From mboxrd@z Thu Jan 1 00:00:00 1970 From: Kun Wang Subject: Re: KVM on IBM PowerEN processor Date: Wed, 31 Aug 2011 00:49:20 +0800 Message-ID: <4E5D1490.40000@gmail.com> References: <4E5CB12D.1070802@suse.de> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: kvm-ppc@vger.kernel.org, "kvm@vger.kernel.org" , Scott Wood , Yoder Stuart-B08248 To: Alexander Graf Return-path: In-Reply-To: <4E5CB12D.1070802@suse.de> Sender: kvm-ppc-owner@vger.kernel.org List-Id: kvm.vger.kernel.org =E4=BA=8E 2011-8-30 17:45, Alexander Graf =E5=86=99=E9=81=93: > Hi Kun, > > On 08/29/2011 11:31 AM, Kun Wang wrote: >> Hi, everyone, >> >> This is Kun Wang from IBM Research China. I and my team have been=20 >> working >> on IBM PowerEN processor in recent years, including its simulation, >> lib/runtime optimization and etc. Now we start the work to enable KV= M on >> PowerEN processor. Since the A2 core of PowerEN follows Power ISA v2= =2E06 >> (more specifically, book3e and 64-bit), I believe 99% of our work wi= ll >> stick to the ISA, and hence can be leveraged by others. >> >> As the new one to this KVM world, I and my team definitely need your= =20 >> help. >> Looking forward to talking and working with you guys in the future. > > Welcome to the PowerPC KVM world! I'm looking very much forward to=20 > working with you there. Please always CC kvm@vger.kernel.org for=20 > emails you send to kvm-ppc@vger, unless you think of it as completely= =20 > off-generic-kvm discussions. I like the idea of having more clever=20 > people involved in the PPC KVM process and by CC'ing kvm@vger, we get= =20 > more exposure. > > The Freescale e500 cores are basically BookE 2.06 compliant with a fe= w=20 > specialties here and there. Have you looked at their code? Sure, it's= =20 > 32-bit as is now but there is definitely work going on at Freescale t= o=20 > get it 64-bit too. > > Maybe it makes sense to take the current code as a starting point and= =20 > slowly move towards an architecture that is more generic. Maybe it=20 > makes sense to start from scratch and actually design something=20 > flexible that splits guest and host TLB code, so we could potentially= =20 > run cross-book KVM in the future (which I would love to see happen!=20 > Just imagine running a PowerEN guest on a POWER7 system). > > Either way, please make sure to coordinate any efforts with me and=20 > Scott so we don't walk off in different directions and only get to=20 > realize we went down the wrong path when there's already a 100 patche= s=20 > patch set on the mailing list :) > > > Again, welcome on board! > > Alex > > Thanks, Alex. Yes, we are aware of the e500 work done by Freescale folks. I agree wit= h=20 you that we should coordinate together to achieve as more synergy. Best regards, Kun Wang