Re: [RFC PATCH] wfcqueue: implement __wfcq_enqueue_head() (v2)

[Eric Wong <normalperson@yhbt.net>]

Mathieu Desnoyers <mathieu.desnoyers@efficios.com> wrote:
> Implement enqueue-to-head. It can run concurrently with enqueue, splice
> to queue, and iteration, but requires a mutex against dequeue and splice
> from queue operations.
> 
> Useful for special-cases where a queue needs to have nodes enqueued into
> its head.
> 
> This patch is only compile-tested.
> 
> Changes since v1:
> * Don't require mutual exclusion between traversals and
>   __wfcq_enqueue_head().
> 
> Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

Thanks!  The first hunk (sync table comment) conflicted with
my __wfcq_enqueue patch, but other than that I could not benchmark any
regression with my 4-core machine with v4 of my
"epoll: avoid spinlock contention with wfcqueue" patch.

All I needed was "s/__wfcq_prepend/__wfcq_enqueue_head/g" to my original
patch to use the updated API.

I was worried about the cmpxchg at first, but it does not seem to hurt
performance on my 4-core system.  In fact, it was slightly better
(but within margin of error)

time ./eponeshotmt -c 1000000 -w 4 -t 4 -f 10
real    0m 5.78s
user    0m 1.20s
sys     0m 21.90s

Tested-by: Eric Wong <normalperson@yhbt.net>

Hopefully somebody can test my epoll patches with more cores/threads :)