From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S266395AbUA3DXL (ORCPT ); Thu, 29 Jan 2004 22:23:11 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S266416AbUA3DXL (ORCPT ); Thu, 29 Jan 2004 22:23:11 -0500 Received: from trantor.dso.org.sg ([192.190.204.1]:34240 "EHLO trantor.dso.org.sg") by vger.kernel.org with ESMTP id S266395AbUA3DXJ (ORCPT ); Thu, 29 Jan 2004 22:23:09 -0500 Message-ID: <4019CDBC.3060900@starhub.net.sg> Date: Fri, 30 Jan 2004 11:21:32 +0800 From: R CHAN User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.6) Gecko/20040116 X-Accept-Language: en-us, en MIME-Version: 1.0 To: linux-kernel@vger.kernel.org Subject: [CRYPTO]: Miscompiling sha256.c by gcc 3.2.3 and arch pentium3,4 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org 2.6.2-rc2 sha256.c is miscompiled with gcc 3.2.3. (User space is RedHat 3ES.) Just an observation: gcc 3.2.3 is miscompiling sha256.c when using -O2 -march=pentium3 or pentium4. gcc 3.3.x is ok, or the problem disappears if I use arch=i686 or reduce the optimisation level to -O2. Sympthoms are all the sha256 test vectors fail. If I extract the guts of the file to compile in user space the same problem occurs.