Leo Fang
Spoke to Vyas while I was on site. We did not have a clear vision whether this is even fixable or just something we have to document/keep track, and be...
We should add a CI job that tests running newer `cuda.bindings` + projects built against older `cuda.bindings` (from the same major release) to catch this in the future, either on...
Since nothing is actionable, closing.
This could be challenging if a PR closes more than one issue... Do we copy the labels from all issues?
An internal discussion pointed out that we should also document how to pass texture/surface objects: https://github.com/NVIDIA/cuda-python/blob/27317a00c8406a542716072f1f0956d66672147b/cuda_bindings/examples/0_Introduction/simpleCubemapTexture_test.py#L170
To add: - benchmarks (#407)
Unfortunately I have been unable to follow the recent, rapid advancement in DLPack, and do not fully understand what we are trying to accomplish with the added complexity. I worry...
cc @kmaehashi for vis
> FWIW [here's my proposal for cudf](https://github.com/rapidsai/cudf/issues/17626). It is not that simple to just do a dynamic detection for which default stream is in use, unfortunately. `cuda-bindings` was implemented in...
This requires a bit of grep'ing to see. Try to grep in `$CUDA_PATH/include` for suffixes `_ptds` and `_ptsz`, and the macros `__CUDA_API_PTDS` and `__CUDA_API_PTSZ`. You will see that for `CUDA_API_PER_THREAD_DEFAULT_STREAM`...