Graham Markall comments

Results 455 comments of


                                            Graham Markall

CUDA: Replace use of deprecated NVVM IR features, questionable constructs

gpuci run tests

CUDA: Replace use of deprecated NVVM IR features, questionable constructs

There's a couple of issues I need to address: - Now we need to know the NVVM version for the datalayout, the current implementation is always attempting to load NVVM....

CUDA: Replace use of deprecated NVVM IR features, questionable constructs

gpu ci run tests

CUDA: Replace use of deprecated NVVM IR features, questionable constructs

gpuci run tests

CUDA: Replace use of deprecated NVVM IR features, questionable constructs

gpuci run tests

CUDA: Replace use of deprecated NVVM IR features, questionable constructs

@guilhermeleobas Many thanks for the review - I've now merged main and responded to the comment above. Note I had to make a couple of edits to resolve the conflict,...

CUDA: Replace use of deprecated NVVM IR features, questionable constructs

gpuci run tests

CUDA: Replace use of deprecated NVVM IR features, questionable constructs

@stuartarchibald Many thanks for the review - I have responded to both comments.

meta issue: fp16 support

I can't edit the issue to tick the box, but it looks like: > llvmlite: > > * [ ] add `half` type is done in https://github.com/numba/llvmlite/pull/509, and > *...

meta issue: fp16 support

There is some support in CUDA from #7556 and #7460 - this work is ongoing though and there are more PRs that will be needed for full `float16` support in...