ofxProfile

by Damian Stewart <http://damianstewart.com>

A thread-safe profiler. Stores and manages execution count, total execution time and average time per execution of each labelled section, separated by thread.

To use in a particular cpp file, you need to #define PROFILE either globally or in that file, and #include the ofxProfile.h header.

define PROFILE

include "ofxProfile.h"

Then use the following macros to profile parts of code:

ofxProfileSectionPush(label) and ofxProfileSectionPop()

Profile code between these calls, storing results in a tree. For example:

.. unprofiled preprocessing ofxProfileSectionPush("partial section"); .. code to be profiled ofxProfileSectionPop(); .. unprofiled postprocessing

Or for nested output:

ofxProfileSectionPush("two-step process"); .. preprocessing code ofxProfileSectionPush("step 1"); .. step 1 code ofxProfileSectionPop(); ofxProfileSectionPush("step 2"); .. step 2 code ofxProfileSectionPop(); ofxProfileSectionPop();

will produce output like this: + two-step process <total time for steps 1 + 2 + preprocessing> ... + step 1 <time for step 1> ... + step 2 <time for step 2> ...

NOTE: all labels at a given level in the tree must be unique.

ofxProfileThisFunction()

Automatically wraps the current function in a pair of ofxProfileSectionPush(functionName) and ofxProfileSectionPop() calls.

ofxProfileThisBlock(label)

Wraps the current block (the code between the current level { and }) in a pair of ofxProfileSectionPush(label) and ofxProfileSectionPop() calls.

eg:

if ( test_condition ) { ofxProfileThisBlock( "test passed" ); // code to run on test condition }

To display profile results, call ofxProfile::display().

To fetch results as a string, call ofxProfile::describe().

To clear current results, call ofxProfile::clear(). Notes If PROFILE is not #defined, the profiling code is not executed. This means you can leave profiling code in place for release builds and not have it impact on performance. Profiling is invasive, this means it will have an impact on execution times. Output from ofxProfile is therefore not 100% accurate but is usually good enough to help nail down performance bottlenecks. OpenGL may do all sorts of interesting things behind your back, that ofxProfiler doesn't know about. If you want to profile OpenGL calls please use a proper OpenGL profiler instead.

original GitHub url: <http://github.com/damiannz/ofxProfile>