Re: Idle priority

4 Aug 2009

      site_archiver@lists.apple.com
Delivered-To: darwin-kernel@lists.apple.com

-- Terry
Hi!
So, any suggestions? Is then this a bug in the scheduler?

Mike
--- On Fri, 7/31/09, Mike Mimic <ppagee@yahoo.com> wrote:
From: Mike Mimic <ppagee@yahoo.com>
Subject: Re: Idle priority
To: "Terry Lambert" <tlambert@apple.com>
Cc: "darwin-kernel@lists.apple.com" <darwin-kernel@lists.apple.com>
Date: Friday, July 31, 2009, 1:54 AM
Hi!
--- On Thu, 7/30/09, Terry Lambert <tlambert@apple.com>
wrote:

You should probably read:

Thanks. I have read it and improved my test program,
"nice":
#include <mach/mach.h>
#include <mach/thread_policy.h>
#include <stdio.h>
int set_idle() {
    struct thread_precedence_policy policy;
    policy.importance = 0; // IDLE_PRI
    return
thread_policy_set(mach_thread_self(),
THREAD_PRECEDENCE_POLICY, (thread_policy_t)&policy,
THREAD_PRECEDENCE_POLICY_COUNT) == KERN_SUCCESS;
}
int main(int argc, char *argv[]) {
    if (!set_idle()) {
        fprintf(stderr,
"set_idle() failed\n");
        return 1;
    }
    int j;
    for (j = 0; ; j++) {
        printf("%d\n", j);
        // Busy wait
        int i;
        for (i = 0; i <
100000000; i++);
    }
    return 0;
}
In documentation it is written about the importance value:
While this is a signed 32-bit value, the minimum legal
value is zero (IDLE_PRI). threads set to IDLE_PRI will only
execute when no other thread is scheduled to execute.
Measurements are as follows.
In the first case I got such values:
- user time
  - mean: 10.0282
  - standard deviation: 0.0008
- real time
  - mean: 10.0939
  - standard deviation: 0.1165
In the seconds case I got such values:
- user time
  - mean: 10.0285
  - standard deviation: 0.0005
- real time
  - mean: 20.0722
  - standard deviation: 0.1155
There is even bigger difference. Am I doing something
wrong?

_______________________________________________
Do not post admin requests to the list. They will be ignored.
Darwin-kernel mailing list      (Darwin-kernel@lists.apple.com)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.a...
If you think there is a bug there, then you need to file a bug and
attach your code examples so that the engineers who work on the
scheduler can decide whether or not it's a problem of implementation
or of perception.
I will point out that the reported real time on the second one is very
close to 20.0722, double the CPU time, which implied that it didn't
actually get any CPU time until the competing (non-idle priority)
process finished, which is what you were appeared to be asking to have
happen.  However, you can't take the "top" or "time" numbers
literally, since they operate on digested resource usage information
obtained after the child exits before it is reaped, using BSD rusage()
calls.
To get actual time-on-cpu on a per-process basis you'd need to examine
the thread state structure before the thread exits.  This is why I
suggested you use shark to gather the information, back when this
thread started.
On Aug 4, 2009, at 2:39 PM, Mike Mimic wrote:

    <http://developer.apple.com/documentation/Darwin/Conceptual/KernelProgramming...
...
which documents the Mach scheduler interfaces.

Mike

This email sent to site_archiver@lists.apple.com

Terry Lambert

tags

participants (1)