You mis understand what we mean by variance. The question isn't whether or
not you can change the variables and see a different result; the question
is can you /not/ change the variables and /not/ see a change. And the
core of the problem is that all context switch tests other than Richard's
have this attibute, at least for the small process case which is what is
being measured here. The fact that his stuff varies when you haven't
changed any variables is what I view as a problem.
: I'd like to see the full spectra of results and do my own stats
: on it.
OK. All tests are with 2 processes. The numbers in parens are the results,
the first number is the median. FIFO == SCHED_FIFO was used. RR == SCHED_RR
was used. OTHER means the default scheduling policy was used.
OTHER: 4.94 (5.02 4.97 4.97 4.95 4.95 4.94 4.94 4.94 4.93 4.92 4.92)
OTHER + 6 bg: 6.29 (6.87 6.67 6.60 6.38 6.30 6.29 6.27 6.26 6.25 6.24 6.23)
FIFO: 6.62 (6.73 6.69 6.64 6.64 6.62 6.62 6.61 6.61 6.61 6.59 6.58)
FIFO + 6 bg: 10.75 (10.88 10.83 10.81 10.77 10.76 10.75 10.73 10.73
10.71 10.71 10.70)
RR: 5.84 (5.95 5.88 5.85 5.85 5.85 5.84 5.84 5.84 5.83 5.83 5.83)
RR + 6 bg: 8.30 (8.47 8.44 8.42 8.37 8.35 8.30 8.27 8.22 8.21 8.19 8.18)
Let me know if you want other data. I can also give you prepackaged
results w/ std deviation, min, max, mean, median, etc., but it sounds
like you can do all that.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/