How can i do a selective profiling of my code using intel vtune or advisor. I dont want to perform the profiling of my full code. is there any options to do it for ex. encoding the section inside some pragma to tell vtune/advisor to do profiling of particular section of code
Hello!
The most common method, applicable to both advisor and vtune, is using itt_pause() / itt_resume() API calls. Quick summary with example can be found here: https://software.intel.com/content/www/us/en/develop/articles/intel-advisor-collection-control-apis.html
Furthermore, selective profiling is an area where intel tools and specifically Intel Advisor have invested a lot and there are many different techniques provided.
So for Advisor please take a look at relatively complete materials with 5 sub-chapters at: https://software.intel.com/content/www/us/en/develop/documentation/advisor-user-guide/top/minimizing-analysis-overhead.html As you can see - with advisor you can select specific modules/libraries to be profiled or you can "mark-up" different loops and functions for profiling even without recompiling your code. And for itt_pause/itt_resume - both vtune and advisor have more fine-tuning usage models (like start-paused mode), all described at the link above.
Another article (advisor-specific) on the topic is: https://software.intel.com/content/www/ru/ru/develop/articles/managing-overhead-of-intel-advisor-analyses.html
链接已复制
Hello!
The most common method, applicable to both advisor and vtune, is using itt_pause() / itt_resume() API calls. Quick summary with example can be found here: https://software.intel.com/content/www/us/en/develop/articles/intel-advisor-collection-control-apis.html
Furthermore, selective profiling is an area where intel tools and specifically Intel Advisor have invested a lot and there are many different techniques provided.
So for Advisor please take a look at relatively complete materials with 5 sub-chapters at: https://software.intel.com/content/www/us/en/develop/documentation/advisor-user-guide/top/minimizing-analysis-overhead.html As you can see - with advisor you can select specific modules/libraries to be profiled or you can "mark-up" different loops and functions for profiling even without recompiling your code. And for itt_pause/itt_resume - both vtune and advisor have more fine-tuning usage models (like start-paused mode), all described at the link above.
Another article (advisor-specific) on the topic is: https://software.intel.com/content/www/ru/ru/develop/articles/managing-overhead-of-intel-advisor-analyses.html
Hi we have accepted your issue. Please let us know if you have any further questions.
This issue has been resolved and support will no longer respond to this thread. If you require additional assistance from Intel, please start a new thread. Any further interaction in this thread will be considered community only