From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dario Faggioli Subject: Re: [PATCH 6 of 8 [RFC]] libxc: introduce xc_domain_move_memory Date: Thu, 11 Apr 2013 16:16:40 +0200 Message-ID: <1365689800.5135.24.camel@Solace> References: Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="===============1992152043423804433==" Return-path: In-Reply-To: List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Dan Magenheimer Cc: Olaf Hering , Ian Campbell , Konrad Wilk , George Dunlap , Andrew Cooper , Juergen Gross , "Tim (Xen.org)" , xen-devel , David Vrabel , Keir Fraser , Andres Lagar-Cavilla , Jan Beulich , Roger Pau Monne List-Id: xen-devel@lists.xenproject.org --===============1992152043423804433== Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="=-HB+bGCQUSyzfvl/BnXLx" --=-HB+bGCQUSyzfvl/BnXLx Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable On mar, 2013-04-09 at 18:43 +0100, Dan Magenheimer wrote: > > From: Dario Faggioli [mailto:dario.faggioli@citrix.com] > > Subject: [PATCH 6 of 8 [RFC]] libxc: introduce xc_domain_move_memory >=20 > (NUMA discussion...) >=20 Hi Dan, > >=20 > > XXX Open issues are: > > - TMEM: how to "move" it? >=20 > (Konrad added to cc list.) >=20 > Tmem memory is, by definition, the lowest priority memory > for the domain and the hypervisor may already be storing it as > efficiently as possible (i.e. the page may be deduplicated). > When it is accessed by the domain (it is never directly > addressable by a domain, and a hypercall is required > to access it), an entire page is sequentially copied from > a physical page in the hypervisor to the domain. Juergen may > know otherwise, but I'd guess this inter-node copy would be > very efficiently pipelined, cache-line by cache-line > possibly even with hardware pre-fetching. >=20 Ok, thanks for the clarification. > So the best answer to "how to move it?" may be "don't > move it at all!". =20 > Ok. I sort of got the feeling that "not touching" would have been TRT but, again, thanks for making it clear. :-) > In fact, a good design for a NUMA-aware > implementation of tmem might intentionally store the data on > "any node other than the node making this tmem-put hypercall". >=20 Well, we'll get there too, sooner or later. For now, and for the purpose of this specific work, I'll put things in such a way that they live TMEM alone. Regards, Dario --=20 <> (Raistlin Majere) ----------------------------------------------------------------- Dario Faggioli, Ph.D, http://about.me/dario.faggioli Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK) --=-HB+bGCQUSyzfvl/BnXLx Content-Type: application/pgp-signature; name="signature.asc" Content-Description: This is a digitally signed message part Content-Transfer-Encoding: 7bit -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.13 (GNU/Linux) iEYEABECAAYFAlFmxcgACgkQk4XaBE3IOsTjaQCdFSUtikmP8uVsKj8p9qyXnKc3 xxQAn1Y09ARMTXcrmGlUTkkcOwB6hL8x =4ASX -----END PGP SIGNATURE----- --=-HB+bGCQUSyzfvl/BnXLx-- --===============1992152043423804433== Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Disposition: inline _______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org http://lists.xen.org/xen-devel --===============1992152043423804433==--