From mboxrd@z Thu Jan 1 00:00:00 1970 Message-ID: <51EA4243.4020904@web.de> Date: Sat, 20 Jul 2013 09:54:43 +0200 From: Jan Kiszka MIME-Version: 1.0 References: In-Reply-To: Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Subject: Re: [Xenomai] shielding fully CPU or CORE List-Id: Discussions about the Xenomai project List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Skander Bahloul Cc: xenomai@xenomai.org On 2013-07-19 11:17, Skander Bahloul wrote: > Dear Xenomai list members, > = > On a successful instal (but without NVIDIA driver, cf my previous message= ), I am now starting using Xenomai. > = > We need to : > 1* shield fully a CPU or core, against the scheduler, interruption, etc > 2* run a single thread on this isolated CPU or core > = > Step 2 is OK with the Xenomai API. > But how should I achieve step 1 ? > = > I did not find advice in the Xenomai documentation. > I guess it should be done with : > $ sudo shield -a 0-3 -m -1 > = > Do you agree ? > When may I find detailed info on CPU shielding ? With "shield fully", do you mean 99,9...% or true 100%? The latter is not possible with current Linux kernels, thus also not with Xenomai. For the former, try kernel parameter isocpus (to keep Linux off certain cores), possibly in combination with xeno_hal.rthal_supported_cpus (to limit Xenomai to those cores). True CPU domination is possible via virtualization, e.g. I'm working on such a solution. But you will not get a Linux process environment for your RT task, rather a set of bare-metal CPU cores you will have to boot up yourself. Also, a lot of noise comes from other shared resources (besides the CPU core itself), like memory and specifically the links to the I/O devices (PCI bridges/buses...). So no miracle to expect that way as well, at least on x86. Jan -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 263 bytes Desc: OpenPGP digital signature URL: