From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from gate.crashing.org (gate.crashing.org [63.228.1.57]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client did not present a certificate) by ozlabs.org (Postfix) with ESMTP id BBB2D679F8 for ; Sat, 21 May 2005 08:29:13 +1000 (EST) From: Benjamin Herrenschmidt To: Dan Malek In-Reply-To: References: <17037.32285.607543.859030@cargo.ozlabs.ibm.com> <1116575115.5153.115.camel@gaston> Content-Type: text/plain Date: Sat, 21 May 2005 08:27:32 +1000 Message-Id: <1116628052.5153.123.camel@gaston> Mime-Version: 1.0 Cc: linuxppc-dev@ozlabs.org Subject: Re: [PATCH] Set cpu type explicitly, take 2 List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Fri, 2005-05-20 at 10:30 -0400, Dan Malek wrote: > On May 20, 2005, at 3:45 AM, Benjamin Herrenschmidt wrote: > > > The problem is the same as ppc64 isn't it ?, -maltivec doesn't work > > with > > -mcpu=power4 on gcc 3.4 does it ? That would mean no RAID6... > > I don't understand "doesn't work." Does power4 imply altivec? > Do we need to add -maltivec based on the configuration option? No, -mcpu=970 is equivalent in a way to -mcpu=power4 and -maltivec, howevever, there is a bug in gcc 3.4 which makes -maltivec cause an error if you have -mcpu=power4. The RAID6 code needs -maltivec. So on gcc 3.4, we need to use -mcpu=970 instead of -mcpu=power4. However, we must _NOT_ do that on gcc-4.0 as that causes it to implicitely generate altivec code all over the place. (Fortunately, the bug is also fixed there, so with gcc-4.0, we just need -mcpu=power4 and -maltivec only on the RAID6 code) Messy ? Yeah ! Ben.