Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932157AbXACW3l (ORCPT ); Wed, 3 Jan 2007 17:29:41 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S932167AbXACW3l (ORCPT ); Wed, 3 Jan 2007 17:29:41 -0500 Received: from mxout.hispeed.ch ([62.2.95.247]:44993 "EHLO smtp.hispeed.ch" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S932157AbXACW3j (ORCPT ); Wed, 3 Jan 2007 17:29:39 -0500 X-Greylist: delayed 2586 seconds by postgrey-1.27 at vger.kernel.org; Wed, 03 Jan 2007 17:29:39 EST Subject: Re: kernel + gcc 4.1 = several problems From: Thomas Sailer To: Linus Torvalds Cc: Grzegorz Kulewski , Alan , Mikael Pettersson , s0348365@sms.ed.ac.uk, 76306.1226@compuserve.com, akpm@osdl.org, bunk@stusta.de, greg@kroah.com, linux-kernel@vger.kernel.org, yanmin_zhang@linux.intel.com In-Reply-To: References: <200701030212.l032CDXe015365@harpo.it.uu.se> <20070103102944.09e81786@localhost.localdomain> Content-Type: text/plain Date: Wed, 03 Jan 2007 22:44:19 +0100 Message-Id: <1167860659.13394.41.camel@unreal> Mime-Version: 1.0 X-Mailer: Evolution 2.8.2.1 (2.8.2.1-2.fc6) Content-Transfer-Encoding: 7bit X-DCC-spamcheck-02.tornado.cablecom.ch-Metrics: smtp-06.tornado.cablecom.ch 1378; Body=11 Fuz1=11 Fuz2=11 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 950 Lines: 21 On Wed, 2007-01-03 at 08:03 -0800, Linus Torvalds wrote: > and assuming the branch is AT ALL predictable (and 95+% of all branches > are), the branch-over will actually be a LOT better for a CPU. IF... Counterexample: Add-Compare-Select in a Viterbi Decoder. If the compare can be predicted, you botched the compression of the data (if you can predict the data, you could have compressed it better), or your noise is not white, i.e. you f*** up the whitening filter. So in any practical viterbi decoder, the compares cannot be predicted. I remember cmov made a big difference in Viterbi Decoder performance on a Cyrix 6x86. But granted, nowadays these things are usually done with SIMD and masks. Tom - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/