Leo Fang comments

Results 1175 comments of


                                            Leo Fang

WIP: Introducing `cuda.py`: pythonic access to CUDA core functionalities

Update: To unblock parallel developments I will merge this PR soon (~tomorrow) so that subsequent PRs can start targeting the main branch of this repo, after checking in the following...

WIP: Introducing `cuda.py`: pythonic access to CUDA core functionalities

Merging! cc @ksimpson-work

`expm` precision inconsistency for multiples of pi

FYI @danlkv `expm` can be further improved (https://github.com/cupy/cupy/issues/7769) if you are interested in contributing 😉

Support loading cuSPARSELt from PyPI wheel

If it'd make CuPy's life easier for nvmath-python to take care of this, please kindly file an issue 🙂

cupy.arange changes system's default encoding

This is not a CuPy issue, but an NVRTC one. Tracked in nvbugs 3833924.

`cupyx.scipy.sparse.linalg.splu` solve is slower than `scipy` equivalent

Q for @pentschev: Didn't @jrbourbeau already figure out the source of the data roundtrip in #8570? Or it's a different thing?

[RFC] Adding CUDA Graphs conditional node support with user-friendly APIs

Thanks for putting thoughts in this. A few drive-by comments: 1. I don't think we need to support `else` nodes, because without compiler help we have no way to validate...

Building CuPy against CUDA Python 12 fails

Thanks for reminder... I had completely forgotten about the other issue 😛

`#include "cutlass/gemm/device/gemm_universal_adapter.h"` is causing the named symbol to not be found

Based on the discussion in https://github.com/NVIDIA/cutlass/issues/1811, it seems to be irrelevant to CuPy. Removing the bug tag. @mrakgr please kindly link the nvbug ID here if you filed one, I...

Adopt cuda-python new module layout

xref: leofang/cuda-python#1