I have been profiling the same FORTRAN OpenMP code on 4 threads on a node comprising 2x 6-core Westmere, under the SGE batch system where I have reserved all 12 cores exclusively for my use.
Out of my 20 runs, I have one that failed with the following to standard error:
amplxe: Collection started. To stop the collection, either press CTRL-C or enter from another console window: amplxe-cl -r /mnt/iusers01/support/mccssmb2/ResearchIT/applications_support/Popelier/ferebus/vtune/various_tests/Schedule/r029hs -command stop.
amplxe: Error: [Instrumentation Engine]: Function IMG_FindByUnloadAddress called without holding lock. Call PIN_LockClient()/PIN_UnlockClient()
amplxe: Collection failed.
amplxe: Internal Error
I am running Amplifier XE version:
$ amplxe-cl --version
Intel(R) VTune(TM) Amplifier XE 2015 (build 367959) Command Line Tool
Copyright (C) 2009-2014 Intel Corporation. All rights reserved.
The code was compiled with ifort version:
$ ifort --version
ifort (IFORT) 14.0.3 20140422
Copyright (C) 1985-2014 Intel Corporation. All rights reserved.
and the Linux version on the compute nodes is
$ uname -a && cat /etc/*release
Linux int00 2.6.32-358.18.1.el6.x86_64 #1 SMP Tue Aug 27 14:23:09 CDT 2013 x86_64 x86_64 x86_64 GNU/Linux
Alces Core HPC Configuration package release 3.0
Scientific Linux release 6.2 (Carbon)
All help appreciated. Yours, M
BTW - with VTune Amplifier XE 2015 Update 1 you can try the new OpenMP efficiency analysis feature with Advanced Hotspots. The details can be found here: https://software.intel.com/en-us/node/529832. Advanced Hotspots are less intrusive and allows to sample with less sampling interval that can be useful if elapsed time of a parallel regions instance is short.