On 11/1/2017 7:26 PM, Sagi Grimberg wrote:

quoted

(4) Send signal with invalidation signal w/o moderation 3.4M / 4.6M

What about remote invalidation? That is how it will usually be used...

We haven't tested remote invalidation yet.

quoted

(5) Send signal with invalidation signal w/ moderation 6.2M / 11.8M

What about latency effects? Any noticeable change?

quoted

I understand that (1), (2) and (3) are invalid, the performance 
reduction induce by mode (4) need to motivate us to enable cqe/event 
moderation as soon as possible.

Well good, it means you can help test my adaptive coalescing code :)

Of course.

May I ask what moderation parameters did you use? it will be useful
to pin down the profile levels the adaptive moderator would bounce
between.

Sure, I mentioned the params earlier (cq_completions=32 cq_timeout=16).

Sagi,
As we can see from our results, the changes made in this patchset must 
be pushed together with some cq moderation (adaptive or manual option 
meanwhile). In the testing, I created a module parameter but we can 
create a different option for the user to configure.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html