From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 101226 invoked by alias); 18 May 2015 21:48:55 -0000 Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Archive: List-Post: List-Help: Sender: gcc-patches-owner@gcc.gnu.org Received: (qmail 100529 invoked by uid 89); 18 May 2015 21:48:55 -0000 Authentication-Results: sourceware.org; auth=none X-Virus-Found: No X-Spam-SWARE-Status: No, score=-2.3 required=5.0 tests=AWL,BAYES_00,KAM_LAZY_DOMAIN_SECURITY,SPF_HELO_PASS,T_RP_MATCHES_RCVD autolearn=no version=3.3.2 X-HELO: mx1.redhat.com Received: from mx1.redhat.com (HELO mx1.redhat.com) (209.132.183.28) by sourceware.org (qpsmtpd/0.93/v0.84-503-g423c35a) with (AES256-GCM-SHA384 encrypted) ESMTPS; Mon, 18 May 2015 21:48:54 +0000 Received: from int-mx11.intmail.prod.int.phx2.redhat.com (int-mx11.intmail.prod.int.phx2.redhat.com [10.5.11.24]) by mx1.redhat.com (8.14.4/8.14.4) with ESMTP id t4ILme8Z016388 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-GCM-SHA384 bits=256 verify=FAIL); Mon, 18 May 2015 17:48:40 -0400 Received: from [10.36.5.53] (vpn1-5-53.ams2.redhat.com [10.36.5.53]) by int-mx11.intmail.prod.int.phx2.redhat.com (8.14.4/8.14.4) with ESMTP id t4ILmcRK011245; Mon, 18 May 2015 17:48:39 -0400 Subject: Re: [PATCH] add self-tuning to x86 hardware fast path in libitm From: Torvald Riegel To: Andi Kleen Cc: Nuno Diegues , gcc-patches@gcc.gnu.org, Paolo Romano In-Reply-To: <20150518213953.GE19417@two.firstfloor.org> References: <87y4m2ogvi.fsf@tassilo.jf.intel.com> <20150408175457.GR2366@two.firstfloor.org> <87iod4npcv.fsf@tassilo.jf.intel.com> <1431984546.8533.90.camel@triegel.csb> <20150518213953.GE19417@two.firstfloor.org> Content-Type: text/plain; charset="UTF-8" Date: Mon, 18 May 2015 22:04:00 -0000 Message-ID: <1431985717.8533.100.camel@triegel.csb> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit X-IsSubscribed: yes X-SW-Source: 2015-05/txt/msg01630.txt.bz2 On Mon, 2015-05-18 at 23:39 +0200, Andi Kleen wrote: > > Are there better options for the utility function, or can we tune it to > > There is nothing better that isn't a lot slower. Do you care to elaborate why? As-is, I find this statement to not be convincing; at the very least we need to document why we think that something time-based is the best option. Other tuning attempts have looked at rates of aborted, attempted, and committed txns, for example. Why do we measure nb. of transactions in a whole period, and can't get the same info through measuring smaller but more specific time intervals (e.g., how long we wait for a serial txn)?