From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [patch 0/3] Allow inlined spinlocks again V3 Date: Fri, 14 Aug 2009 15:23:58 -0700 (PDT) Message-ID: <20090814.152358.16603090.davem@davemloft.net> References: <20090814.131900.229343660.davem@davemloft.net> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Return-path: Received: from 74-93-104-97-Washington.hfc.comcastbusiness.net ([74.93.104.97]:48435 "EHLO sunset.davemloft.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753215AbZHNWXs (ORCPT ); Fri, 14 Aug 2009 18:23:48 -0400 In-Reply-To: Sender: linux-arch-owner@vger.kernel.org List-ID: To: torvalds@linux-foundation.org Cc: heiko.carstens@de.ibm.com, akpm@linux-foundation.org, a.p.zijlstra@chello.nl, mingo@elte.hu, linux-arch@vger.kernel.org, schwidefsky@de.ibm.com, arnd@arndb.de, horsth@linux.vnet.ibm.com, ehrhardt@linux.vnet.ibm.com From: Linus Torvalds Date: Fri, 14 Aug 2009 14:10:45 -0700 (PDT) > On Fri, 14 Aug 2009, Linus Torvalds wrote: >> >> But not inline in the code, though. So yeah, it has a memory footprint, >> but shouldn't have a cache footprint. > > An example of this: stack usage. This is the code with BUG_ON(): ... > and note how in the code, we're just jumping over something like four > bytes ("ud2" plus that silly endless loop-jump just to make gcc happy > are both 2 bytes on x86[-64]). > > Here's the same code with that horrid skb_under_panic(): Yes on x86 it's better even with verbose debugging because of the inlining of the debug message pointer etc. I'll have to do something more intelligent than I currently do on sparc64, and I had always been meaning to do that. :-)