DPRG
DPRG List  



DPRG: OS-9 anyone?

Subject: DPRG: OS-9 anyone?
From: MIKE MCCARTY -- 93789 JMCCARTY at DEVEL.USA.ALCATEL.COM
Date: Fri Aug 27 12:38:10 CDT 1999

[snip]

)remember the name though.  For most hobbyist applictions, Linux makes
)a lot of sense.. It's free. I'm not sure how much the OS needs to be
)designed for Real time, if the application uses only a fraction of the
)CPU, as is the case for most of my current projects. If you can have
)some control over the thread priorities, you can make sure high
)priority tasks get done first. This is all the real time most of us
)need. 

You may be right for most robotics applications, I dunno. The issues I
have with using non-real-time OSs is not that they consume enormous
amounts of CPU time, but rather that they sometimes and unpredictably
disable interrupts or just don't run my tasks for enormous amounts of
time. IOW, I don't think that the OS is "eating" the CPU, just not
allowing my interrupt handler to run until after it is too late. So
while the OS may not use more than 1% of the CPU, it unpredictably
disables interrupts for 20ms, when I have events happening once every 5
seconds, but which require action within 100us. Lots of cycles lying
around unused, but I'm still not meeting real-time constraints.

Of course, interrupts is just an example. The point is that fast,
predictable context switching is what I want, not lots of available
CPU. Normally, anyway. A design which needs a high percentage of the
CPU is probably underdesigned for hardware power, and needs another
CPU, in my experience. And fast, predictable context switching (to
another task, to an interrupt handler, etc.) is what an RTOS
guarantees, and which non-RT OSs do NOT guarantee.

YMMV

Mike
- ----
char *p="char *p=%c%s%c;main(){printf(p,34,p,34);}";main(){printf(p,34,p,34);}
I do not speak for Alcatel. <- They make me say that.
This message made from 100% recycled bits.

------------------------------

More information about the DPRG mailing list