[pypy-issue] Issue #2087: NumPy Any/All character and (pypy/pypy)

Richard Plangger issues-reply at bitbucket.org
Mon Jul 20 17:26:47 CEST 2015


New issue 2087: NumPy Any/All character and
https://bitbucket.org/pypy/pypy/issues/2087/numpy-any-all-character-and

Richard Plangger:

>>> numpy.all([1,2,256]*1000) # pypy 2.6.0
False
>>> numpy.all([1,2,256]*1000) # python 2.7.10
True

That is the trace:

label(p27, i26, p1, i32, i9, i33, descr=TargetToken(51234744))
(numpy_reduce: no get_printable_location)
guard_early_exit(descr=<Guard0x26ca860>)
i36 = raw_load(i9, i32, descr=<ArrayS 8>)
guard_not_invalidated(descr=<Guard0x26ca8c0>)
i37 = i36 & 255
guard(i37 is true)
i38 = getarrayitem_raw(140216103876730, 1, descr=<ArrayU 1>)
guard(i38 == 0)
i39 = getarrayitem_raw(140216103876728, 0, descr=<ArrayU 1>)
i40 = i26 + 1
i41 = i32 + 8
i42 = i40 >= i33
guard(i42 is false) show bridge  (run 9800 times, ~0%)
guard(i39 == 1)
(numpy_reduce: no get_printable_location)
jump(p27, i40, p1, i41, i9, i33, descr=TargetToken(51234744))




More information about the pypy-issue mailing list