diff options
| author | Developer-Ecosystem-Engineering <65677710+Developer-Ecosystem-Engineering@users.noreply.github.com> | 2022-02-01 10:59:29 -0800 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2022-02-01 12:59:29 -0600 |
| commit | 49c560c22f137907ea6a240591e49b004f28444b (patch) | |
| tree | ce9096f4242d6a42a15baabfc51b5ca010e8a2df /numpy/array_api/_creation_functions.py | |
| parent | b6e6b838f1dd5f689ef8bac7a40ee814be7cd362 (diff) | |
| download | numpy-49c560c22f137907ea6a240591e49b004f28444b.tar.gz | |
BUG: Loss of precision in longdouble min (#20872)
* BUG: Loss of precision in longdouble min
Generic reuse in the latest changes around `min` works unless the macro is redefined for SIMD.
This change avoids `scalar_min_f` for generic comparisons (as it can be redefined) and defines is separately as `scalar_min`
* Add tests as requested
Adding tests like suggested in https://github.com/numpy/numpy/issues/20863
* MAINT: Use `np.longdouble` rather than float128 for min/max tests
Co-authored-by: Sebastian Berg <sebastian@sipsolutions.net>
Diffstat (limited to 'numpy/array_api/_creation_functions.py')
0 files changed, 0 insertions, 0 deletions