[lfs-dev] Automake SBUs may be incorrect

Ken Moffat zarniwhoop at ntlworld.com
Wed Mar 18 10:40:02 PDT 2015


On Wed, Mar 18, 2015 at 11:22:46AM -0500, Douglas R. Reno wrote:
> Alright I understand now. Thanks Bruce! I have noticed that it all depends
> on whats running on the system at hand as well.
> 
> Ken, I am building on a physical laptop as well now (I will be away from
> home for a week and I still want to work on stuff), so I will see what it
> is after that. The VM is really only for the BluRay stuff, as my new box
> uses UEFI and I am still trying to get my RAID and stuff setup. *fun times*
> LOL.
> 
> Douglas R. Reno

There is nothing wrong with building in a VM, my concern is only
that any VM imposes some overhead on the measurement.  An SBU is
already a somewhat tenuous measurement : when we introduced it, the
number of SBUs to build later packages were not too dissimilar across
different CPUs.

But with modern CPUs, and two versions (32,64), things can change.
When I first got my Phenom x4 and my low-end SandyBridge i3, the
Phenom was faster even for -j1 builds.  Nowadays, the i3 is
apparently faster (although it still takes longer to run a -j4
firefox build).

A small difference in how long it takes to build binutils can lead
to increasing differences in the number of SBUs for slow BLFS
packages.

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 lfs-dev mailing list