This is the mail archive of the
mailing list for the GCC project.
Re: getting rid of ggc_push_context
Jan Hubicka <email@example.com> writes:
> I wonder how desirable is to get rid of ggc_push_context/ggc_pop_context
> machinery. It should not be terribly dificult deal to make cse.c ggc
> safe but without using ggc_free that would bring other expenses.
> What GGC guys prefer?
> Perhaps it might be possible to trottle the cse memory consumption too.
> Does someone recall why this call has been added?
All else being equal, it is highly desirable to get rid of this
machinery. In ggc-page, it's expensive and it ruins memory locality.
In ggc-zone, I'm not sure if it works at all. Dan Jacobowitz would
know, but he's out of town.
I seriously wonder whether the ggc_collect call inside CSE is still
necessary. With the tree optimizers in place, it probably isn't
generating nearly as much garbage as it used too. And our
memory-pressure heuristic lets us chew up tons of memory before the GC
actually runs, anyway. Dan had numbers on this, but again, he's out
of town. :(