Profiling.rst

Profiling

> This page is very out-of-date and has some inconsistencies with current versions

The following block of code is profiled by :pycProfile:

../tool/profQCLayers.py

where PQLiu.json is the following file:

../ErwinJr2/example/PQLiu.json

Current version with population and gain

ncalls	tottime	percall	cumtime	percall	filename:lineno(function)
1	0.001	0.001	5.955	5.955	./ErwinJr2/QCLayers.py:1493(auto_gain)
1	0.000	0.000	4.753	4.753	./ErwinJr2/QCLayers.py:1386(full_population)
1	0.024	0.024	4.753	4.753	./ErwinJr2/QCLayers.py:921(full_population)
3640	1.358	0.000	3.361	0.001	./ErwinJr2/QCLayers.py:701(_ifr_transition)
14428	0.014	0.000	1.545	0.000	./ErwinJr2/QCLayers.py:433(psi_overlap)
14436	0.316	0.000	1.531	0.000	{built-in method builtins.sum}
3640	0.038	0.000	1.367	0.000	./ErwinJr2/QCLayers.py:621(_lo_transition)
57712	0.292	0.000	1.216	0.000	./ErwinJr2/QCLayers.py:437(<genexpr>)
45069	0.090	0.000	1.024	0.000	./ErwinJr2/QCLayers.py:273(_shift_psi)
12164	0.012	0.000	0.935	0.000	<__array_function__ internals>:2(interp)
12164	0.015	0.000	0.914	0.000	[.python-packages]/numpy/lib/function_base.py:1289(interp)
12164	0.871	0.000	0.871	0.000	{built-in method numpy.core._multiarray_umath.interp}
1754	0.783	0.000	0.800	0.000	./ErwinJr2/OneDQuantum/OneDSchrodinger.py:144(cLOphononScatter)
1	0.017	0.017	0.678	0.678	./ErwinJr2/QCLayers.py:1394(full_gain_spectrum)
1785	0.078	0.000	0.656	0.000	./ErwinJr2/QCLayers.py:587(_dipole)
1	0.002	0.002	0.513	0.513	./ErwinJr2/QCLayers.py:322(solve_whole)
1	0.036	0.036	0.510	0.510	./ErwinJr2/QCLayers.py:344(_solve_whole_ode)
1	0.469	0.469	0.469	0.469	./ErwinJr2/OneDQuantum/OneDSchrodinger.py:101(cBandSolve1D)
343980	0.372	0.000	0.372	0.000	./ErwinJr2/QCLayers.py:732(interpZ)
116482	0.062	0.000	0.355	0.000	<__array_function__ internals>:2(argmax)
3570	0.281	0.000	0.281	0.000	./ErwinJr2/QCLayers.py:993(_xBandMassInv)
7179	0.005	0.000	0.248	0.000	<__array_function__ internals>:2(trapz)
7179	0.169	0.000	0.237	0.000	[.python-packages]/numpy/lib/function_base.py:4006(trapz)
116482	0.053	0.000	0.234	0.000	[.python-packages]/numpy/core/fromnumeric.py:1114(argmax)
116483	0.059	0.000	0.182	0.000	[.python-packages]/numpy/core/fromnumeric.py:52(_wrapfunc)
3570	0.002	0.000	0.138	0.000	<__array_function__ internals>:2(gradient)
3570	0.101	0.000	0.129	0.000	[.python-packages]/numpy/lib/function_base.py:802(gradient)
116482	0.103	0.000	0.103	0.000	{method 'argmax' of 'numpy.ndarray' objects}
7181	0.005	0.000	0.063	0.000	{method 'sum' of 'numpy.ndarray' objects}
232976	0.058	0.000	0.058	0.000	./ErwinJr2/QCLayers.py:1247(layerVc)
7181	0.003	0.000	0.058	0.000	[.python-packages]/numpy/core/_methods.py:45(_sum)
7210	0.056	0.000	0.056	0.000	{method 'reduce' of 'numpy.ufunc' objects}
12164	0.006	0.000	0.021	0.000	<__array_function__ internals>:2(iscomplexobj)

