[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-rdma
Subject:    Re: [PATCH 04/18] ib_srp: Micro-optimize completion handlers
From:       David Dillow <dillowda () ornl ! gov>
Date:       2012-02-26 6:32:14
Message-ID: 1330237934.1026.82.camel () obelisk ! thedillows ! org
[Download RAW message or body]

On Sat, 2012-01-14 at 12:42 +0000, Bart Van Assche wrote:
> Reduce completion queue lock contention by polling for multiple
> work completions at once. Help the CPU branch predictor by making
> it clear that IB_WC_SUCCESS is the most likely case. Move the
> error handling code into the new function srp_handle_qp_err().
> 
> Also, convert srp_target_port.qp_in_error from int to bool and
> move the initialization of that variable into srp_connect_target().

Don't merge the two different changes here. Introducing
srp_handle_qp_err() is fine, but the change to polling for multiple WC
is completely unrelated to that and to the stated intent of this series.
It also hurt performance when I measured it during the IOP scaling work.

Let's get this series out of the way, then we can talk about the WC
changes.
-- 
Dave Dillow
National Center for Computational Science
Oak Ridge National Laboratory
(865) 241-6602 office

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
[prev in list] [next in list] [prev in thread] [next in thread]