A newer version of this document is available. Customers should click here to go to the newest version.
PCTRACE [on|off|<trace levels>|<skip levels>:<trace levels>]
off for performance analysis, on otherwise
Some platforms support the automatic stack sampling for MPI calls and user-defined events. Intel Trace Collector then remembers the Program Counter (PC) values on the call stack and translates them to source code locations based on debug information in the executable. It can sample a certain number of levels (<trace levels>) and skip the initial levels (<skip levels>). Both values can be in the range of 0 to 15.
Skipping levels is useful when a function is called from within another library and the source code locations within this library shall be ignored. ON is equivalent to 0:1 (no skip levels, one trace level).
The value specified with PCTRACE applies to all symbols that are not matched by any filter rule or where the relevant filter rule sets the logging state to ON. In other words, an explicit logging state in a filter rule overrides the value given with PCTRACE.
Did you find the information on this page useful?