[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Tue, 12 Dec 2017 09:29:08 -0800
From: Tejun Heo <tj@...nel.org>
To: Nikolay Borisov <nborisov@...e.com>
Cc: axboe@...nel.dk, linux-kernel@...r.kernel.org, oleg@...hat.com,
peterz@...radead.org, kernel-team@...com, osandov@...com
Subject: Re: [PATCH 6/6] blk-mq: remove REQ_ATOM_STARTED
Hello, Nikolay.
On Tue, Dec 12, 2017 at 01:17:52PM +0200, Nikolay Borisov wrote:
> On 9.12.2017 21:25, Tejun Heo wrote:
> > After the recent updates to use generation number and state based
> > synchronization, we can easily replace REQ_ATOM_STARTED usages by
> > adding an extra state to distinguish completed but not yet freed
> > state.
> >
> > Add MQ_RQ_COMPLETE and replace REQ_ATOM_STARTED usages with
> > blk_mq_rq_state() tests. REQ_ATOM_STARTED no longer has any users
> > left and is removed.
>
> Where are the promised in patch 5/6 performance results?
Opos, I thought I removed all of those. I couldn't reliably show that
this performed better. I was testing with nullblk but the run-to-run
deviations were too great (they generally kept getting faster, maybe
better locality?) to draw a reliable conclusion. Whatever difference
in performance is unlikely to be material in actual workloads anyway.
I dropped the sentence from the description.
Thanks.
--
tejun
Powered by blists - more mailing lists