Charles Harris
Charles Harris
Well, here goes. Thanks @DimitriPapadopoulos . Would be good to get those linter failures fixed up.
Same at https://github.com/numpy/numpy/actions/runs/11110913872/job/30870108510?pr=27486
Another case at https://github.com/numpy/numpy/actions/runs/14505328602/job/40693911668. Checking by hand, the expected value is correct and the returned value differs by 1 ulp. However, the result is in the range where there are...
And another https://github.com/numpy/numpy/actions/runs/14541363659/job/40800139903?pr=28771. These are no longer rare, we might want to look into them.
Is this a regression?
@h-vetinari AVX512 was disabled on 32 bit windows in #29910. @seiko2plus thought that was sufficient.
How does argsorting compare? Do you actually move strings around, or just change the offsets?
Closing, the current spin test seems adequate.
Thanks @mdavidsaver .
@seberg I put up a backport for this (#27397) in order to simplify some other backports.