Leo Fang
For NVCC it's probably blocked by https://github.com/NVIDIA/cuda-python/issues/762 (cc @rwgk for vis), though I don't think it is critical. We cannot pip install a host compiler anyway, so what we currently...
Hi @0x1abcd We have discussed with the NVIDIA RAPIDS team. We've noticed this account and several others have jointly started opening a large quantity of issues in the last 24-48...
With my human read: This is definitely AI generated. We've disabled the CUB path in histogram API in the entirely v13 series, and the revert in #8950 was not backported...
/test mini
/test mini
FYI, the low-level bindings to the `cudaFuncSetAttribute` (runtime) or `cuFuncSetAttribute` (driver) API are exposed in `cuda.bindings` (part of `cuda-python`). As a stop-gap solution, I think it should be possible to...
@Kairoven We have discussed with the NVIDIA RAPIDS team. We've noticed this account and several others have jointly started opening a large quantity of issues in the last 24-48 hours....
Discussed briefly with @shwina. This is the kind of tasks that I'd love to easily support using `cuda.cccl.parallel` ([example](https://github.com/NVIDIA/cccl/blob/8759390985a6616cf9015b43801ed88424b6135d/python/cuda_cccl/tests/parallel/examples/scan/logcdf_example.py#L37), [RFC](https://github.com/cupy/cupy/issues/9238)), instead of homegrown solutions! 😄
@kmaehashi WDYT?
One obvious alternative solution that I did not mention is to rely on WheelNext's variant plugin to allow `pip install cupy` for selecting the right variant package to install. I...