[pypy-issue] [issue1279] Numerical code for list of floats 3x slower than for list of ints.

Jorge tracker at bugs.pypy.org
Sat Oct 6 19:00:05 CEST 2012

New submission from Jorge <jorgecanardo at gmail.com>:

The attached file is an extract of a very simple matrix class in Python.

I'm comparing its performance to a similar class in C++ by multiplying 500x500 
matrices. Timings were similar when the python matrices had only ints (C++ -O1 
0.9 s; PyPy 1.9.0 1.5 s) but using floats in the python side turned the 
multiplication over 3 times slower (!), much slower than even -O0. Does this 
make sense?

(As a workaround, I'm using array.array('d', ...), which is actually slightly 
faster than the int-only list (1.3 s).)

nosy: +jorge
status: unread -> chatting

PyPy bug tracker <tracker at bugs.pypy.org>

More information about the pypy-issue mailing list