In general, vtune drivers will be built and loaded to the Linux* system automatically during installing VTune™ Amplifier XE product, then hardware PMU event-based sampling can work.
However sometime, vtune drivers were built/loadeded unsuccessfully, because of one of below reason:
1. There was no Linux* kernel source, required by building vtune drivers
2. The user has no “root” privilege to load vtune drivers to system
3. There is no development tool to build vtune drivers, either the user works on their customized operation system.
4. Customer are not familiar with VTune command line, but be familiar with Linux* Perf tool.
Since VTune Amplifier XE 2015 initial release, command amplxe-perf provided by VTune can support Perf’s functionality seamlessly – please refer to this article, which teach you how to use Perf utility within VTune.
In VTune Amplifier XE 2015 Update 2, amplxe-perf command is not required again, simply use amplxe-cl to collect performance data of Perf via VTune, when there is no vtune driver in system (otherwise, will do regular PMU Event-based sampling collections).
Another key points is, using amplxe-cl to collect hotspots from Perf also supports call stack info gathering.
First at all, check there is no vtune driver: lsmod | grep sep, lsmod | grep pax, lsmod | grep vtsspp
If there are drivers, you can do – vtune_amplifier_xe/sepdk/src/rmmod-sep3, unload all vtune drivers
$ amplxe-cl -collect advanced-hotspots –knob collection-detail=stack-sampling -- ./primes.gcc
$ amplxe-cl -R callstacks
amplxe: Using result path `/home/peter/r009ah'
amplxe: Executing actions 50 % Generating a report
Function Function Stack Module CPU Time
---------- -------------- ---------- --------
findPrimes primes.gcc 2.000s
start_thread primes.gcc 0.564s
__clone primes.gcc 0s
Note: please use this feature on latest operation system, which was released after 02/04/2014, and you can check supported events at /sys/devices/cpu/events in your system.
