There is a failure due to -Werrror=unused-function

In file included from ../numpy/_core/src/npysort/x86_simd_qsort_16bit.dispatch.cpp:4:
../numpy/_core/src/npysort/x86-simd-sort/src/x86simdsort-static-incl.h:193:6: error: ‘void x86simdsortStatic::partial_qsort(T*, size_t, size_t, bool, bool) [with T = _Float16]’ defined but not used [-Werror=unused-function]
  193 | void x86simdsortStatic::partial_qsort<_Float16>(

There is a failure due to -Werrror=unused-function

In file included from ../numpy/_core/src/npysort/x86_simd_qsort_16bit.dispatch.cpp:4:
../numpy/_core/src/npysort/x86-simd-sort/src/x86simdsort-static-incl.h:193:6: error: ‘void x86simdsortStatic::partial_qsort(T*, size_t, size_t, bool, bool) [with T = _Float16]’ defined but not used [-Werror=unused-function]
  193 | void x86simdsortStatic::partial_qsort<_Float16>(

thanks! Hopefully it should be fixed now.

LGTM. The previous commit was c306ac581a59f89585d778254c4ed7197e64ba2d, and there seems to be many commits since, mainly around enabling OPENMP threading which is not relevant for NumPy. Does the update influence any perfomance benchmarks? If so, maybe a release note would be nice.

Thanks for catching that, @mattip. You’re right — this patch enables OpenMP support for the 16-bit dtype. That does assume NumPy is built with OpenMP, which is disabled by default. Would you like me to add a release note about this?

Not specifically about OpenMP, but if there are other performance enhancements then maybe a small release note would be nice. I will merge this to unblock the bug-fix.