Context Navigation

Changes between Version 3 and Version 4 of MaskedArrayAlternative

v3	v4
138	138
139	139	A quick benchmark gives the following results:
140		* {{{numpy.ma.divide}}} : 2.84 ms per loop
141		* classical division : 2.99 ms per loop
142		* division w/ prefilling : 2.20 ms per loop
143		* division w/o filling : 1.54 ms per loop
	140	* {{{numpy.ma.divide}}} : 2.69 ms per loop
	141	* classical division : 2.21 ms per loop
	142	* division w/ prefilling : 2.34 ms per loop
	143	* division w/o filling : 1.55 ms per loop
144	144
145	145	So, is it worth filling the arrays beforehand ? Yes, if we are interested in avoiding floating-point exceptions that may fill the result with infs and nans. No, if we are only interested into speed...