Hey All,

first thanks for the replies on my Yocto mail.

Now I have a real PTXdist question;

What do you use to speed up your builds?

For example I use webkitgtk and that clips my 24 core Xeon at 100% for
like 30 minutes.

ccache doesn't really help by a first run, it even makes it worse. Also
I must  build several versions and so the ccache hast to be massive or
else the cache is always just updating and before you can use it it is
flushed from it already.

So get faster hardware? Of course one could get a >100k euro multi CPU
system, but I don't think that is the most economic choice.

Distributed CC (distcc/icecc) seem a possibly to scale up things in
smaller steps (ad cheap compile server one by one) Sadly I don't have
any experience with it, but it is something I want to look into.

PTXDist's -j functionality I can not use for all my builds, for example
I am stuck on kernel 3.1.10 (thanks nVidia), which means I am stuck on a
older systemd and can't use gcc 5.4 etc. So I am stuck at a rather old
ptxdist, and never got to try the -j stuff. (But with the webkitgtk
example it won't help much, the CPU is at 100% already)

So what do ppl do to make things faster ? I would really like to hear
other ppls experiences with large builds.

- Erwin

_______________________________________________
ptxdist mailing list
[email protected]

Reply via email to