[lkml]   [2002]   [Sep]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [patch] lockless, scalable get_pid(), for_each_process() elimination, 2.5.35-BK
On Wed, Sep 18, 2002 at 08:21:38PM +0200, Ingo Molnar wrote:
> On Wed, 18 Sep 2002 wrote:
> > So Solaris made a stupid design decision to encourage people to use a
> > thread per call on these big systems and Linux should make the same
> > decision for compatibility?
> if it can be done, why not? Do you advocate the using of big select()
> loops and userspace threading libraries?

I like request/servers and state machines. I also think that there
may be some smart ways to fold multiple threads into one. And there is
nothing wrong with userspace threading libraries.

> i have to admit that there's an inherent simplicity in using one thread
> per line, and for the more critical systems it's the simplicity that

There is a simplicity acquired by pushing the complexity to somewhere else.

> matters alot. One process per line would be even nicer - but that has a
> much higher resource footprint. While it could all be rewritten into a
> IRQ-driven state-machine as well, i dont think that is economic nor
> manageable for every case. Guess why Apache is still using the model of
> threads/processes, with a serial workflow done by them, and not the model
> of an async state-machine that TUX uses.

Because Linux does not provide an easy and highly efficient API for
building state machines and because even if it did, a crappy threads
model that runs on NT as well as Linux is more attractive.

> > I can see why people want this: they have huge ugly systems that they
> > would like to port to Linux with as little effort as possible. But it's
> > not free for the OS either.
> i believe if you do not see the dangers of O(N^2) algorithms then you
> should not do RT coding. While it's not at all the goal of the patch i
> sent, with some effort we can make Linux to behave in an RT-correct way if
> a subset of APIs are used and drivers are carefully controlled.

Just for information, don't try to apply asymptotic analysis blindly to
bounded conditions.

Victor Yodaiken
Finite State Machine Labs: The RTLinux Company.

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2005-03-22 13:22    [W:0.078 / U:6.544 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site