From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752562Ab1HHMZs (ORCPT ); Mon, 8 Aug 2011 08:25:48 -0400 Received: from merlin.infradead.org ([205.233.59.134]:46116 "EHLO merlin.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751583Ab1HHMZq convert rfc822-to-8bit (ORCPT ); Mon, 8 Aug 2011 08:25:46 -0400 Subject: Re: regarding [PATCH] Add libpfm4 support (v3) From: Peter Zijlstra To: Jiri Olsa Cc: Ingo Molnar , Frederic Weisbecker , Arnaldo Carvalho de Melo , linux-kernel@vger.kernel.org, Arun Sharma Date: Mon, 08 Aug 2011 14:25:25 +0200 In-Reply-To: <20110808120432.GB2076@jolsa.brq.redhat.com> References: <20110808120432.GB2076@jolsa.brq.redhat.com> Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 8BIT X-Mailer: Evolution 3.0.2- Message-ID: <1312806326.10488.30.camel@twins> Mime-Version: 1.0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Mon, 2011-08-08 at 14:04 +0200, Jiri Olsa wrote: > hi, > > any chance of getting this included? > http://article.gmane.org/gmane.comp.linux.perfmon2.devel/2912 > http://article.gmane.org/gmane.comp.linux.perfmon2.devel/2912/raw > > it needs just some little changes to fit to current tip tree While I don't mind extending the current way to input events I don't think using libpfm4 for it is desired. One of the ideas that has been floating about is to provide a format file in the sysfs representation of the pmu. This format file would describe the config space. The tool would parse this and use it to provide better input capabilities. This is not an easy task however, since the format syntax needs to be capable enough to describe all the various config muck we need to support, think for example of the P4 mess. For Intel/AMD (excluding P4) its mostly simple, with Intel having a slightly harder time for the offcore bits. As to the various arch event lists, again that is not something we need a whole library for, parsing those text files isn't hard. Maintaining the various arch event lists in a sane format is the 'hard' part and is something I think we should push to the vendors, because they're basically the only ones capable.