From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from mailman by lists.gnu.org with tmda-scanned (Exim 4.43)
	id 1K1NJV-0004sN-3k
	for qemu-devel@nongnu.org; Wed, 28 May 2008 11:11:41 -0400
Received: from [199.232.76.173] (port=39280 helo=monty-python.gnu.org)
	by lists.gnu.org with esmtp (Exim 4.43) id 1K1NJU-0004sF-IM
	for qemu-devel@nongnu.org; Wed, 28 May 2008 11:11:40 -0400
Received: from mx1.polytechnique.org ([129.104.30.34]:49327)
	by monty-python.gnu.org with esmtps
	(TLS-1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.60)
	(envelope-from <fabrice@bellard.org>) id 1K1NJU-0007tA-9O
	for qemu-devel@nongnu.org; Wed, 28 May 2008 11:11:40 -0400
Message-ID: <483D7625.2030807@bellard.org>
Date: Wed, 28 May 2008 17:11:33 +0200
From: Fabrice Bellard <fabrice@bellard.org>
MIME-Version: 1.0
Subject: Re: [Qemu-devel] [4600] variable dynamic translation buffer size
References: <E1K1KnZ-00018e-0T@cvs.savannah.gnu.org>
	<200805281355.53574.paul@codesourcery.com>
In-Reply-To: <200805281355.53574.paul@codesourcery.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Reply-To: qemu-devel@nongnu.org
List-Id: qemu-devel.nongnu.org
List-Unsubscribe: <http://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.gnu.org/pipermail/qemu-devel>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <http://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: qemu-devel@nongnu.org
Cc: Paul Brook <paul@codesourcery.com>

Paul Brook wrote:
> On Wednesday 28 May 2008, Fabrice Bellard wrote:
>> +        code_gen_buffer = mmap(NULL, code_gen_buffer_size,
>> +                               PROT_WRITE | PROT_READ | PROT_EXEC,
>> +                               flags, -1, 0);
> 
> I'd expect this to cause a significant performance regression on hosts that 
> have limited branch range (i.e. everything except x86). The mmap region is 
> almost never close to the application text segment, so all helper calls have 
> to be indirect.

I knew that when I did the patch, but having a dynamic size on x86 hosts 
prevails.

The other hosts can still implement hacks if they really need to, either 
by reserving a huge data in BSS as it was done before or by using the 
proc/x/maps entries to compute a possible mmap() address. IMHO I think 
this is unnecessary as helpers should be the exception.

Fabrice.