[Bug 54979] Futexes which protect memory allocation congest, causing 100x performance degradation

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=54979

--- Comment #4 from John Nagle  ---
Any progress on this?

What usually sets this off is growing an array. That calls realloc, and
attempts to grow the buffer in place. If that's not possible due to a lack of
free space after the buffer, the realloc code allocates a new buffer, copies
the data, and releases the old buffer. The bottleneck seems to be at the point
where the old buffer is released. That's somewhat unexpected.

A test case might involve a large number of compute-bound threads, all growing
arrays slowly.