From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756058AbcHWUzH (ORCPT ); Tue, 23 Aug 2016 16:55:07 -0400 Received: from merlin.infradead.org ([205.233.59.134]:52956 "EHLO merlin.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754286AbcHWUzF (ORCPT ); Tue, 23 Aug 2016 16:55:05 -0400 Date: Tue, 23 Aug 2016 22:55:02 +0200 From: Peter Zijlstra To: Paul Turner Cc: Dietmar Eggemann , Ingo Molnar , LKML Subject: Re: [PATCH] sched/fair: Fix fixed point arithmetic width for shares and effective load Message-ID: <20160823205502.GX10153@twins.programming.kicks-ass.net> References: MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.5.23.1 (2014-03-12) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, Aug 23, 2016 at 01:40:01PM -0700, Paul Turner wrote: > Independent of this patch: > When we initially introduced load scaling, it was ~uniform on every > value. Most of the current pain has come from, and will continue to > come from, that with v2 of the load-tracking this is no longer the > case. We have a massive number of scaled and unscaled inputs floating > around, many of them derived values (e.g. source_load above) which > require chasing. > > I propose we simplify this. Yes please... I had been thinking of trying to use sparse to give the different FP scales different types, but gave up real quick on that.