[nycbug-talk] make -jN performance

Isaac Levy ike
Thu Jan 20 11:59:45 EST 2005


Hey All,

On Jan 19, 2005, at 5:34 PM, Pete Wright wrote:

> hmm...then it looks like I should submit a PR to the doc team 
> then...well
> atleast it was an interesting exercise :)
>
> -pete

Yeah- I wanted to pipe in quicklike with this one with some super-crude 
test results,

Basically I've been compiling a lot of FreeBSD jails lately, so what 
the hey, I've thrown in some -j flags informally to see what happens, 
here's my results:

--
Single CPU (crufty old office 800mhz PIII):

- No flags, make world takes 49 min average.
- j2, j4, j6, j10, all take 50 min average.

--
Dual CPU (new-ish Dual Xeon 2.6mhz, under other server load):

- No flags, make world takes 22 min average.

- j2, j4, j6, j10, all take about 12 min average. (NICE.)



--
Lots of other stuff could affect my results here, but in a nutshell, it 
seems that the j flag has no effect unless there's more than 1 
processor.

That stated, I'd think that in my scripted jail-build stuff, it would 
be worth my time to throw a -j4 flag in there, as it doesn't seem to 
hurt much, and what the hey- the systems may port to 4 cpu machines in 
the not-too-distant future...

Reporting from the Why-the-heck-not Dept.,
.ike





More information about the talk mailing list