From mboxrd@z Thu Jan 1 00:00:00 1970 Subject: Re: [Xenomai-help] pcspeaker tsc From: Philippe Gerum In-Reply-To: <45DC42D5.9020605@domain.hid> References: <45DB5A29.9080506@domain.hid> <45DB5FC0.6060004@domain.hid> <45DC42D5.9020605@domain.hid> Content-Type: text/plain Date: Wed, 21 Feb 2007 15:53:55 +0100 Message-Id: <1172069636.3551.40.camel@domain.hid> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Sender: Philippe Gerum Reply-To: rpm@xenomai.org List-Id: Help regarding installation and common use of Xenomai List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: rolandtollenaar@domain.hid Cc: Xenomai-help@domain.hid, Jan Kiszka On Wed, 2007-02-21 at 14:02 +0100, Roland Tollenaar wrote: > Hi, > > my kernel was compiled for processor family 486. If I change this I > discover that I have to recompile all the modules as well. For normal > kernel recompiles this is normal practice, for my live distribution it > is a house load of really crappy work. > > So I am considering to leave my kernel in 486 state until later. > > My final application will only have a fastest cycle-time of 1ms and > according to Philippes interpretation of my latency test read-out I > should be able to achieve that easily without TSC support in the kernel. > > My question is, can I do this? What problems may I experience later on > as a result of this choice now. > > I want to get some feel for the application space programming with > xenomai and it does not seem worth spending forever getting this right > now. Or am I wrong? > > > Are there any settings in the kernel that will improve behaviour without > the TSC support that I could or should apply? > On a low-end machine, using TSC improves latency by ~7-10 us, depending how crappy your 8253/8254 PIT and/or ISA bus are under high load conditions, so if your period is 1ms, just don't bother; latency will come from other issues, like cache refills. Functionally speaking, TSC or not does not make any difference Xenomai-wise, except if you need to use the PC speaker while RT stuff is running. Since both the PC speaker (vanilla kernel code) and Xenomai's TSC emulation code would use the PIT's channel #2, Xenomai either voids the speaker code for 2.4 kernels, or disables it at compile-time with 2.6 kernels. > Kind regards, > > Roland Tollenaar. > > Jan Kiszka wrote: > > Roland Tollenaar wrote: > >> Hi > >> > >> I have some form of kernel recompiled in which I have the power > >> management turned off. The PCspeaker is also unselected but it comes > >> alive anyway. This is the result -or so it seems- of the soundcard > >> driver. I would like to rmmod the sound card driver after booting if > >> this interferes with xenomai but there are so many dependencies snd-this > >> and snd-that, that its not trivial. > >> > >> -Is this sound-card driver a problem for xenomai? > > > > Nope. > > > >> -how can I be sure that the pc speaker is harmless for xenomai > > > > Yes, as long as you are not targeting a very low-end x86 platform (any > > pre-Pentium system without TSC support). But Xenomai will warn you then > > during kernel configuration. > > > > Jan > > > > _______________________________________________ > Xenomai-help mailing list > Xenomai-help@domain.hid > https://mail.gna.org/listinfo/xenomai-help -- Philippe.