System information

Perfmon2—Hardware-Based Performance Monitoring 99

pfmon --cpu-list=0-1 --system-wide -u -e

CPU_OP_CYCLES_ALL,IA64_INST_RETIRED -- ls -l /dev/null

crw-rw-rw- 1 root root 1, 3 27. Mär 03:30 /dev/null

CPU0 38925 CPU_OP_CYCLES_ALL

CPU0 7510 IA64_INST_RETIRED

CPU1 9825 CPU_OP_CYCLES_ALL

CPU1 1676 IA64_INST_RETIRED

3 Press the Enter key to stop a session:

4 If you want to aggregate counts, use the -aggr option after the previous com-

mand:

pfmon --cpu-list=0-1 --system-wide -u -e

CPU_OP_CYCLES_ALL,IA64_INST_RETIRED --aggr

52655 CPU_OP_CYCLES_ALL

53164 IA64_INST_RETIRED

7.3.3 Monitoring Running Tasks

Perfmon can also monitor an existing thread. This is useful for monitoring system

daemons or programs which take a long time to start. First determine the process ID

you wish to monitor:

ps ax | grep foo

10027 pts/1 R 2:23 foo

Use the found PID for the --attach-task option of pfmon:

pfmon --attach-task=10027

3682190 CPU_OP_CYCLES_ALL

7.4 Retrieving Metrics From

DebugFS

Perfmon can collect statistics which are exported through the debug interface. The

metrics consists of mostly aggregated counts and durations.

Access the data through mounting the debug file system as root under /sys/ker

nel/debug