Studying MTD wrote: > > Hi all, > > I am getting on my i386 box :- > --------- > LINUX TICK AT 100 HZ > CALIBRATED CPU FREQUECY 298424000 HZ > CALIBARED APIC-INTERRUPT-TO-SCHEDULER LATENCY 1749 ns > CALIBRATED ONE SHOT APIC SETUP TIME 251 ns > > FOR 30 TASKS: TIME 222 (ms), SUPS/RES SWITCHS 60000, > SWITCH TIME (INCLUDING > FULL FP SUPPORT) 3697(ns)
On a list of 30 tasks a full task switch, including the support of saving restoring FPU registers, using suspend/resume (SUSP/RES not SUPS/RES) takes 3.7 us. > FOR 30 TASKS: TIME 213 (ms), SUPS/RES SWITCHS 60000, > SWITCH TIME (INCLUDING > FULL FP SUPPORT) 3557(ns) Same has above. I suspect you typed the above result by hand. In fact, unless you changed the test yourself, the above lines should refer to semaphores signal/wait. On my PII 233 MHz I get 2.5 us (UP). What 300 Mhz machine you are using, UP/SMP/MUP, with what version of RTAI? I know from the above lines that you should be using the SMP scheduler, which has more overhead, since it has to prepare also for messages to be sent to other CPUs having ready tasks runnable on them. Am I right? > Can someone please explain to me :- > > 1. what is the meaning to APIC-INTERRUPT-TO-SCHEDULER > LATENCY The time that passes between the timer interrupt and its programming for the next shot. > 2. why APIC-INTERRUPT-TO-SCHEDULER LATENCY is very big > number 1749 ns > 3. If possible, show me the path of the > APIC-INTERRUPT-TO-SCHEDULER (means how can we divide > 1749 ns) > It is just a point of view, to me 1749 nanosecs is not too big. The path is described above. BTW, you should buy a navigation system because it seems you lost your way in the Linux real time maze. Paolo. -- [rtl] --- To unsubscribe: echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED] -- For more information on Real-Time Linux see: http://www.rtlinux.org/
