 
the following:

    if( GC_gcollect_hook )
       GC_gcollect_hook( GC_heapsize,
			 (unsigned long)(GC_atomic_in_use)+
			 (unsigned long)(GC_composite_in_use) );

at the end of the function. Here is attached the whole modified alloc.c
file.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: alloc.c
Type: text/x-csrc
Size: 46344 bytes
Desc: not available
Url : http://napali.hpl.hp.com/pipermail/gc/attachments/20100812/abba3eb6/alloc-0001.c
-------------- next part --------------

-- 
Manuel


More information about the Gc mailing list