From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:33377) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1W6ybZ-0006Li-40 for qemu-devel@nongnu.org; Sat, 25 Jan 2014 03:28:49 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1W6ybO-0005AK-Q6 for qemu-devel@nongnu.org; Sat, 25 Jan 2014 03:28:41 -0500 Received: from mail-qc0-x22b.google.com ([2607:f8b0:400d:c01::22b]:36570) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1W6ybO-0005A1-L7 for qemu-devel@nongnu.org; Sat, 25 Jan 2014 03:28:30 -0500 Received: by mail-qc0-f171.google.com with SMTP id n7so5614900qcx.2 for ; Sat, 25 Jan 2014 00:28:30 -0800 (PST) Sender: Paolo Bonzini Message-ID: <52E375A8.1090703@redhat.com> Date: Sat, 25 Jan 2014 09:28:24 +0100 From: Paolo Bonzini MIME-Version: 1.0 References: <1390506565-8880-1-git-send-email-trent.tong@gmail.com> <87bnz14oag.fsf@linaro.org> In-Reply-To: Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 8bit Subject: Re: [Qemu-devel] [PATCH v2] cpu: implementing victim TLB for QEMUsystem emulated TLBB List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: BALATON Zoltan , =?UTF-8?B?QWxleCBCZW5uw6ll?= Cc: Richard Henderson , Xin Tong , QEMU Developers , afaerber@suse.de Il 24/01/2014 23:32, BALATON Zoltan ha scritto: > On Fri, 24 Jan 2014, Alex Bennée wrote: >> trent.tong@gmail.com writes: >>> Attaching data in excel which could not be sent with the patch at the >>> same time. >>> >> >> >> If you can attach the summary of the data as plain text that would be >> useful. Not all of us have access to a Windows box with Excell! > > Opens on LibreOffice as well but plain text is easier to read. > > I wonder how much of the measured performance increase is due to the > increased effective cache size and how much is the reduction in number > of instructions by the victim TLB mechanism. Wouldn't it be a more fair > measurement to use a bigger TLB size for the TLB only case that matches > the effective size of the TLB+victim case instead of using the same TLB > size for both? That's quite unlikely, the TLB size is 256 while TLB+victim is 264. Paolo