[blfs-dev] bash script

Ken Moffat zarniwhoop at ntlworld.com
Tue Jun 3 08:21:12 PDT 2014


On Tue, Jun 03, 2014 at 11:16:32AM +1200, Christopher Gregory wrote:
> 
> It is really weird that the times do vary between runs of the script.
> 
 On a non-realtime OS you should always expect _some_ variation.
If you look in 'top', even on a system without 'cron' or 'at' there
will be a lot of processes which might wake up.  Also memory pressure
if you are doing other things during the build and disk-head
movement (on rotating hard disks).

 For different toolchain versions there will also be differences
(usually, newer means slower to compile, but with 4.9 I have seen
shorter compiles in some packages.

 Also, after watching the frequencies of my four AMD cores during
compilation with -j1 on a mostly idle system (I was worried about
why the ondemand governor showed excessive variation in my SBUs for
recent system builds) I realised that work is being shuffled around
the cores.  On modern intel, the 'performance' governor has replaced
'ondemand' and works as well as ondemand used to (i.e. it quickly
speeds up and quickly slows down when the load has gone).

ĸen
-- 
Nanny Ogg usually went to bed early. After all, she was an old lady.
Sometimes she went to bed as early as 6 a.m.


More information about the blfs-dev mailing list