From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from eggs.gnu.org ([2001:4830:134:3::10]:52779)
	by lists.gnu.org with esmtp (Exim 4.71)
	(envelope-from <rth7680@gmail.com>) id 1byhlP-0001pK-2l
	for qemu-devel@nongnu.org; Mon, 24 Oct 2016 12:06:15 -0400
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
	(envelope-from <rth7680@gmail.com>) id 1byhlL-0000Ns-UO
	for qemu-devel@nongnu.org; Mon, 24 Oct 2016 12:06:15 -0400
Sender: Richard Henderson <rth7680@gmail.com>
References: <1477288523-10819-1-git-send-email-vijay.kilari@gmail.com>
	<1477288523-10819-4-git-send-email-vijay.kilari@gmail.com>
	<538a5505-fef3-fc2b-4411-bbaa8537c9c7@redhat.com>
From: Richard Henderson <rth@twiddle.net>
Message-ID: <cbdd5daf-41db-a897-8bbd-9558adb90011@twiddle.net>
Date: Mon, 24 Oct 2016 08:51:52 -0700
MIME-Version: 1.0
In-Reply-To: <538a5505-fef3-fc2b-4411-bbaa8537c9c7@redhat.com>
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
Subject: Re: [Qemu-devel] [PATCH v3 3/3] utils: Add prefetch for Thunderx
 platform
List-Id: <qemu-devel.nongnu.org>
List-Unsubscribe: <https://lists.nongnu.org/mailman/options/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.nongnu.org/archive/html/qemu-devel/>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <https://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: Paolo Bonzini <pbonzini@redhat.com>, vijay.kilari@gmail.com, qemu-arm@nongnu.org, peter.maydell@linaro.org
Cc: qemu-devel@nongnu.org, Vijaya Kumar K <Vijaya.Kumar@cavium.com>

On 10/24/2016 04:25 AM, Paolo Bonzini wrote:
>> >          for (; p + 8 <= e; p += 8) {
>> > -            __builtin_prefetch(p + 8, 0, 0);
>> > +            __builtin_prefetch(p +
>> > +               (8 * cache_line_factor * prefetch_line_dist), 0, 0);
> You should precompute cache_line_bytes * prefetch_line_dist /
> sizeof(uint64_t) in a single variable, prefetch_distance.  This saves
> the effort of loading global variables repeatedly.  Then you can do
> 
>     __builtin_prefetch(p + prefetch_distance, 0, 0);
> 

Let's not complicate things by dividing by sizeof(uint64_t).
It's less complicated to avoid both that and the implied multiply.

  __builtin_prefetch((char *)p + prefetch_distance, 0, 0)


r~