SMP profiling broken in 2.1.78?

Emil Briggs (briggs@bucky.physics.ncsu.edu)
Mon, 19 Jan 1998 00:58:38 -0500 (EST)


I wanted to compare the SMP statistics for 2.0 and 2.1 series
kernels so I turned on the profiling and took a look. The results
don't look quite right to me however. Am I missing something here?
Or are the 2.1 kernels only supposed to show total interrupts and
not per CPU?

Emil

>From 2.1.78
CPUS: 2
SUM P0 P1
0: 233374 0 0 timer
1: 5055 0 0 keyboard
2: 0 0 0 cascade
3: 37870 0 0 serial
13: 4 0 0 fpu
14: 39074 0 0 BusLogic BT-958
15: 9 0 0 ide1
LCK: 0 0 0 spins from int
LCK: 0 0 0 spins from syscall
LCK: 0 0 0 spins from sysidle
IDLE 0 0 0 idle ticks
TICK 466720 233360 233360 local APIC timer ticks
MULT: 1 1 profiling multiplier
COUNT: 1 1 profiling counter
IPI: 0 received

and for 2.0.33
CPUS: 2
SUM P0 P1
0: 51339 51339 0 timer
1: 1287 1287 0 keyboard
2: 0 0 0 cascade
3: 15102 15102 0 + serial
4: 13589 13589 0 + serial
7: 73 73 0 DS21041 Tulip
8: 1 1 0 + rtc
13: 24550 10328 14222 + IPI
14: 6476 6476 0 + BusLogic BT-958
15: 0 0 0 + ide1
LCK: 9373612 2039366 7334246 spins from int
LCK: 14831436 7601292 7230144 spins from syscall
LCK: 3348152 1825244 1522908 spins from sysidle
IDLE 97504 48851 48653 idle ticks
IPI: 24550 received