said:
>The fastest response comes from dedicated hardware, and I don't mean
>"computer" hardware. I'm talking analog -- opamps with fast slew rates
>or whatever. It's been a few years since I designed analog pc boards
>(remember Bishop Graphics and all their circuit design stickies?), but
>even back then the response was "instantaneous". I _think_ slew rates
>are about 20 volts/usec these days for an off-the-shelf opamp.
You are preaching to choir. You've got to remember I've been involved in
some aerospace stuff. Things like a positioning systems with arc-sec
accuracy requirements. The inner loops where invariably analog. Some of
this was to avoid the effects of digital math on the servo loops. Most
was because it was the cheapest way to achive the required loop response.
My stuff was the outer loops and command and control. Even this was
required a fair amout of processing power. The DSP we used was capable
of implementing a 50 block control loop with several filters in 800 uSec.
>That's the kind of thing I'm interested in. _Why_ is "1 msec" the point
>where you feel OS/2 "hits the wall"?
It's not so much as hits the wall, as diminishing returns. Ideally you
want to write as much of the application as possible in ring 3. This
requires context switches. My understanding is the the OS/2 scheduler
will not guarantee an context switch to even the highest priority process
in much less than this.
Also, the style guidelines allow driver's to keep control, in ring 0, from
longer than you would prefer in a hard read-time system.
>Is the amount of latency in the OS/2 interrupt handler enough to keep
>OS/2 from achieving a 100 usec response time?
No, the interrupt latency is not that bad. It's the context switching
that is the issue.
>Hah, you got me. What is a "context hook"?
You know those DDK books you misplaced...
A context hook allows an interrupt handler to schedule part of itself to
run at task time (i.e. just before the sceduler would return control to an
interrupted process). This code is interruptable, so it can do the work
it needs without holding off other pending interrupts.
Steven
--
----------------------------------------------------------------
"Steven Levine" MR2/ICE 2.19z #10183 Warp4/FP11
www.scoug.com irc.webbnet.org #scoug (Mon 7pm PST)
----------------------------------------------------------------
=====================================================
To unsubscribe from this list, send an email message
to "steward@scoug.com". In the body of the message,
put the command "unsubscribe scoug-programming".
For problems, contact the list owner at
"rollin@scoug.com".
=====================================================
<< Previous Message <<
>> Next Message >>
Return to [ 30 |
April |
2000 ]
The Southern California OS/2 User Group
P.O. Box 26904
Santa Ana, CA 92799-6904, USA
Copyright 2001 the Southern California OS/2 User Group. ALL RIGHTS
RESERVED.
SCOUG, Warp Expo West, and Warpfest are trademarks of the Southern California OS/2 User Group.
OS/2, Workplace Shell, and IBM are registered trademarks of International
Business Machines Corporation.
All other trademarks remain the property of their respective owners.
