Intel® Trace Analyzer and Collector

Profiling System Calls

On Linux* OS, use this capability to track I/O calls.

By default, system call profiling is off. To collect system calls, add the following line into a configuration file:


To enable collecting of an exact function add the following line into a configuration file:

STATE SYSTEM:<func_name> ON



int VT_countsetcallback( VT_CountCallback_t callback, void *custom, int ncounters )


Sets a callback for counter sampling for the calling thread.

The data provided by the callback is logged with the same time stamp as the event that triggered the callback. The callback must be set for each thread individually. Setting NULL disables sampling for the thread.

Level of Detail

Tracing all available events over time can generate billions of events even for a moderate program runtime of a few minutes and a handful of CPUs. The sheer amount of data is a challenge for any analysis tool that has to cope with this data. This is even worse as in most cases the analysis tool cannot make use of the same system resources as the parallel computer on which the trace was generated.

Filtering Events

You can filter which events are displayed in the views. If you use filtering, you see only those events that satisfy specific condition(s). All other events are suppressed as if they were not written to the trace file.

To define the filter conditions, select Advanced > Filtering.

To remove all defined filters and tags, select Advanced > Reset Tagging/Filtering. This command removes all tags and filters.

Filtering and Tagging

In the Event Timeline, you see the tagged functions in a frame and with their function labels in bold. The lines of the tagged messages and collective operations are thicker.

When you filter the Event Timeline for a particular process, you see only the MPI functions of this process, since only they pass the filter and all the other functions and processes are filtered out.

Assine o Intel® Trace Analyzer and Collector