From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755440Ab2AXQ0P (ORCPT ); Tue, 24 Jan 2012 11:26:15 -0500 Received: from casper.infradead.org ([85.118.1.10]:45709 "EHLO casper.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755204Ab2AXQ0N convert rfc822-to-8bit (ORCPT ); Tue, 24 Jan 2012 11:26:13 -0500 Message-ID: <1327422364.2446.75.camel@twins> Subject: Re: [RFCv3 0/9] perf tool: parser generator for events parsing From: Peter Zijlstra To: Jiri Olsa Cc: acme@redhat.com, mingo@elte.hu, paulus@samba.org, cjashfor@linux.vnet.ibm.com, linux-kernel@vger.kernel.org, Greg KH Date: Tue, 24 Jan 2012 17:26:04 +0100 In-Reply-To: <1326717103-10287-1-git-send-email-jolsa@redhat.com> References: <20120109152855.GA1597@m.brq.redhat.com>--suppress-cc=author> <1326717103-10287-1-git-send-email-jolsa@redhat.com> Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7BIT X-Mailer: Evolution 3.2.1- Mime-Version: 1.0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Mon, 2012-01-16 at 13:31 +0100, Jiri Olsa wrote: > Here's new version of event parsing patchset. To summarize: > - adding event parser bison/flex generator > - the event syntax stays as it was > - adding event format file sysfs attribute > for pmu device Added Greg to CC, please also see: https://lkml.org/lkml/2012/1/16/148 Right, lets hope we can convince the sysfs people that a few lines are ok.. if not we need to find another way to represent all this muck, one alternative is exploding the whole thing into a sysfs hierarchy like: format/event: config:0-7 ... format/config: config:0-63 format/config1: config1:0-63 format/config2: config2:0-63 that will of course consume loads more resources but if that's what it takes :-( > - adding new syntax to specify raw events based > on the sysfs pmu's format attribute > eg. "cpu/config=1,config1=2,config2=3/u" fair enough, so we have the attributes from the $FORMAT thingy and (for now) one hardcoded 'period' thing. Should we worry about namespace collisions for the built-in vs format provided names? I suppose that once we get all this in (and it looks good from my pov), we can start thinking about doing things like: events/cycles: event=0x3c Which together with: format/event: config:0-7 would allow us to write something like: cpu/event=cycles/u