Closing as a duplicate of bug #105.  Please add extra notes to that bug if you want.

Hi,
  If I have a lot of jobs to be run with -j I might know that some take a lot longer to run than others; on a really bad day what might happen is that it runs lots of the faster jobs in parallel and then starts the longest to run last when there is very little else to run in parallel.

So what I'd like to do is to be able to influence the choice of job that gets run when there are a bunch of jobs that are all able to be run.  I suggest a way to do this would be to attach a weight to each job and always run the job with the heighest weight first (or earlier). I guess this wouldn't be too hard with some form of phony target?

Now once that is in things could start to get even more interesting; lets imagine that weight was an estimated time in seconds - of course I could supply it, or make could spit out a log of the times taken by a previous run and feed it back in at the top!

Dave (linux@treblig.org)