Fail at seeing transitions in OpenMP application

Fail at seeing transitions in OpenMP application

Hi,

I am a newbie as far as as VTune is concerned. I have an OpenMP application that I am trying to analyse, since the delivered performance is way too far from expected. However, when I ran "Locks and Waits" and "Thread Concurency" I get no transitions whatsoever. The check box is selected, though. Please look at the following picture:

[url=http://postimg.org/image/flogeemu5/full/][img]http://s14.postimg.org/crl...
[url=http://postimage.org/index.php?lang=german]kostenlose bilder[/url]

Additionally, I have different execution times than in "Thread Concurrency" - the latter match the execution times I have in practice.

I am using OpenMP to create threads (I have some sync points that are not shown at all by VTune) and I am compiling the code without optimizations - O0 - and g++.

Anything I am doing wrong?

Thanks

4 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.

I am sorry about the picture, here it goes:

I don't know why you can't see transition, but I tried my example without any problem. 

Have you tried latest VTune Amplifier XE 2013 Update 15? I also used Intel C/C++ compiler.

[root@prc-mic01 pi_solution]# icc -g omp_pi.c -openmp -openmp-report -o omp_pi
omp_pi.c(16): (col. 1) remark: OpenMP DEFINED LOOP WAS PARALLELIZED
[root@prc-mic01 pi_solution]# export  KMP_FORKJOIN_FRAMES=1
[root@prc-mic01 pi_solution]# amplxe-cl -c concurrency -- ./omp_pi

Please try attached example, I am attaching screen shot on my side. 

Attachments: 

AttachmentSize
Downloadimage/png concurrency_0.png119.33 KB
Downloadtext/x-csrc omp_pi.c501 bytes

Hi Artur:

Are you using the Intel compiler?  Did you check the "Analyze Intel runtimes and user synchronization" checkbox in the analysis configuration?

Attachments: 

AttachmentSize
Downloadimage/png analyzerintelruntimes.PNG63.61 KB

Leave a Comment

Please sign in to add a comment. Not a member? Join today