1.0 and earlier versions

Before optimizing code performamce, there are 814603 function calls taking 10.835 seconds in total. The output of :pycProfile sorted by cumulative time is listed in the following table.

ncalls	tottime	percall	cumtime	percall	filename:lineno(function)
1	0.000	0.000	10.803	10.803	QCLayers.py:425(calc_FoM)
35	9.742	0.278	10.803	0.309	QCLayers.py:383(loTransition)
2	0.000	0.000	10.449	5.224	QCLayers.py:419(loLifeTime)
2	0.000	0.000	10.449	5.224	QCLayers.py:422(<listcomp>)
135301	0.132	0.000	0.797	0.000	fromnumeric.py:1821(sum)
135303	0.155	0.000	0.640	0.000	fromnumeric.py:64(_wrapreduction)
135331	0.467	0.000	0.467	0.000	{method 'reduce' of 'numpy.ufunc' objects}
135180	0.262	0.000	0.262	0.000	{built-in method builtins.abs}
1	0.000	0.000	0.037	0.037	QCLayers.py:260(solve_whole)
1	0.036	0.036	0.036	0.036	OneDSchrodinger.py:61(cSimpleSolve1D)
135313	0.025	0.000	0.025	0.000	{built-in method builtins.isinstance}
135303	0.019	0.000	0.019	0.000	{method 'items' of 'dict' objects}
1	0.001	0.001	0.002	0.002	QCLayers.py:178(populate_x)
65	0.000	0.000	0.001	0.000	QCLayers.py:169(avghwLO)
1	0.001	0.001	0.001	0.001	OneDSchrodinger.py:74(cSimpleFillPsi)
227	0.000	0.000	0.001	0.000	{built-in method builtins.sum}
28	0.001	0.000	0.001	0.000	QCLayers.py:227(<listcomp>)
1885	0.001	0.000	0.001	0.000	QCLayers.py:173(<genexpr>)

ErwinJr2.QCLayers

After optimizing code performance by replace most time consuming function call (:pyQCLayers.loLifeTime), by C code and add openMP support, there are 3613 function calls taking 1.255 seconds in total. The output of :pycProfile sorted by cumulative time is listed in the following table.

ncalls	tottime	percall	cumtime	percall	filename:lineno(function)
1	0.000	0.000	1.205	1.205	QCLayers.py:427(calc_FoM)
35	0.003	0.000	1.205	0.034	QCLayers.py:383(loTransition)
30	1.199	0.040	1.199	0.040	OneDSchrodinger.py:113(cLOphononScatter)
2	0.000	0.000	1.166	0.583	QCLayers.py:421(loLifeTime)
2	0.000	0.000	1.165	0.583	QCLayers.py:424(<listcomp>)
1	0.000	0.000	0.048	0.048	QCLayers.py:260(solve_whole)
1	0.046	0.046	0.046	0.046	OneDSchrodinger.py:65(cSimpleSolve1D)
1	0.001	0.001	0.002	0.002	QCLayers.py:178(populate_x)
91	0.000	0.000	0.001	0.000	fromnumeric.py:1821(sum)
65	0.000	0.000	0.001	0.000	QCLayers.py:169(avghwLO)
227	0.000	0.000	0.001	0.000	{built-in method builtins.sum}
1	0.001	0.001	0.001	0.001	OneDSchrodinger.py:78(cSimpleFillPsi)
93	0.000	0.000	0.001	0.000	fromnumeric.py:64(_wrapreduction)
121	0.001	0.000	0.001	0.000	{method 'reduce' of 'numpy.ufunc' objects}
1885	0.001	0.000	0.001	0.000	QCLayers.py:173(<genexpr>)
68	0.000	0.000	0.001	0.000	ctypeslib.py:196(from_param)
28	0.000	0.000	0.000	0.000	QCLayers.py:227(<listcomp>)
65	0.000	0.000	0.000	0.000	{built-in method numpy.core.multiarray.array}