From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:42845) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1bjrtS-0004nx-G3 for qemu-devel@nongnu.org; Tue, 13 Sep 2016 13:53:15 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1bjrtP-0007YM-Pn for qemu-devel@nongnu.org; Tue, 13 Sep 2016 13:53:14 -0400 Received: from mx1.redhat.com ([209.132.183.28]:45162) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1bjrtP-0007Y6-KK for qemu-devel@nongnu.org; Tue, 13 Sep 2016 13:53:11 -0400 Received: from int-mx10.intmail.prod.int.phx2.redhat.com (int-mx10.intmail.prod.int.phx2.redhat.com [10.5.11.23]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mx1.redhat.com (Postfix) with ESMTPS id 2C5DE4A547 for ; Tue, 13 Sep 2016 17:53:11 +0000 (UTC) Date: Tue, 13 Sep 2016 18:53:08 +0100 From: "Daniel P. Berrange" Message-ID: <20160913175307.GE30949@redhat.com> Reply-To: "Daniel P. Berrange" References: <1470756748-18933-1-git-send-email-berrange@redhat.com> <87vay6ed9t.fsf@fimbulvetr.bsc.es> <20160908134305.GL30602@redhat.com> <87y431bai0.fsf@fimbulvetr.bsc.es> <20160909120838.GJ25802@redhat.com> <87bmzxb4cd.fsf@fimbulvetr.bsc.es> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <87bmzxb4cd.fsf@fimbulvetr.bsc.es> Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [PATCH for-2.8 v1 00/60] Modular build of trace event files List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org, Stefan Hajnoczi On Fri, Sep 09, 2016 at 03:16:50PM +0200, Llu=C3=ADs Vilanova wrote: > > The various _DSTATE variables are still arbitrarily scattered in > > memory, as opposed to in a contiguous cache friendly array, which > > is one of the goals of Paolo's original change. >=20 > > That said, I'm unclear on how much of the performance win from > > Paolo's change came from eliminating the struct field de-reference, > > vs having the contiguous array. I'm guessing most of the win is > > from the former, the latter only being important if we hit multiple > > related tracepoints on close succession. >=20 > The latter can also be achieved by packing them all together in a singl= e > section, but I don't know if that's acceptable within portable QEMU cod= e. For > example: >=20 > bool ___TRACE_EVENTNAME_DSTATE __attribute__((section("dstate_array")= )) Acutally, it should be sufficient if we just generate all the dstate variables in one place in the .c file - the compiler isn't going to scatter a set of variables declared one after each other. Regards, Daniel --=20 |: http://berrange.com -o- http://www.flickr.com/photos/dberrange= / :| |: http://libvirt.org -o- http://virt-manager.or= g :| |: http://autobuild.org -o- http://search.cpan.org/~danberr= / :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vn= c :|