[PATCH] malloc: Enable merging of remainders in memalign (bug 30723)

[DJ Delorie]

"Zack Weinberg" <zack@owlfolio.org> writes:
> but could we have a more aggressive consolidation pass, capable of
> pulling stuff back out of the fastbins, that was triggered at least
> some fraction of the times we think we need to expand the heap?

Yes, except:

1. We don't currently keep track of the status of fastbins (how many
   chunks, now much memory) to have as a metric, and

2. We don't have a good bit of logic to decide which API call will be
   interrupted to do the consolidation.

In this day and age, I'd suggest a separate consolidation thread, but
that would mean we'd have to review our atomics etc and make sure there
weren't any unneeded delays or locks.