%FILENAME%
python-cupy-13.6.0-11-x86_64.pkg.tar.zst

%NAME%
python-cupy

%BASE%
python-cupy

%VERSION%
13.6.0-11

%DESC%
NumPy/SciPy-compatible array library for GPU-accelerated computing with Python (CUDA version)

%CSIZE%
30287823

%ISIZE%
70511434

%MD5SUM%
2e00f67f93fcd8cbdefa58e7d299107c

%SHA256SUM%
84cb5fc21387bfaae77e0a42a1b3afcf032878f9b1ecebdb417eeb4a0cd4b222

%PGPSIG%
iHUEABYKAB0WIQQEzwzW9u6TrhiW9YQH0GNRylsxvgUCaXttKgAKCRAH0GNRylsxvoJFAQDskelG4/xlWnoOibqbf8DVnIK8bVxTv3A05sw97sk2zwD/ScITakAvpR/Ef89X0PmxRiGv7xTTwz0kZA0xGJCJ8Aw=

%URL%
https://github.com/cupy/cupy

%LICENSE%
MIT

%ARCH%
x86_64

%BUILDDATE%
1769692680

%PACKAGER%
Torsten Keßler <tpkessler@archlinux.org>

%DEPENDS%
cuda
gcc-libs
glibc
python
python-fastrlock
python-numpy

%OPTDEPENDS%
python-scipy: for copying sparse matrices from GPU to CPU
python-optuna: for using using Automatic Kernel Parameters Optimizations (cupyx.optimizing)
cutensor: for acceleratedd tensor operations
nccl: for collective multi-GPU computations

%MAKEDEPENDS%
git
cython
python-build
python-installer
python-wheel
python-setuptools
python-fastrlock
python-numpy
cuda
cutensor
nccl
hip-runtime-amd
hipblas
hipcub
hipfft
hiprand
hipsparse
rccl
rocfft
rocprim
rocrand
rocsolver
rocsparse
rocthrust
roctracer
rocm-toolchain

