From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1760513AbYHHWbI (ORCPT ); Fri, 8 Aug 2008 18:31:08 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1764748AbYHHWaj (ORCPT ); Fri, 8 Aug 2008 18:30:39 -0400 Received: from an-out-0708.google.com ([209.85.132.242]:1805 "EHLO an-out-0708.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1764695AbYHHWah (ORCPT ); Fri, 8 Aug 2008 18:30:37 -0400 DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=message-id:date:from:to:subject:cc:in-reply-to:mime-version :content-type:content-transfer-encoding:content-disposition :references; b=S9t1lUQpN9U/XnotyegrZ3zREd2pGb/Z5MG7+bCvrWO7rtY1n4fMK8kFRdcpQwSymK igv43CyOTFaDYEqp91pVwxWbKSQFQcgzD/+trj/qS6f2UksAfvof2dDY5gM9nQQvPOwx 1U/fnzEFOhnPZsNPGoXRLejjCAb91e4I8uKh0= Message-ID: <86802c440808081530o6c07c3bbl1a52f7a267dc2936@mail.gmail.com> Date: Fri, 8 Aug 2008 15:30:35 -0700 From: "Yinghai Lu" To: "H. Peter Anvin" Subject: Re: [PATCH 00/42] dyn_array/nr_irqs/sparse_irq support v5 Cc: "Ingo Molnar" , "Thomas Gleixner" , "Eric W. Biederman" , "Dhaval Giani" , "Mike Travis" , "Andrew Morton" , linux-kernel@vger.kernel.org In-Reply-To: <489CC7E2.7040209@zytor.com> MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Content-Disposition: inline References: <1218232368-31228-1-git-send-email-yhlu.kernel@gmail.com> <489CC246.5000202@zytor.com> <86802c440808081514w3a7ea1d6hda2442e7bf6729a4@mail.gmail.com> <489CC7E2.7040209@zytor.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, Aug 8, 2008 at 3:25 PM, H. Peter Anvin wrote: > Yinghai Lu wrote: >>>> >>> So I'm still clearly missing something about this... if we need sparse >>> IRQs >>> in the first place (which we do), what's the point of the dyn_array? >> >> x86_64: support CONFIG_HAVE_SPARSE_IRQ and CONFIG_DYN_ARRAY >> x86_32: support CONFIG_DYN_ARRAY >> >> some arches could use dyn_array with probing nr_irqs and it could be >> 32 and much less than 224. >> some could have that like 512. and those arch may not need to mess up >> with sparse_irq at first point. >> but still could get some flexibilty about that array size. >> > > As an x86 maintainer, I definitely do not want x86-64 and x86-32 to diverge > unless there is an extremely strong reason to. > > Other architectures may speak for themselves, but why not just support > sparse IRQs on x86-32 *and* -64 and skip the dyn_array variant? after we merged io_apic_32.c into io_apic_64.c. also I want to kill irq balance in io_apic_32.c, but no one say anything about it. also dyn_array could have other user in addition to nr_irqs. i will dig it out like NR_CPUS/nr_cpu_ids related array. and Mike tried to put every thing to PER_CPU instead of array, maybe some case array would be effient than that. that make dyn_array some usage. YH