Graham Markall comments

Results 455 comments of


                                            Graham Markall

@testhound I understand that `numba.cuda.tests.cudapy.test_operator.TestOperatorModule.test_fp16_comparison` failed on the buildfarm because this PR accidentally removes the decorator to skip it when the CC is less than 5.3 Could you add it...

Testhound/fp16 operators1

@esc Could this have another buildfarm run please?

Reorder typeof checks to avoid infinite loops on StructrefProxy hash

Also needs a quick flake8 fix.

Add get_const_mem_size method to Dispatcher

gpuci run tests

Use `omp_set_max_active_levels()` instead of `omp_set_nested()`

(cc @seanlaw)

Use `omp_set_max_active_levels()` instead of `omp_set_nested()`

Argh: "Visual C++ supports the OpenMP 2.0 standard." - https://docs.microsoft.com/en-us/cpp/parallel/openmp/openmp-in-visual-cpp?view=msvc-170

Use `omp_set_max_active_levels()` instead of `omp_set_nested()`

The Linux failures look pretty weird, but I wonder if this is just a knock-on effect of the function not being supported in the environmentfor that test: Errors ``` Parallel:...

Use `omp_set_max_active_levels()` instead of `omp_set_nested()`

@stuartarchibald Shall we give this a buildfarm run? I'd like to get some disappointments in early on in the year.

Use `omp_set_max_active_levels()` instead of `omp_set_nested()`

> I wonder if at compile time the OpenMP version could be detected and added to the current 2 vs 3+ split implied by _MSC_VER, this to accommodate e.g. software...