From owner-freebsd-performance@FreeBSD.ORG Tue Sep 27 22:36:08 2011 Return-Path: Delivered-To: freebsd-performance@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id B22A01065670 for ; Tue, 27 Sep 2011 22:36:08 +0000 (UTC) (envelope-from gnn@freebsd.org) Received: from vps.hungerhost.com (vps.hungerhost.com [216.38.53.176]) by mx1.freebsd.org (Postfix) with ESMTP id 827F38FC0A for ; Tue, 27 Sep 2011 22:36:08 +0000 (UTC) Received: from [209.249.190.124] (helo=gnnmac.hudson-trading.com) by vps.hungerhost.com with esmtpsa (TLSv1:AES128-SHA:128) (Exim 4.69) (envelope-from ) id 1R8fS7-0006fQ-Hn; Tue, 27 Sep 2011 17:44:35 -0400 Mime-Version: 1.0 (Apple Message framework v1244.3) Content-Type: text/plain; charset=us-ascii From: George Neville-Neil In-Reply-To: Date: Tue, 27 Sep 2011 17:44:42 -0400 Content-Transfer-Encoding: quoted-printable Message-Id: <3BE45FDD-1A7A-4F9D-B32B-30D19C897112@freebsd.org> References: To: Anton Yuzhaninov X-Mailer: Apple Mail (2.1244.3) X-AntiAbuse: This header was added to track abuse, please include it with any abuse report X-AntiAbuse: Primary Hostname - vps.hungerhost.com X-AntiAbuse: Original Domain - freebsd.org X-AntiAbuse: Originator/Caller UID/GID - [47 12] / [47 12] X-AntiAbuse: Sender Address Domain - freebsd.org Cc: freebsd-performance@freebsd.org Subject: Re: hwpmc overflow X-BeenThere: freebsd-performance@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Performance/tuning List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 27 Sep 2011 22:36:08 -0000 On Sep 21, 2011, at 01:48 , Anton Yuzhaninov wrote: > On Wed, 21 Sep 2011 09:27:33 +0800, Adrian Chadd wrote: > AC> Is this an AMD class CPU? > AC>=20 > AC> The way the counters work on some platforms is that they count = _up_ to > AC> an overflow value, trigger the overflow interrupt, and PMC then is > AC> supposed to calculate what the original value was. > AC> It sounds like maybe that isn't working correctly for a certain = class > AC> of counters? >=20 > CPU: Intel(R) Xeon(R) CPU E5645 @ 2.40GHz >=20 If you count them individually do you still get the overflows? That is, = if you run the same workload but only count instructions, then run it again with = the next counter, etc. I am wondering if something is wrong because of the number of = things you're trying to count at the same time. Best, George From owner-freebsd-performance@FreeBSD.ORG Wed Sep 28 10:06:14 2011 Return-Path: Delivered-To: freebsd-performance@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id B99CE106566C for ; Wed, 28 Sep 2011 10:06:14 +0000 (UTC) (envelope-from gofp-freebsd-performance@m.gmane.org) Received: from lo.gmane.org (lo.gmane.org [80.91.229.12]) by mx1.freebsd.org (Postfix) with ESMTP id D64658FC0C for ; Wed, 28 Sep 2011 10:06:13 +0000 (UTC) Received: from list by lo.gmane.org with local (Exim 4.69) (envelope-from ) id 1R8r1o-0004l9-Dk for freebsd-performance@freebsd.org; Wed, 28 Sep 2011 12:06:12 +0200 Received: from 178.214.36.169 ([178.214.36.169]) by main.gmane.org with esmtp (Gmexim 0.1 (Debian)) id 1AlnuQ-0007hv-00 for ; Wed, 28 Sep 2011 12:06:12 +0200 Received: from citrin by 178.214.36.169 with local (Gmexim 0.1 (Debian)) id 1AlnuQ-0007hv-00 for ; Wed, 28 Sep 2011 12:06:12 +0200 X-Injected-Via-Gmane: http://gmane.org/ To: freebsd-performance@freebsd.org From: Anton Yuzhaninov Date: Wed, 28 Sep 2011 10:05:58 +0000 (UTC) Organization: Vega Lines: 35 Sender: Anton Yuzhaninov Message-ID: References: <3BE45FDD-1A7A-4F9D-B32B-30D19C897112@freebsd.org> X-Complaints-To: usenet@dough.gmane.org X-Gmane-NNTP-Posting-Host: 178.214.36.169 X-Comment-To: George Neville-Neil User-Agent: tin/1.9.6-20101126 ("Burnside") (UNIX) (FreeBSD/9.0-BETA2 (i386)) X-Mailman-Approved-At: Wed, 28 Sep 2011 10:53:46 +0000 Subject: Re: hwpmc overflow X-BeenThere: freebsd-performance@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Performance/tuning List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 28 Sep 2011 10:06:14 -0000 On Wed, 28 Sep 2011 01:44:42, George Neville-Neil wrote: GNN> If you count them individually do you still get the overflows? That is, if you run GNN> the same workload but only count instructions, then run it again with the next counter, GNN> etc. I am wondering if something is wrong because of the number of things you're trying GNN> to count at the same time. for one counter problem the same. This test on other hardware: CPU: Intel(R) Xeon(R) CPU E5335 @ 2.00GHz (2004.98-MHz K8-class CPU) hwpmc: TSC/1/64/0x20 IAP/2/40/0x3ff IAF/3/40/0x61 Counter for DTLB_MISSES.ANY should overflow at 2^40 - 1 But again I see overflow near 2^31 # pmcstat -C -w1 -d -p DTLB_MISSES.ANY -o pmc.log -t 10924 from log: 2022802283 2041525661 2061836744 2083646634 2103665903 2123998093 2142902046 1097377998076 1097397893694 1097424659734 1097446393399 1097469768564 1097495460707 1097519426191 -- Anton Yuzhaninov