On Wed, May 25, 2016 at 01:17:21 +0300, Sergey Fedorov wrote:
> >> With this implementation we could:
> >> (1) get rid of qht_map::stale
> >> (2) don't waste cycles waiting for resize to complete
> > I'll include this in v6.
>
> How is it by perf?
Not much of a difference, since resize is a slow path. Calling
qht-bench with lots of update and resize threads performs very
poorly either way =D
I like the change though because using the mutex here simplifies
the resize code; there's no guilt anymore attached to holding the lock
for some time (e.g. when allocating a new, possible quite large,
map), whereas with the spinlock we would allocate it before
acquiring the lock, without knowing whether the allocation would
be needed in the end.
Emilio
