Linux was originally written as a general-purpose operating system without any consideration for real-time applications. Recently Linux has become attractive to the real-time community due to its low cost and open standards. In order to make it more practical for the real-time community, patches have been written to affect such things as interrupt latency and context switch. These patches are public domain and are becoming part of the main Linux tree. LinuxJournal tests the preemptible patch.
Gentoo has made it easy to apply the preemptive patch on 2.4
I can’t watch for 2.6/3.0 (whatever they decide the name is) stable to be released
These patches are public domain and are becoming part of the main Linux tree
Note these patches (like everything else) are not public domain but are copyrighted to myself and MontaVista and licensed under the GNU GPL v2.0.
Oh, and they already are part of the 2.5 kernel
The preemptible kernel patches are already a part of kernel 2.5, they went in quite early in the development cycle. The performance of the patches will get better over time, as more long-held locks are broken up. I’m using the 2.5.44 kernel right now, and they do improve the feel of the system under load. For the preemptible patches to behave their best, you have to make sure X and your GUI programs run at a higher priority so they will preempt I/O programs.
PS> The article made a little boo-boo. The patches are not in the public domain, they are GPL’ed, like the rest of the kernel. I believe the copyright still belongs to the original writer of the patches (either Robert Love or Montavista).
Gentoo has made it easy to apply the preemptive patch on 2.4
Like patch -p0 <preempt-kernel-2.4.19 is *that* difficult? IIRC, debian distributes the patch too. I like to roll my own kernels, though.
As someoone pointed out in the comments to the story at LinuxJournal, the gentoo sources don’t seem to make alot of sense. I don’t notice any speedup using a kernel compiled from a gentoo tarball versus a vanilla kernel.org tarball or stock debian package.
When I started using the preempt patch, it was still umm…unblessed. On my old machine, it made a difference in day-to-day use. MP3 playback wouldn’t dropout while something else big (i.e. compiling something) was going on in the background. On my newer, faster, machines, however, I haven’t had those problems, even using stock kernels. 🙂
why is IBM going to use Linux on it new supercomputers (http://www.wired.com/news/technology/0,1282,35113,00.html“>see )…
beats me since numerous companies like sgi, disney etc have had to overhall X and other daemons for performance. I can’t imagine what else will be rewritten for performance in Linux and then made closed source so the rest of us CANT benefit.
X is a spec that can be implimented any way that a group would like as long as it matches the specs.
Disney can develope there own X-server, there are many comercial X servers that you can use with Linux.
also, if disney wanted to, they could make a kernal module and keep it Closed as Linus says that is ok (like what Nvidia has done).
why is IBM putting Linux on super computers? becasue they want to move over to an operating system that is cheaper to develope for them. Linux is that.
you need to not worry about such things as they have not yet happened.
Hmm, your comment sounds more trollish. The kerenl has the pacthes, it doesn’t mean you have to enable them.
As for perofrmance, you won’t see it during normal work, but you sure notice it when the CPU is under a heavy load (e.g: when compiling kde from cvs). The system stays quite responsive even in those cases.
kernel.kolivas.net
From Con Kolivas.
Include the O(1) scheduler, low latency, preemptible and more.
There is also one for compressed caching for low-memory situations but it hurt performance in my case, I talked to the programmer of the patch though and he has a new version in the works that performs well.
XFS, ALSA and -aa patches are there too.
D
or we can wait until 2.6 comes out.
A better test of the patches to improve the Linux kernel’s responsiveness was done back in March by a RedHat engineer. Read more at http://www.linuxdevices.com/articles/AT3595894022.html