Re: [RFC PATCH 00/11] pipe: Notification queue preparation [ver #3]

[Linus Torvalds]

On Fri, Nov 1, 2019 at 3:05 PM David Howells <dhowells@xxxxxxxxxx> wrote:
>
> Changing those to non-sync:

Your benchmark seems very insensitive to just about any changes.

I suspect it is because you only test throughput. Latency is what the
pipe wakeup has been optimized for, and which tends to be much more
sensitive to other changes too (eg locking).

That said, I'm not convinced a latency test would show much either.

               Linus