This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Patch: Support IA-64 speculation [4/5]
- From: Andi Kleen <ak at suse dot de>
- To: Jan Hubicka <hubicka at ucw dot cz>
- Cc: gcc at gcc dot gnu dot org
- Date: 29 Jul 2006 21:05:55 +0200
- Subject: Re: Patch: Support IA-64 speculation [4/5]
- References: <43AC335D.1080201@ispras.ru.suse.lists.egcs-patches> <43FF1398.9010007@redhat.com.suse.lists.egcs-patches> <4418DE23.4020508@ispras.ru.suse.lists.egcs-patches> <20060726225816.GA16035@atrey.karlin.mff.cuni.cz.suse.lists.egcs-patches> <44C8AE61.60408@ispras.ru.suse.lists.egcs-patches> <20060728100647.GD16035@atrey.karlin.mff.cuni.cz.suse.lists.egcs-patches> <44C9EC10.6000400@ispras.ru.suse.lists.egcs-patches> <20060728123527.GE16035@atrey.karlin.mff.cuni.cz.suse.lists.egcs-patches> <44CA0725.8000303@ispras.ru.suse.lists.egcs-patches> <20060728135408.GF16035@atrey.karlin.mff.cuni.cz.suse.lists.egcs-patches>
Jan Hubicka <hubicka@ucw.cz> writes:
>
> Adding backward pointers will increase their size by 30%, so perhaps
> some mark dead and collect later scheme would work here better?
Can't you use a xor list? It works as long as you can keep
enough context to know about the following (or previous) entry too.
However if it's garbage collected it might be difficult to teach the GC
about it because the pointers will be encoded. I guess it could
be done however with some tricks.
-Andi