nonlinear timeslice gap?

[Jason Baietto]

ChangeSet 1.1832.6.22 (2004/08/24 11:10:51 mingo@xxxxxxx) introduced the
following time slice mapping for nice values:

   [ -20 ... 0 ... 19 ] => [800ms ... 100ms ... 5ms]

This met the goal of a 1:160 ratio for nice-20/nice+19, however it
introduced a 320ms gap between 0 and -1.  Was this gap intentional or
accidental, and is it really acceptable?

Here's the full mapping (verified by sched_rr_get_interval):

-20 => 800ms
-19 => 780ms
-18 => 760ms
-17 => 740ms
-16 => 720ms
-15 => 700ms
-14 => 680ms
-13 => 660ms
-12 => 640ms
-11 => 620ms
-10 => 600ms
 -9 => 580ms
 -8 => 560ms
 -7 => 540ms
 -6 => 520ms
 -5 => 500ms
 -4 => 480ms
 -3 => 460ms
 -2 => 440ms
 -1 => 420ms *** gap ***
  0 => 100ms
  1 => 95ms
  2 => 90ms
  3 => 85ms
  4 => 80ms
  5 => 75ms
  6 => 70ms
  7 => 65ms
  8 => 60ms
  9 => 55ms
 10 => 50ms
 11 => 45ms
 12 => 40ms
 13 => 35ms
 14 => 30ms
 15 => 25ms
 16 => 20ms
 17 => 15ms
 18 => 10ms
 19 => 5ms

Take care,
Jason
--
jason.baietto@xxxxxxxx
http://www.ccur.com/oss

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