Source: cupy
Section: contrib/science
Priority: optional
Maintainer: Adam Cecile <acecile@le-vert.net>
Build-Depends: debhelper (>= 9), dh-python, python3-all-dev, python3-setuptools,
# Documentation says 7.5 but
# *** WARNING: CUDA version is too old: 7050
# *** WARNING: CUDA v7.0 or newer is required
# And build stops then
 nvidia-cuda-toolkit (>= 8~),
 libcudnn8-dev | libcudnn7-dev | libcudnn6-dev,
 libthrust-dev (>= 1.8.1~),
 libnccl-dev,
 python3-numpy (>= 1.9.0~),
 python3-six (>= 1.9.0~),
 python3-fastrlock (>= 0.3~),
 cython3 (>= 0.28.0~),
 python3-pytest <!nocheck>,
 python3-mock <!nocheck>,
 lsb-release,
Standards-Version: 4.1.3
Homepage: https://github.com/cupy/cupy/archive/v5.4.0.tar.gz
X-Python-Version: >= 2.7
X-Python3-Version: >= 3.4

Package: python3-cupy
Architecture: any
Depends: ${python3:Depends}, ${misc:Depends}, ${shlibs:Depends}
Suggests: python-cupy-doc
Description: NumPy-like API accelerated with CUDA (Python 3)
 CuPy is an implementation of NumPy-compatible multi-dimensional array on
 NVIDIA CUDA.
 .
 CuPy consists of the core multi-dimensional array class, cupy.ndarray, and
 many functions on it. It supports a subset of numpy.ndarray interface.
 .
 This package installs the library for Python 3.
