Graham Markall comments

Results 776 comments of


                                            Graham Markall

Clarify explanation of setup for `shutting_down()`

Thanks, I probably will do when I circle back to this.

Port refprune pass to NewPassManager

> Clang format does not report anything for me locally. What version of clang-format is used in the checks? It is clang-format-14, I think.

Add llvmlite usage example python notebook

@yashssh Is this still a draft, or ready for a review?

Add llvmlite usage example python notebook

I think that's OK - I think we can review it independently of the LLVM 19 support, and merge it with / after the LLVM 19 support is merged so...

`test_target_extension.TestTargetHierarchySelection.test_overload_allocation` fails when run in isolation

Also: https://github.com/numba/numba/blob/467136ab48871c36505ab54dda7eff49f1cc5eef/numba/tests/test_target_extension.py#L667-L668

`test_target_extension.TestTargetHierarchySelection.test_overload_allocation` fails when run in isolation

When the test is run in isolation, the wrong `intrin_alloc` is called. It is using https://github.com/numba/numba/blob/467136ab48871c36505ab54dda7eff49f1cc5eef/numba/np/arrayobj.py#L4302-L4313 instead of https://github.com/numba/numba/blob/467136ab48871c36505ab54dda7eff49f1cc5eef/numba/tests/test_target_extension.py#L660-L673

Move Azure to use macos-13

Looks like #1097 will be the way forward here.

Numpy vecdot support

Labelled as "good first issue" - this would require implementing an overload (`@overload`) with the High-level extension API, whose implementation is similar to / based on the example in the...

Source code location of IR instruction

My interpretation of this request is that we could support this by wrapping the [`getDebugLoc()`](https://llvm.org/doxygen/classllvm_1_1Instruction.html#a4b8faae4ff9e7434a1d226d03d15dcd2) function (and similar / related APIs as necessary). I don't think there's availability for anyone...

@jit(target_backend='cpu') select loop mode for huge for loop

If you're looking to run pandas code with CUDA, it may be worth looking at the cuDF Pandas accelerator and writing your code to the pandas API, rather than attempting...