Development Team/Performance

From Sugar Labs
Jump to: navigation, search

Just a quick dump of the snippets I use for profiling:

First use a system-wide profiler like sysprof

Next, profile python code, open the output with kcachegrind:

import os
import cProfile
import lsprofcalltree  # http://www.gnome.org/~johan/lsprofcalltree.py

profiler = cProfile.Profile()
profiler.enable()

### code to profile ### 

profiler.disable()

k = lsprofcalltree.KCacheGrind(profiler)
data = open('/tmp/import.kgrind', 'w+')
k.output(data)
data.close()

Check for leaks:

        import gobject
        gobject.timeout_add(10000, self._log_mem_usage) 

    def _log_mem_usage(self): 
        import analysis, os, gc 
        
        for obj in gc.get_objects():
            if hasattr(obj, '__class__') and 'ChooserListView' in obj.__class__.__name__:
                logging.debug(obj) 

        for i in range(0, 10):
            gc.collect()
        logging.debug('mem: %i' % analysis.Analysis(os.getpid()).ApproxRealMemoryUsage()) 
        return True 

Rendering time of a widget/window:

    def do_expose_event(self, event):
        import time
        t = time.time()
        hippo.Canvas.do_expose_event(self, event)
        logging.debug('listview expose event: %r' % (time.time() - t))