Dan Creswell wrote:
>This is a Dual 2GHz G5 - of course, the test is single threaded so the
>extra CPU means nothing.
Not entirely true. Turn one of 'em off and see what happens. The extra
CPU has to take out the garbage, mow the grass, clean out the garage, and
pick up the kids from soccer. In other words, the extra CPU is busy doing
all the OTHER things that need doing.
Or more precisely, the OS's scheduler will assign CPUs as needed to
maximize the concurrency of ALL the things that are happening.
>FileOutputStream write 8M (avg) 47
>FileOutputStream write 8M (mean) 45
8M what? 47 what? 45 what? Ohms? Furlongs? Quatloos?
"M" is simply a multiplier meaning "million". It does not mean "megabytes".
Mr. McDougall loses 10 points for neglecting units.
And what's the difference between "average" and "mean"?
After inspecting the code, I see the terms should actually be "mean" and
"median", respectively. Mean is not median.
"Mean" is the sum of all measured times divided by the number of
measurements (arithmetic mean).
"Median" is the central value, where half the measurements are below it,
and half above it.
Another 10 points lost for incorrect terminology.
Ref:
<http://en.wikipedia.org/wiki/Mean>
<http://en.wikipedia.org/wiki/Median>
>FileOutputStream fill and write 8K avg 60
>FileOutputStream fill and write 8K mean 59
8K? Kilo? One thousand? Fill and write takes over 1000 times longer than
write alone?
Inspecting the code again, I see the values are mislabeled. The units are
actually megabytes.
The applicant randomly loses either 1000 points or 0.001 points for
incorrect statement of scale.
As a suggested improvement, I'd like to see the min and max measurements
reported, too. Say, min median max.
-- GG
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Java-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/java-dev/email@hidden
This email sent to email@hidden
