[yocto] Procedure to setup icecc for performing a distributed build

[Paul Eggleton]

On Thursday 06 September 2012 09:33:26 Gary Thomas wrote:
> On 2012-09-06 09:29, Paul Eggleton wrote:
> > On Thursday 06 September 2012 19:23:53 Elvis Dowson wrote:
> >> Unfortunately no. I've got two machines, both with quad-core intel i7
> >> processors, but I just couldn't get icecc to work with yocto. I end up
> >> regularly perform fresh builds at least 5 to 6 times a day, and it takes
> >> me
> >> 2 hours to build core-image-minimal.
> > 
> > At its core, icecc.bbclass just sets things up so that icecc is called
> > instead of the normal compiler. It would be worth verifying that that is
> > actually taking effect by looking at the commands being used as reported
> > in log.do_configure for specific recipes.
> 
> Is this only for the host compiler or the target (cross) compilers as well?
> If it's only the host compiler being used for "native" builds, how much is
> there to be gained from distributing the work?

It covers all according to the comments at the top of the file.

I have to be honest and say I've never used icecc.bbclass; I've been told that 
the overhead mostly wipes out any advantage of individual compile jobs being 
farmed out to different machines, but I've never actually verified that first 
hand.

Cheers,
Paul

-- 

Paul Eggleton
Intel Open Source Technology Centre