Great news for the Linux desktop users. Robert Love’s patch which turns the Linux kernel to behave like a preemptive one, has been accepted to the development 2.5.x Linux source tree. Preemptiveness greatly improves UI responsiveness (for example mp3s won’t skip when you do something CPU heavy at the same time and you won’t experience as many UI “locks” during normal usage). Our Take: Let’s hope that SGI’s XFS will also make it to be included by default to the 2.5.x Linux kernel.
Robert Love’s Preemptive Patch Made it Into Linux 2.5.x
2002-02-10 Linux 19 Comments
As an “Amigan” I`m very pleased to hear that. 🙂 Hopefully within the coming years good and generally accepted standards are set within the Linux community for better Desktop usage.
Still I’m using a 2.5.3-dj kernel, before I tried both a non preempt 2.4.17 and a preempt 2.4.17, Not many changes… Now I can see a DivX more or less w/out skip during a compile before I had no issues listening to my mp3 and compiling X…
Is THAT needed for having a pleasant use of desktop, or is preeempt something that may help, but a better use of resource may enhance more the UI responsiveness?
[Notice I use xmms as music player and mplayer as movie player, I’m still using enlightenment .16 and few epplets (E applets) ]
My mp3’s is skipping while running programs like ‘find’ so I’d say it’s badly needed.
I haven’t had mp3’s skipping in Windows since I upgraded from my P200…(to a PIII 450)
xmms never skips on me on my PII 300… maybe you don’t pre-buffer? (ie arts/Noatun doesn’t support pre-buffer at all)
Skipping ofently happens because DMA mode is net set for hdd. Try hdparm for this issue.
For those who want a bleeding edge kernel try the mjc branch : it has a lotta optimisations like a new scheduler, the preemp patch, etc…
Maybe it’s coz the FreeBSD folks works on a preemptive kernel a.k.a. SMPng and KSE.
I heard that this patch doesn’t scale (to multiple processors) or port (to other architectures) very well. Has this been addressed?
Nevertheless, the patch works great on my uniprocessor x86 system with a 2.4.17-mjc3 kernel.
To get the best use of this patch, make good use of the <tt>nice</tt> and <tt>renice</tt> commands. For example, I always renice my X and xfs processes to -10. If I’m compiling something and I want to do something else in the meanwhile, I’ll renice the conpilation process to 10. It works great
amen to that, tso.
Does this mean I’ll finally be able to have two browser windows downloading pages at the same time without one timing out?
I was always under the impression that the Great 32-bitOS of the Future is, indeed, a preemptive multitasking OS.
Are you telling me that it was the same as the friggin’ Win 3.1 (albeit with memory protection and 32 bits)??
The sad thing is, when I would point out ot Linux zealots that gzip would slow down X, I would get attacked from right and left with “you don’t know how to configure Linux”.
Linux is gaining the ability to pre-empt the kernel. this will make for very smooth desktop usage once it gets all tuned up. It is also for embedded systems to get better responeses.
Linux has always had preemptive multi-tasking for user space applications but not for the kernel itself. Now everything is pre-emptive. Oh and by the way in case you missed this Linux is a 64 bit operating system as well as 32 bit(Alpha, mips,etc). We are still waiting for microsoft in the 64 bit arena(and no a xp alpha does not count). And I think win3.1 used cooperative muti-tasking.
Why don’t you read up on a subject before making silly comments?
Now all we need is O(1) to added as well and Linux will be getting there.
I wouldn’t trust a PDA with Linux. Try QNX, it’s better.
Eugenia keeps on calling this a “preemptive” patch even when she knows better. At least when Slashdot’s editors are incompetent or just trolling they have a ready supply of people to shoot them down. I’d rather not bother, but since people apparently read OSNews articles as fact <sigh> …
How about something like:
Robert Love’s Preemptible Kernel Patch Makes it Into Linux 2.5.x
Great news for the future of Linux desktop and real time users. Robert Love’s patch which makes the Linux kernel fully preemptible, has been appieed to the development 2.5.x Linux source tree. Kernel preemptibility improves latency by allowing processes to be interrupted even when the kernel is running on their behalf. This feature is common in real time systems (and is apparently included in Windows NT).
Bet there are lots of places where drivers and filesystems depend on the old behaviour.
Yes, I know that Linux also runs on 64 bit CPUs in 64 bit mode. Thanks for nothing.
I wonder if the new Linux users will put up with your personal attacks any time they notice that something is wrong with the OS.
Your personal attack didn’t achieve absolutely anything. It may have made your day, to a certain extent, but it didn’t convince me or make me any more willing to suggest the use of Linux for any professional purposes. On the contrary, I’ll be mindful of the foul-mouthed, unfriendly and opinionated community that comes with the OS.
You make a nasty comment which was plain wrong. Do you really think no one was going to call you on it? You make a personal attack on Linux and then get all upset when you sound stupid. let just remind you of what you said
“I was always under the impression that the Great 32-bitOS of the Future is, indeed, a preemptive
multitasking OS. Are you telling me that it was the same as the friggin’ Win 3.1 (albeit with memory protection and 32 bits)??
The sad thing is, when I would point out ot Linux zealots that gzip would slow down X, I would getattacked from right and left with “you don’t know how to configure Linux”. ”
So its okay for you to make incorrect nasty coments but when you get challenged you react with
“how dare those Linux zelotes make me look stupid” ?
Nice double standard.
hmm the new kernel 2.5.4 is out, I’ll wait the next dj patch and I’ll try it
Is great to see that the developement is progressing nicely and many people are interested.
Probably if an Half of them were interested on the Hurd we got another choice and maybe better for certain uses…
(BTW wy don’t use the Hurd for BlueOS?)
Eugenia, please stop bothering us with XFS each time a news article speaks of the linux kernel.
XFS has nothing better than reiserfs, beside attributes which are no use without apps.
BTW, reiser performs way better on most systems.
If you want xfs, patch your own kernel, but stop your propaganda. No need to speak of it in an article about the preemptible patch or whatever.