distcc not very efficient because I can't emerge in parallel

[geckosenator]

I have two computers, 1.7ghz p4 and 3000+ amd64. I set them up to do distcc together (cross compiling) and it is working, but it is not very efficient. I don't know if this is completely gentoo specific. The trouble is, the remote computer can only help out when the emerge is compiling. This is not true for configuring or any of the other time consuming tasks the local host must perform (cleaning copying etc...)

The reason distcc is possible is because make has a -j option. Why doesn't emerge have this option too? Then multiple packages could be built in parallel (assuming non-conflicting dependancies) This would help more than distcc users. This would allow compiling one package while another downloads, especially helping users with very slow connections.

distcc needs a way to distribute gccx too.. this is taking forever

any ideas?

[NeddySeagoon]

geckosenator,

A few points from your post.

The host has to do the configure and link but MAKEOPTS="-j5" allows five parts of the same packahe to build at the same. There would be no advantage to emerging several packages at the same time - the object being to get through the whole lot as fast as possbile.

You should do some tuning to your distcc set up. You may only want the host to run one make job while the remote Pc does more. This helps balance the load. On larger compile farms, the controlling host often does no compiling at all. It has to do configure and linking.

You can run the downloading and compiling at the same time. Run your emerge command in two xterms, the first with the -f option (fetchonly) and the second without. Proper locking is implemented now so it should not get in a mess if the second instance catches up with the first.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

[Dikkiedik]

Hi,

I realise this post is a bit old but still I have a question.

How exactly do you tweak distcc to make the secondary computer run the makes, and the primary computer do
the linking, etc??

Could anybody give me a good example of how to use the distcc configuration file? I'm trying to use distcc
with a Pentium 4 and an old 800 something mhz pentium3 coppermine.

I can't see any output in the distcc monitor, not the text based and not the gui, and when I use the special
command to start the gui it sais it can't complete the pol since the directory which is specified doesn't exist.

I was wondering if anybody has succesfully used distcc, what their options were and how and what they've
configured to make this happen because I'm not getting it.

Thanks allready,

Kev.

[NeddySeagoon]

Dikkiedik,

I've used distcc briefly on a pair of athlon XPs.
It does work although two PCs is probably not enough. I used MAKEOPS="-j5" for the two systems.

You set the host not to do any compiling in distcc setup when you tell distcc how many jobs it can allocated to each PC. You can set none for the controlling host.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

[Dikkiedik]

How exactly do I run this setup?

There is distcc-config but there's no information on the total of jobs each pc has to do.. So I'm a bit lost there,

thanks for the information though. I do have 2 other pc's and when I get the hang of distcc I'll surely use them .

[NeddySeagoon]

Dikkiedik,

Its in