From mboxrd@z Thu Jan  1 00:00:00 1970
From: Knut Petersen <Knut_Petersen@t-online.de>
Subject: Re: X11 performance regressions
Date: Wed, 11 May 2011 23:22:19 +0200
Message-ID: <4DCAFE0B.5020909@t-online.de>
References: <4DC6DF5D.2050006@t-online.de>	
	<15db10$kicfca@fmsmga002.fm.intel.com>
	<4DCAA134.4020705@t-online.de>
	<1305143375.19920.211.camel@atropine>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Return-path: <intel-gfx-bounces+gcfxdi-intel-gfx=m.gmane.org@lists.freedesktop.org>
Received: from mailout05.t-online.de (mailout05.t-online.de [194.25.134.82])
	by gabe.freedesktop.org (Postfix) with ESMTP id 7190C9E7B1
	for <intel-gfx@lists.freedesktop.org>;
	Wed, 11 May 2011 14:22:30 -0700 (PDT)
In-Reply-To: <1305143375.19920.211.camel@atropine>
List-Unsubscribe: <http://lists.freedesktop.org/mailman/options/intel-gfx>,
	<mailto:intel-gfx-request@lists.freedesktop.org?subject=unsubscribe>
List-Archive: <http://lists.freedesktop.org/archives/intel-gfx>
List-Post: <mailto:intel-gfx@lists.freedesktop.org>
List-Help: <mailto:intel-gfx-request@lists.freedesktop.org?subject=help>
List-Subscribe: <http://lists.freedesktop.org/mailman/listinfo/intel-gfx>,
	<mailto:intel-gfx-request@lists.freedesktop.org?subject=subscribe>
Sender: intel-gfx-bounces+gcfxdi-intel-gfx=m.gmane.org@lists.freedesktop.org
Errors-To: intel-gfx-bounces+gcfxdi-intel-gfx=m.gmane.org@lists.freedesktop.org
To: Adam Jackson <ajax@redhat.com>
Cc: intel-gfx <intel-gfx@lists.freedesktop.org>
List-Id: intel-gfx@lists.freedesktop.org

As I do have only a few minutes now, a few comments:

1: The complete trees are compared, all modules/libraries are either old or new. No debug-versions.

2: Speculating about cores is definitely wrong -- the Pentium M Dothan definitely is a single core cpu.

3. There often is a "choked most" (1) -- "choked least" (10) -- "choked a bit more again" (100,500)
     result:

   1450000.0    0.50   1x1 stippled rectangle (8x8 stipple)
    134000.0    1.11   10x10 stippled rectangle (8x8 stipple)
      2540.0    1.05   100x100 stippled rectangle (8x8 stipple)
       110.0    0.95   500x500 stippled rectangle (8x8 stipple)

    Heavy per call impact of factor A on those small requests, light impact of a factor B with growing numbers?
    A = compiler / library overhead?

Yes, there is
>    15400.0    0.54   GetProperty
>    15500.0    0.54   QueryPointer
but we also see
 
     8150000.0    1.21   X protocol NoOperation


4. No, it's not the kernel. I did
    a) boot
    b) x11perf on old X
    c) x11perf on new X
    d) reboot
    e) x11perf on new X
    f) x11perf on old X
    and saw only very marginal differences between those two runs.

5.  Yes,  I do agree to that:
>  I'm actually pretty pleased with the results you've shown, 10%
> or better speedup for basically all text ops, about half of window
> management ops, and almost all window exposure ops. 
6. More later.

cu,
 Knut