Re: [PATCH] Make GC dependent on % of total memory consed
Marcus Crestani wrote:
>>>>>"SJT" == Stephen J Turnbull
<stephen(a)xemacs.org> writes:
>>>>>
>>>>>
SJT> I'd like to run with this for a while before making comments on
SJT> the algorithm change, but I think we should bump to
SJT> gc-cons-threshold to at least 2MB while we're at it.
Ben didn't add his changes to the new allocator's code, so you won't
get different behavior if you are using the default-enabled new
allocator. Ben, could you please add these changes to the new
allocator?
To let people know: I have a working incremental garbage collector in
my workspace. I won't have time the next three weeks, but I will
commit this workspace in the second half of October.
Ben, do you have any other GC-related patches in your queue?
Not that I can see. One of my ws's adds pdump descriptions for
face_cachels, but that's about all.
Marcus, I went and added the changes to your allocator code but
currently it depends on the code in Flrecord_stats(), which is only
conditionally defined. Is there some other way of determining at the
end of GC how much total memory is allocated?
ben