Momchil Velikov wrote: > On Thu, 31 Oct 2002 09:06:25 -0800 (PST) > > shino korah <shino_korah@yahoo.com> wrote: > > This results shows HT improves performance if we have > > 4 tasks on DP system and performance is hit if we use > > 2 tasks on a DP system. > > Probably the two tasks were scheduled in the same physical CPU, on each > HW thread, leaving the other CPU idle. that's the chance-dependent behavior that the HT-aware scheduler fixes. what surprises me is that Shino gets consistent numbers. maybe the 2.4 scheduler has a low cpu-affinity, so the tasks keep moving, averaging the balanced-inbalanced case. in contrast, Ingo's tests were very different from one try to the next, but he used the latest 2.5 kernel as the reference. sometimes the processes would be balanced by chance, and some other times they wouldn't; and cpu affinity would prevent from jumping to the other cpu at the next context switch. does somebody know the history of the cpu affinity feature? ------ Javier -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
