Graham Markall

AArch64 OrcJIT segfault

4

The conda-forge AArch64 build of llvmlite 0.41 is failing tests with a segfault. See https://github.com/conda-forge/llvmlite-feedstock/pull/74 I note that there seems to be an issue in one of the OrcJIT tests,...

libxml2 is required for building on Windows since LLVM 14

2

Identified in #922. * @apmasell Suggested that it may be required for emitted remarks: https://github.com/numba/llvmlite/pull/922#issuecomment-1476061341 * On Linux and macOS, we explicitly disable it: https://github.com/numba/llvmlite/blob/main/conda-recipes/llvmdev/build_llvm14.sh#L31 We should determine whether it's...

Azure on Windows does not use `setup_conda_environment.cmd`

1

Identified in #922. Instead of running `setup_conda_environment.cmd` to set up the environment, it runs commands inside `azure-windows.yml`. It seems that this was originally due to some issues with calling `setup_conda_environment.cmd`,...

For working on OrcJITv2 support. Current status: - Passes tests with LLVM 14.0.6 - Numba support needs work. - MCJIT example ported to LLJIT works. Current known to-do list: -...

Can AArch64 Numba tests be re-renabled?

2

### Describe the question. https://github.com/NVIDIA/DALI/pull/4862 disabled the Numba tests on AArch64. The AArch64 issues in Numba described in numba/numba#8567 and numba/numba#9001 were fixed by the merge of numba/llvmlite#1009 and in...

Tutorial requires sign-in?

3

I built grist-electron as documented in the README, and I want to start learning how to use it, but I've come up against some behaviour that surprised me and I...

Move a couple of CUDA-specific items into the CUDA target

6

Dummy arrays and device ufuncs are only used by the CUDA target, and I think are not generally useful even for someone implementing a new target: - The device ufunc...

Quick notes on this based on a verbal description from @guilhermeleobas: The `specialize()` function doesn't create the specialized dispatcher with any reference to caching, so caching is broken for specialized...

[RFC] Support for `__array__()`

14

Pushing this up early to request comments / feedback - I think this is one way to solve rapidsai/cudf#15694, with broad applicability to other use cases - this should allow...

This should address inconsistencies in the exceptions raised by `numpy_support.from_dtype()` and ensure that we catch its exceptions. I did consider whether we ever need to catch a `ValueError` - perhaps...