From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from smtp1.linuxfoundation.org (smtp1.linux-foundation.org [172.17.192.35]) by mail.linuxfoundation.org (Postfix) with ESMTP id 48178A83 for ; Fri, 9 May 2014 17:52:36 +0000 (UTC) Received: from mail-yh0-f44.google.com (mail-yh0-f44.google.com [209.85.213.44]) by smtp1.linuxfoundation.org (Postfix) with ESMTPS id CFAE02034B for ; Fri, 9 May 2014 17:52:33 +0000 (UTC) Received: by mail-yh0-f44.google.com with SMTP id b6so4032781yha.3 for ; Fri, 09 May 2014 10:52:33 -0700 (PDT) MIME-Version: 1.0 In-Reply-To: <1399657343.2166.61.camel@dabdike.int.hansenpartnership.com> References: <1399595490.2230.13.camel@dabdike.int.hansenpartnership.com> <20140509122451.5228a038@gandalf.local.home> <1399657343.2166.61.camel@dabdike.int.hansenpartnership.com> Date: Fri, 9 May 2014 13:52:32 -0400 Message-ID: From: Matthew Wilcox To: James Bottomley Content-Type: multipart/alternative; boundary=089e014952961b4d2204f8fb43d3 Cc: Sarah Sharp , ksummit-discuss@lists.linuxfoundation.org, Greg KH , Julia Lawall , Darren Hart , Christoph Lameter , Dan Carpenter Subject: Re: [Ksummit-discuss] [CORE TOPIC] Kernel tinification: shrinking the kernel and avoiding size regressions List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , --089e014952961b4d2204f8fb43d3 Content-Type: text/plain; charset=UTF-8 On 2014-05-09 1:42 PM, "James Bottomley" < James.Bottomley@hansenpartnership.com> wrote: > In theory, we could use link time optimization to place all the most > used functions in the first TLB entry. However, as Steve said, have you > got measurements showing this helps? If it's down in the noise, it's a > lot of work for no benefit. It's going to be highly workload dependent. For example, TPC-C randomly accesses all of memory. Even doubling the number of 2MB TLB entries isn't going to help more than a couple of percent. On the other hand, for a scientific workload which juuuust overflows the number of 2MB entries, you might see a 100% speedup with the freeing of a single 2MB entry to userspace. And there are many workloads in between (most exhibit at least some locality of reference). --089e014952961b4d2204f8fb43d3 Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable

On 2014-05-09 1:42 PM, "James Bottomley" <James.Bottomley@hansenpartnership.c= om> wrote:
> In theory, we could use link time optimization to place all the most > used functions in the first TLB entry. =C2=A0However, as Steve said, h= ave you
> got measurements showing this helps? =C2=A0If it's down in the noi= se, it's a
> lot of work for no benefit.

It's going to be highly workload dependent. For example, TPC-C rando= mly accesses all of memory. Even doubling the number of 2MB TLB entries isn= 't going to help more than a couple of percent. On the other hand, for = a scientific workload which juuuust overflows the number of 2MB entries, yo= u might see a 100% speedup with the freeing of a single 2MB entry to usersp= ace. And there are many workloads in between (most exhibit at least some lo= cality of reference).

--089e014952961b4d2204f8fb43d3--