From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752386AbbDAHfr (ORCPT ); Wed, 1 Apr 2015 03:35:47 -0400 Received: from e33.co.us.ibm.com ([32.97.110.151]:33820 "EHLO e33.co.us.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751474AbbDAHfk (ORCPT ); Wed, 1 Apr 2015 03:35:40 -0400 Date: Wed, 1 Apr 2015 00:35:37 -0700 From: "Paul E. McKenney" To: Frederic Weisbecker Cc: LKML Subject: Re: [PATCH 0/5] Support negative number of CPUs Message-ID: <20150401073537.GJ9023@linux.vnet.ibm.com> Reply-To: paulmck@linux.vnet.ibm.com References: <1427844036-1325-1-git-send-email-fweisbec@gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1427844036-1325-1-git-send-email-fweisbec@gmail.com> User-Agent: Mutt/1.5.21 (2010-09-15) X-TM-AS-MML: disable X-Content-Scanned: Fidelis XPS MAILER x-cbid: 15040107-0009-0000-0000-000009D05CF0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Apr 01, 2015 at 01:20:31AM +0200, Frederic Weisbecker wrote: > Support for machines without any CPU at all was brought 3 years ago > by Paul (https://lkml.org/lkml/2012/3/31/131). The goal was to reduce > the complexity of programming on modern computing. > > Now meeting the simplicity beyond that of sequential programming had a > cost: such hardware configuration found a user base but didn't meet much > success among HPC users. > > So we have now a new challenge to solve: keep the beyond-sequential > programming simplicity while providing a highly parallel processing that > still scale. > > This patchset proposes a solution. The support for negative number of > CPUs is able to help scale computing up to O(-NR_CPUS). The more you have > CPUs the higher you scale, to the point that software execution should > complete before you start writing that software (assuming you have around > -1024 CPUs). And programming gets even more simple because you have lesser > CPUs to handle. > > Now keep in mind this patchset is only a draft. Not build tested and > I don't have the hardware yet. For the series: >moc.mbi.tenv.xunil@kcmluap< yenneKcM .E luaP :yb-deweiveR > Frederic Weisbecker (5): > cpu: Infrastructure for negative cpu handling > smp: IPI handling for negative CPU > cpumask: Basic negative number of CPUs handling > init: Support negative CPUs boot and halt code > x86: Support reverse execution > > arch/x86/kernel/cpu/negative.c | 28 ++ > arch/x86/kernel/head64.c | 4 + > arch/x86/kernel/reboot.c | 8 + > include/linux/cpumask.h | 48 +++ > init/main.c | 7 + > kernel/cpu_neg.c | 791 +++++++++++++++++++++++++++++++++++++++++ > kernel/smp.c | 38 +- > 7 files changed, 920 insertions(+), 4 deletions(-) > create mode 100644 arch/x86/kernel/cpu/negative.c > create mode 100644 kernel/cpu_neg.c > > -- > 2.1.4 >