I: [OT] R: R: How does Linux work?

Marco Ermini (mail@markoer.org)
Tue, 8 Jun 1999 18:48:20 +0100


> On Tue, 8 Jun 1999, Marco Ermini wrote:
> > > -- Linux coders are not comparable to the monkeys working in Redmond.
> > > Their code does not fail.
> >
> > It fails. The difference is that developers admit it. And fix it.
> > "Every software has bugs" (Linus).
> >
> > > -- In NT 4, not only are the video drivers in kernel space, but so is
> > > the windowing system.
> >
> > And soon, in NT5, IIS too (!). NT4 have a new scheduler
> And soon, in Linux, khttpd too (!).
>
> > algorithm, which have a pseudo real-time options, which
> > could cause processes to lock up the machine (actually,
> > changing a process to real-time will cause NT to lock up
> > until the process ends... ugh!)
>
> This is false. I'm not an NT fan, but its rt scheduling algorithms

Try it yourself: open the task manager and
change the priority of a multimedia player
to real-time on a Pentium 200. Then wait
until the player it's over (mouse locked too).

> are similar to Linux. There is a range of priorities that use an
> adaptive scheduling algorithm for time-sharing, like linux, and

Like almost anything in NT: it's very much teorethical. Very
pretty theory, but can you be sure everything MS says is true?
did you see the NT kernel sources? Using it in a pratical way
leads me to say that I'll find things not working so good. I'm
not an NT-hater or similar, I'm an MCP and I work every day
with NT and I'll always had problems with threads and IPC.

Now me and my team are becoming crazy in trying to get
working a game that creates and manages several threads
and that uses a touchscreen, which is very slow in responsivness;
changing the priority of something causes every kind of
problems (ex. raising the priority of the screen render
causes other processes to stop (!) and other similar
tasks; the pretty things is that the touchscreen driver
seems to use a process which could be delayed, even
if it's a kernel-process). Even not using the RT-priority
but just the highest non-RT one often lead other processes
not be weaked up on Pentium II-350! and it's just
ONE application with 5-6 threads.

Everyone in my team feels that NT's IPC is very
ugly and thread and priority management is buggy.
We are all planning to switch to Linux in the
next version of the game (just for time constraints
we're not scratching up all this NT version).

Cheers.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/