This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Release RTL bodies after compilation (sometimes)


>>>>> Michael Matz writes:

Michael> But it helps _finding_ leaked memory, so it does influence performance
Michael> very much.  If you don't believe me look up the old mails from Honza where
Michael> he used ggc_free to reduce memory overhead by like 2000%.  It is the kind
Michael> of assertion you already mentioned, if it's made not to free the memory,
Michael> but instead mark the object as dead in checking mode.

	Verifying that an object is freed by a specific point is good.
That is different than asserting to the garbage collection system that
something can be and should be freed.  As we have seen, even with the
checks implemented by ggc_free(), we still can have dangling references
that break the compiler, instead of a well-defined ICE from the GC
subsystem.  Hopefully the zone-collector will allow more accurate checking
and better garbage collection performance.  I believe that I am accurately
stating that if you read Jeff and Diego's comments, they support more
powerful verification of GCC's GC that would allow more aggressive garbage
collection.

David


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]