GC statistics (was Re: big project ported)

[Tom Tromey]

>>>>> "Hans" == Boehm, Hans <hboehm@exch.hpl.hp.com> writes:

Hans> But that's not necessarily sufficient, and I have since seen one
Hans> real instance in which the compiler successfully hid a pointer
Hans> from the collector, and thus introduced a bug.  (This occurred
Hans> with Visual C++, which seems to have grown a fairly aggressive
Hans> optimizer over the years.)

The gcc guys tell me that gcc is getting better at doing this sort of
thing, too.  I'm not sure it is that much of a problem for Java, but
you never know.

Hans> Does gcj use the normally default of defining
Hans> ALL_INTERIOR_POINTERS?

Not right now.
We currently define SILENT, NO_SIGNALS, NO_DEBUGGING, and
JAVA_FINALIZATION.
When cross-compiling we also define NO_SIGSET, NO_CLOCK, and
SMALL_CONFIG.

Hans> Even if it doesn't, all interior pointers originating on the
Hans> stack or in registers are recognized.

Good.  The Java heap should only contain pointers to object starts
anyway.

Hans> I would like to eventually see the gcc back-end ensure some form
Hans> GC-safety, at least for potentially large arrays.

Me too.

T