Open
Description
This is a reminder that we have seen a few inconsistencies in the floating-point pow function on linux vs windows vs darwin.
The problems:
- The operation
10.**arange(308)on linux correctly gives the nearest IEEE float64 to all possible powers of 10, with unbiased rounding. In other words we have10.**np.float64(i) == np.float64('1e%d' %i)for alli. On some windows systems this is often not the case fori >= 23(where float64 values are no longer integers), where it gives a slightly different value. I put a list of appveyor envs where this fails below marked with X. See ENH: Use Dragon4 algorithm to print floating values #9941 (comment) - The mac
powfunction fails for2.**expwithexp >= 1024. See BUG: Fix casting from longdouble to long #9971 (comment) MAINT: Fix test_int_from_huge_longdouble on Darwin. #10000
PYTHON=C:\Python27, PYTHON_VERSION=2.7, PYTHON_ARCH=32, SKIP_NOTAG=true, TEST_MODE=full
PYTHON=C:\Python36, PYTHON_VERSION=3.6, PYTHON_ARCH=32, TEST_MODE=fast
X PYTHON=C:\Python27-x64, PYTHON_VERSION=2.7, PYTHON_ARCH=64, TEST_MODE=fast
X PYTHON=C:\Python34-x64, PYTHON_VERSION=3.4, PYTHON_ARCH=64, TEST_MODE=fast
X PYTHON=C:\Python36-x64, PYTHON_VERSION=3.6, PYTHON_ARCH=64, TEST_MODE=full
PYTHON=C:\Python27, PYTHON_VERSION=2.7, PYTHON_ARCH=32, SKIP_NOTAG=true, TEST_MODE=full
PYTHON=C:\Python34, PYTHON_VERSION=3.4, PYTHON_ARCH=32, SKIP_NOTAG=true, TEST_MODE=full
PYTHON=C:\Python35-x64, PYTHON_VERSION=3.5, PYTHON_ARCH=64, SKIP_NOTAG=true, TEST_MODE=full
PYTHON=C:\Python35, PYTHON_VERSION=3.5, PYTHON_ARCH=32, SKIP_NOTAG=true, TEST_MODE=full
from this test
Metadata
Metadata
Assignees
Labels
No labels