From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mx1.suse.de (mx1.suse.de [195.135.220.2]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client CN "mx1.suse.de", Issuer "CAcert Class 3 Root" (verified OK)) by ozlabs.org (Postfix) with ESMTPS id 562CDDDECA for ; Wed, 30 Jul 2008 17:26:17 +1000 (EST) Date: Wed, 30 Jul 2008 09:26:11 +0200 From: Nick Piggin To: Benjamin Herrenschmidt Subject: Re: [PATCH] powerpc/mm: Lockless get_user_pages_fast() Message-ID: <20080730072611.GA7986@wotan.suse.de> References: <1217389038.11188.285.camel@pasglop> <1217391656.11188.292.camel@pasglop> <1217394382.10646.13.camel@localhost> <1217394520.11188.298.camel@pasglop> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <1217394520.11188.298.camel@pasglop> Cc: linuxppc-dev list List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Wed, Jul 30, 2008 at 03:08:40PM +1000, Benjamin Herrenschmidt wrote: > On Wed, 2008-07-30 at 15:06 +1000, Michael Ellerman wrote: > > > + > > > +/* > > > + * The performance critical leaf functions are made noinline otherwise gcc > > > + * inlines everything into a single function which results in too much > > > + * register pressure. > > > + */ > > > > This strikes me as something that is liable to change for compiler > > version n+1, or n with -fsomething - and might leave us shooting > > ourselves in the foot, just a thought. > > > > Not that much I'd say... In fact, I wouldn't be too worried on powerpc, > I wonder if that comment is stale from the x86 variant :-) Nick ? Right... gcc is really poor at over pressuing registers when inlining, and when I checked I don't think it even allocated registers to the inner-most variables in cases such as this. I thought I checked powerpc and sound some spilling there too, but it was quite a long time ago (and yes it was brought over from x86). Should double check.