Martin Pool wrote:
On 8 Jan 2003, Dan Kegel dank@kegel.com wrote:
I think the issue was that I did not have three identical machines; the speedup from distributing to three hosts was minimal compared to just building on the fast machine.
I think the scheduler doesn't handle the case of heterogenous machines as well as it might. I'm going to try to fix that.
You should make sure that the fastest machines are listed first in DISTCC_HOSTS.
OK, so if I'm running the build on the fast host, I should include 'localhost' first in DISTCC_HOSTS.
But what about SMP machines? If the fast machine is a dual processor box, and the others are all single-processor, should I put localhost twice in DISTCC_HOSTS?
That's probably part of the reason I got such a poor speedup -- I had no idea how to configure distcc well in the first place, let alone configuring it to deal with SMP systems. - Dan