Application Performance Snapshot (APS) Playbook 
===============================================

The Playbook contains command lines starting with $
Please change $PRG, $ARGS into the path,name and parameters of your program! 

Version 0.9, 08.10.2018

Please send feedback to Heinrich.Bockhorst@Intel.com

0. Environment
--------------

Please use the provided script: env_sh
---------------------------------------

$ source <path to parallel studio>/psxevars.sh

(works for all tools)

or load module for Vtune

check for important executables

$ which aps

check version 

$ aps -version

Alternative Environment: PSXE 2019 (see below) 

1.0 Application Performance Snapshot (APS) usage:
==================================================

Just put aps in front of the executable:

$ aps $PRG $ARGS

For MPI programs:

$ mpirun -n <N>  aps -r report $PRG $ARGS


Poisson Example:

mpirun -n <N> aps -r report ./poisson.x -n 3200

Output directory name "report" may be omitted or chosen differently

Note: check with $ ls -ltr 
for the last created directory. 

2.0 View Results
================

Summary results can be displayed by 

$ aps-report -s -D report

result-dir is starting with "aps_" if result dir is not specified.

HTML results

$ aps-report -g -D report

generates aps_report_<date>_<time>.html

more MPI statistics (functions) are available.

$ aps-report -f -D report 

or for the full output: 

$ aps-report -a -D report

For more detailed MPI output the program has to run under the environment variable:

$ export MPS_STAT_LEVEL=2

The integer value may be raised to 3 or 4 for even more infos. 

3.0 Code block for analysis may be selected
-------------------------------------------

Insert MPI_Pcontrol(0) right after MPI_Init() to switch off tracing 
Insert MPI_Pcontrol(1) before code block      to switch on tracing
Insert MPI_Pcontrol(0) after code block       to switch off tracing 

see: https://software.intel.com/en-us/get-started-with-application-performance-snapshot

MPI_Pcontrol will be applied only on the MPI part. For limiting the HW counters use the _itt 
library found on web page above