AW: [Mono-devel-list] GC/Memory allocation performance

Varga Zoltan vargaz at
Mon Jun 2 07:45:35 EDT 2003


  During incremental GC, the heap is write protected, which
system calls like read() to fail if they get passed a
pointer to memory
inside the heap. This happens in the current io-layer code.
Even if
this is fixed, enabling incremental GC only shortens GC
pauses, but
leads to longer total time spent doing GC. 



Paolo Molaro <lupus at> írta:

> On 05/30/03 Marcus wrote:
> > Doesn't Boehm's GC have a compacting mode? I know that
SRC Modula-3 (and its 
> > descendants) for over 10 years have used a Boehm GC that
is generational, 
> > incremental, and compacting.
> The standard Boehm GC can do some form of
> collectiong (but no compacting), but enabling it in mono
always results
> in a SEGV in libgc's code for me. It may still be a mono
issue, I didn't
> investigate it. That support also relies on trapping
SIGSEGV that mono
> uses too to handle null ref exceptions, so it may not be
possible to
> deal with it anyway.
> Enabling it for the C microbenchmark makes the numbers
worse, though,
> from 0.95 to 1.2 for the statically compiled version.
> lupus
> -- 
> lupus at                                    
> lupus at                             Monkeys do it
> _______________________________________________
> Mono-devel-list mailing list
> Mono-devel-list at

More information about the Mono-devel-list mailing list