|
In PR 23925 the tests for allowed ULP error for sin and cos were tightened for float32 and float64 to 1.49 and 1.00 respectively. The use of SVML/SIMD routines for sin/cos on float64 was disabled across the board, and the checks all pass. People using the latest HEAD version should be able to try out the code. People using the wheels from https://anaconda.org/scientific-python-nightly-wheels/numpy will have to wait a few days for a new upload. Please try it out and let us know if the slowdown is noticeable, and if the accuracy/precision is now as expected. Matti |
_______________________________________________ NumPy-Discussion mailing list -- numpy-discussion@python.org To unsubscribe send an email to numpy-discussion-le...@python.org https://mail.python.org/mailman3/lists/numpy-discussion.python.org/ Member address: arch...@mail-archive.com
