tech-userlevel archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: [PATCH] Re: sharing mutex between libc stub and libpthread

>> 100 million free(malloc(1024)): slowdown 2%
>> without the patch: mean 27.36s, 106 samples, sd 0.48
>> with the patch: mean 28.04s, 245 samples, sd 0.25

>> 100 million gmtime(&now): slowdown not significant
>> without the patch: 49,15s, 34 samples, sd 0.35
>> with the patch, 48.55s, 51 samples, sf 0.50

> By my quick calculations, that's a 1.2% difference.  While not
> substantial, it seems significant to me.

I don't know what manu meant.  But "significant" is a technical term in
statistics; a measured difference may be "not significant" in the sense
that it can plausibly be explained by measurement noise rather than a
real difference in the thing being measured.

My statistics knowledge is way too rusty to compute levels of
significance here.  But if the difference is .6 and the standard
deviation (I assume "sf" in the fourth measurement is a typo for "sd",
and that "sd" means "standard deviation") is .5, that's about 1.2
standard deviations off the mean.  In a sample size as small as 85,
this very well not be significant.  (The malloc/free test difference is
approximately twice as large, about 2.4%, has a sample size over four
times as large, and is some 2.72 standard deviations instead of 1.2.)

Note also that according to these measurements, gmtime got faster, not
slower.  This makes it even more plausible to me that the difference is
due to measurement noise.

/~\ The ASCII                             Mouse
\ / Ribbon Campaign
 X  Against HTML      
/ \ Email!           7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B

Home | Main Index | Thread Index | Old Index