On Mon, Mar 16, 2009 at 9:10 AM, Christian Packmann <Christian.Packmann@xxxxxx> wrote: > Er, I am in control. This is my desktop machine. It just does what I tell it > to do. Nothing else. Do you accept any other behavior from your computers? > :-) Of your computer? Yes. Of some end user running your program (which is quite likely to be the case in the endless examples of distributed computing projects previously listed in the thread). No. > Finally, a sound technical reason against hard affinity. Thank you. > > Which only becomes a problem if you run two programs using hard affinity at > the same time which try to use the same cores/threads. Otherwise the > scheduler will push the non-bound threads on the idle cores. > That's making an enormous assumption about scheduler behavior right there. You deciding to use hard affinity on a particular thread of your program does not imply that it should get to monopolize that core. Especially if the other process in the picture already had soft affinity for that core and was higher priority than yours, I see no reason why the scheduler should push it off someplace else. > How much of a problem this would be in practice is another thing. I assume > that only a small minority of programs (<0.1%) will ever use hard affinity. > And by providing information about "occupied" hardware threads, this might > be alleviated by allowing programs using hard affinity to simply select > other cores/threads (of which there will be *many* in a few years). > So, this appears solvable for the most part. And if a user tries to run That seems questionable. Given that you're not in charge of the cores, the information you're given about how busy a given set of cores are is dated by the time you get it, since all of those cores are executing things at the same time as you. By the time you've decided which cores look best for you to run on, the situation may have already changed and your choices rendered invalid. And by your own admittance just now, only a tiny percentage of programs would hypothetically even try to use such a feature at all, so I really don't see it being worth the effort. Regards, Rene