http://qa.mandrakesoft.com/show_bug.cgi?id=3172
------- Additional Comments From [EMAIL PROTECTED] 2003-03-12 04:32 ------- Eric, I can confirm what you've written here: I tried reverting to as many kernels as I could try, and they all had the same (bad) performance. In addition, after using smp_affinity to make IRQ's route to both CPUs, mencoder performance didn't improve. For now, I'll see if I can get back the performance I need by recompiling from source here. However, it's still perturbing to see that by default IRQ's are not being routed to both processors when the APIC is enabled. Even if it's not the cause of this particular performance problem, it's still something that's changed for the worse. Perhaps we should open a bug just about that issue? ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is. ------- Reminder: ------- assigned_to: [EMAIL PROTECTED] status: RESOLVED creation_date: description: This bug has been signaled on cooker mailing list, I signal it on behalf of Wesley J Landaker, who could not reach Bugzilla: > IO-APIC is enabled on my system, however all interrupts are routed to > CPU0. This didn't used to be the case, and I see a *drastic* > performance hit because of this (for instance, using mencoder to > capture real-time TV I used to get 35fps, now I can't even get > 20fps). It has been confirme by several people with different kernels: with standard kernel 2.4.21, I observed a very drastic performance problem with mencoder, encoding a video or TV, or encoding sound in oggvorbis (about 5 times slower). This happens also with the smp kernel, and this is even worse with the multimedia kernel.
