kostia.lopuhin <[email protected]> added the comment: On the code from archive I get 26 s for PyPy nighly (OS X 64 bit) and 16 s for CPython 2.7. If I put Point and _get_high_values into top-level scope, then PyPy nighly is much slower (89 s vs 16 for CPython). But then if I replace itememgetter(x) for _clock and _value definitions with lambda y: x[y], or move _clock and _value definitions into top level, than PyPy takes 17 s (and by the way PyPy 2.0 and 2.1 take 13 seconds on this last version). So, one of the reasons for PyPy slowness here might be itemgetter, defined inside a hot function.
---------- nosy: +kostia.lopuhin ________________________________________ PyPy bug tracker <[email protected]> <https://bugs.pypy.org/issue1652> ________________________________________ _______________________________________________ pypy-issue mailing list [email protected] https://mail.python.org/mailman/listinfo/pypy-issue
