From mboxrd@z Thu Jan 1 00:00:00 1970 From: Scott Wood Date: Fri, 27 Apr 2012 16:37:25 +0000 Subject: Re: [PATCH 2/2] KVM: PPC: Book3S: Call into C interrupt handlers Message-Id: <4F9ACB45.3040803@freescale.com> List-Id: References: <1335435543-19690-1-git-send-email-agraf@suse.de> <1335435543-19690-2-git-send-email-agraf@suse.de> <20120427054810.GC1216@drongo> In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Alexander Graf Cc: Paul Mackerras , kvm list , kvm-ppc , Benjamin Herrenschmidt On 04/27/2012 06:23 AM, Alexander Graf wrote: > > On 27.04.2012, at 07:48, Paul Mackerras wrote: > >> Have you measured a performance improvement with this patch? If so >> how big was it? > > Yeah, I tried things on 970 in an mfsprg/mtsprg busy loop. I measured 3 different variants: > > C irq handling: 1004944 exits/sec > asm irq handling: 1001774 exits/sec > asm + hsrr patch: 994719 exits/sec > > So as you can see, that code change does have quite an impact. But > maybe the added complexity isn't worth it? Either way, we should try > and find a solution that works the same way for booke and book3s - I > don't want such an integral part to differ all that much. Is it really added complexity, considering what you can remove from the asm? I went with C handling on bookehv because it seemed simpler (the original internal version had asm handling). -Scott