PAPI_hl_region_end - Man Page

Read performance events at the end of a region and store the difference to the corresponding beginning of the region.


Detailed Description

C Interface:

#include <papi.h>
int PAPI_hl_region_end( const char* region );


region -- a unique region name corresponding to PAPI_hl_region_begin

Return values

PAPI_ENOTRUN -- EventSet is currently not running or could not determined.
PAPI_ESYS -- A system or C library call failed inside PAPI, see the errno variable.
PAPI_EMISC -- PAPI has been deactivated due to previous errors.
PAPI_ENOMEM -- Insufficient memory.

PAPI_hl_region_end reads performance events at the end of a region and stores the difference to the corresponding beginning of the region.

Assumes that PAPI_hl_region_begin was called before.

Note that PAPI_hl_region_end does not stop counting the performance events. Counting continues until the application terminates. Therefore, the programmer can also create nested regions if required. To stop a running high-level event set, the programmer must call PAPI_hl_stop(). It should also be noted, that a marked region is thread-local and therefore has to be in the same thread.

An output of the measured events is created automatically after the application exits. In the case of a serial, or a thread-parallel application there is only one output file. MPI applications would be saved in multiple files, one per MPI rank. The output is generated in the current directory by default. However, it is recommended to specify an output directory for larger measurements, especially for MPI applications via the environment variable PAPI_OUTPUT_DIRECTORY. In the case where measurements are performed, while there are old measurements in the same directory, PAPI will not overwrite or delete the old measurement directories. Instead, timestamps are added to the old directories.

For more convenience, the output can also be printed to stdout by setting PAPI_REPORT=1. This is not recommended for MPI applications as each MPI rank tries to print the output concurrently.

The generated measurement output can also be converted in a better readable output. The python script enhances the output by creating some derived metrics, like IPC, MFlops/s, and MFlips/s as well as real and processor time in case the corresponding PAPI events have been recorded. The python script can also summarize performance events over all threads and MPI ranks when using the option 'accumulate' as seen below.


int retval;

retval = PAPI_hl_region_begin("computation");
if ( retval != PAPI_OK )

 //Do some computation here

retval = PAPI_hl_region_end("computation");
if ( retval != PAPI_OK )
python --type=accumulate

   "computation": {
      "Region count": 1,
      "Real time in s": 0.97 ,
      "CPU time in s": 0.98 ,
      "IPC": 1.41 ,
      "MFLIPS /s": 386.28 ,
      "MFLOPS /s": 386.28 ,
      "Number of ranks ": 1,
      "Number of threads ": 1,
      "Number of processes ": 1
See also





Generated automatically by Doxygen for PAPI from the source code.


Thu Jul 20 2023 00:00:00 Version PAPI