Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Mon, 10 Feb 2020 14:35:46 -0800
From:   Andrew Morton <akpm@linux-foundation.org>
To:     Davidlohr Bueso <dave@stgolabs.net>
Cc:     linux-kernel@vger.kernel.org, mcgrof@kernel.org,
        broonie@kernel.org, alex.williamson@redhat.com
Subject: Re: [PATCH -next 0/5] rbtree: optimize frequent tree walks
Message-Id: <20200210143546.4491d9715f1c4a0a1de999ca@linux-foundation.org>
In-Reply-To: <20200210155611.lfrddnolsyzktqne@linux-p48b>
References: <20200207180305.11092-1-dave@stgolabs.net>
        <20200209174632.9c7b6ff20567853c05e5ee58@linux-foundation.org>
        <20200210155611.lfrddnolsyzktqne@linux-p48b>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Mon, 10 Feb 2020 07:56:11 -0800 Davidlohr Bueso <dave@stgolabs.net> wrote:

> On Sun, 09 Feb 2020, Andrew Morton wrote:
> >Seems that all the caller sites you've converted use a fairly small
> >number of rbnodes, so the additional storage shouldn't be a big
> >problem.  Are there any other sites you're eyeing?  If so, do you expect
> >any of those will use a significant amount of memory for the nodes?
> 
> I also thought about converting the deadline scheduler to use these,
> mainly benefiting pull_dl_task() but didn't get to it and I don't expect
> the extra footprint to be prohibitive.
> 
> >
> >And...  are these patches really worth merging?  Complexity is added,
> >but what end-user benefit can we expect?
> 
> Yes they are worth merging, imo (which of course is biased :)
> 
> I don't think there is too much added complexity overall, particularly
> considering that the user conversions are rather trivial. And even for
> small trees (ie 100 nodes) we still benefit in a measurable way from
> these optimizations.
> 

Measurable for microbenchmarks, I think?  But what benefit will a user
see, running a workload that is cared about?