- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
I'm trying to get wait time and wait count data from some openmp applications, but it return always zero when using locksandwaits or N/A using the concurrency option.
The compiler I'm using is gcc with this command line:
gcc app.c -o app -fopenmp -lm -g
And this line to run:
amplxe-cl -collect locksandwaits -app-working-dir /my_dir -- ./app NumThreads
I also did:
export KMP_FORKJOIN_FRAMES=1
There's this warning: amplxe:
Warning: Cannot locate debugging symbols for file `/opt/intel/vtune_amplifier_xe_2016.3.0.463186/lib64/libtpsstool.so'.
The output I get is like:
Summary
-------
Average Concurrency: 3.983
Elapsed Time: 3.177
CPU Time: 12.620
Wait Time:
Average CPU Usage: 3.835
It works with pthreads, mpi-1 and mpi-2 applications, but doesn't with openmp.
So, anyone knows the way to solve this problem with openmp or another way to do it?
Thanks!
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello Adriano,
Meanwhile I'm looking at this, do you have ability to use intel compiler to compile your OpenMP application? If so - it is recommended to use "-c hpc-performance" to get metrics on OpenMP usage efficiency including serial time, imbalance, locks etc.
Thanks & Regards, Dmitry
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello Adriano,
Meanwhile I'm looking at this, do you have ability to use intel compiler to compile your OpenMP application? If so - it is recommended to use "-c hpc-performance" to get metrics on OpenMP usage efficiency including serial time, imbalance, locks etc.
Thanks & Regards, Dmitry

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page