Charles Harris

Results 283 comments of Charles Harris

@mayeut Thanks. Is it available as plain old `"python3.11"`?

NumPy 1.23.2 has been released with Python 3.11 wheels.

> Any reason someone knows of not to use plain inline in SciPy? The new C++ code uses the standard `inline` and I suggest using it for new code. We...

This sounds like an Anaconda BLAS/LAPACK problem, and possibly a problem with the Apple Accelerate library.

> if it does bump the gfortran version. NumPy 1.23.x is using 0.3.20, do we need to worry, or is our gfortran version used for wheels sufficient?

Who can check? I note that 1.23.2 didn't seem to have the problem (?) and I don't think anything should have changed since then.

@isuruf Thanks for checking. I'll be releasing 1.23.3 this weekend, so that feedback is helpful.

> It was a gfortran compiler bug. Going to close this. @isuruf Do you recall the compiler version?

> Maybe this should get a release note? I just note these as "continuing SIMD improvements" :) A release note wouldn't hurt as FP16 improvements are new.