[Coco] Kernel activity on the O-Scope
Boisy G. Pitre
boisy at tee-boy.com
Thu Mar 4 17:35:45 EST 2010
Correction. My cursor was a little wide on the margins. I zoomed in more closely and the actual time is .542 ms. So it's actually .542/16.67 = 3.25% of the time. Also this is NitrOS-9/6309 Level 2 running in native mode at 1.78MHz on a CoCo 3. The 6809 version of OS-9 Level 2's measurement is .585, so .585/16.67 - 3.5%. Sorry for the confusion. It helps to really zoom in with these scopes and place the cursors in accurate positions.
Boisy G. Pitre
On Mar 4, 2010, at 4:24 PM, Boisy G. Pitre wrote:
> I know posting attachments are discouraged, but in this case I hope folks don't mind.
> Working on my thesis here at school gives me the opportunity to play with some nice test equipment. While doing some other measurements, I figured I would measure the percentage of the time NitrOS-9 runs when there is no process active. Keep in mind that every 16.67ms the clock interrupt kicks in and the code in clock.asm checks interrupts, just to the kernel so it can do its housekeeping (checks the queues, etc.), then inserts the active process into the queue, or in this case, call the CWAI instruction since there is no active process. The shell prompt was sitting there as I captured this.
> The graph below shows the amps in yellow, the volts in green and the LIC line in purple. The little peaks represent the clock/kernel activity... the measured time is 700us, or .7ms.
> Doing the math, .7/16.67 * 100= 4.2%. So 4.2% of the clock period is spent in the kernel.
> Coco mailing list
> Coco at maltedmedia.com
More information about the Coco