Cupy

Fixed the following errors when building [v7.8.0](https://github.com/cupy/cupy/releases/v7.8.0) source published on PyPI:

* `RuntimeError: Missing file: cupy/cuda/cub.cpp` (when CUB is configured via the environment variable or using CUDA 11.0)
* `RuntimeError: Missing file: cupy/cuda/cutensor.cpp` (when cuTENSOR is configured via the environment variable)

This release is only for packaging fix; there is no code difference since v7.8.0.

This is the release note of v7.8.0. See [here](https://github.com/cupy/cupy/milestone/79?closed=1) for the complete list of solved issues and merged PRs.

Highlights

* This release adds support for CUDA 11, NumPy 1.19, and SciPy 1.5.
* We expect this version to be the final release for v7.x series. Please start testing your workloads with the latest v8.x pre-release.

Notes on CUDA 11.0 support

* Update on 2020-09-23: `cupy-cuda110` package is now available on PyPI! ~`cupy-cuda110` wheel packages are currently available only for Windows. We are going to publish Linux wheels once we got [approval](https://github.com/pypa/pypi-support/issues/553) from the PyPI team. (update on 2020-08-21: Meanwhile, Linux wheels can be downloaded from the Assets section below (or `pip install cupy-cuda110 -f https://github.com/cupy/cupy/releases/tag/v7.8.0`). Those wheels will be removed once we published the package on PyPI.)~
* `cupy-cuda110` packages are built with cuDNN support but without bundled cuDNN shared libraries (see 3724 for the discussion). To use cuDNN features, you need to install cuDNN v8.0.x via the system package manager (e.g, `apt install libcudnn8` or `yum install libcudnn8`) or manually install it and set `LD_LIBRARY_PATH` (Linux) or `PATH` (Windows) environment variables.
* When building CuPy from source with CUDA 11.0, g++-6 or later is required. See the [installation guide](https://docs.cupy.dev/en/v7.8.0/install.html#build-fails-with-cuda-11-0-on-ubuntu-16-04-centos-6-or-7) for the detailed instructions.

New Features

- Support CUDA 11.0 (3720)
- Support cuSPARSE generic API (3721)

Enhancements

- Update CUDA 11.0 FP16 header to production release version (11.0.2) (3799)

Performance Improvements

- Improve cuDNN performance when using deterministic mode (3798)

Bug Fixes

- Fix broken message format (3698)
- Support order argument in cupy.ones, cupy.full and cupy.eye (3699, thanks grlee77!)
- Fix sparse matrix related test failures on CUDA11 (3761)
- Allow `MatDescriptor` to be pickle-able (3771)
- Skip FFT input checks for some CUDA >= 10.1 cases (3792)
- Add temporary fixes for cuDNN v8 (3794)
- Fix error message broken (3800)
- Fix cuSparse build failure on Windows (3809)

Documentation

- Fix format of docstrings in creation routines (3767)
- Update requirements (3803)
- Update install doc: `source devtoolset` needed in CentOS (3806)

Tests

- Fix wrong test file name (3754)
- Relax tolerance in `TestArrayElementwiseOp::test_doubly_broadcasted_pow` (3762)
- Skip tests failing due to exception type changes in NumPy 1.19 (3787)
- Avoid testing exception type match on NumPy 1.19 (3797)
- Skip `TestDiaMatrixScipyComparison` failing with `scipy>=1.5.0` (3805)

Others

- Bump version to v7.8.0 (3812)

This is the release note of v7.7.0. See [here](https://github.com/cupy/cupy/milestone/77?closed=1) for the complete list of solved issues and merged PRs.

Enhancements

- Support `cusparse<t>csrgeam2` and `cusparse<t>csrgemm2` (3666)

Bug Fixes

- Fix for `cupy.cuda.thrust` (3422)
- Fix sorting order of COO sparse matrix for `cuSPARSE` (3623)
- Fix array creation for `ndarray` list of arrays of different dtypes (3663)

Code Fixes

- Suppress compile warnings (3580)

Documentation

- Update url and email (3635)
- Add a warning for `sum_duplicates` (3636)
- Update Installation Guide (3660)

Tests

- Fix negative value test in `test_helper` (3622)
- Skip `csc` and `erf` tests for `scipy>1.2` (3628)

Others

- Update style checker version for Python 3.7 (3589)
- Add link to Twitter account (3634)
- Bump version to v7.7.0 (3688)
- Use builtins directly (3667, thanks larsoner!)

This is the release note of v7.6.0. See [here](https://github.com/cupy/cupy/milestone/75?closed=1) for the complete list of solved issues and merged PRs.

New Features

- Support all dtypes in every sorting function in `cupy.cuda.thrust` (3415, thanks leofang!)

Enhancements

- Get arch per device and support CUDA 9.2+ (3396, thanks leofang!)

Bug Fixes

- Fix `_count_non_nan` datatype for windows (3391)
- Properly reset current stream in case null stream is destroyed (3437)
- Fix `TypeError` in parameterize test catching `CUDADriverError` (3459)
- Assert that all the pointers are in the same device in `concatenate` (3472)

Code Fixes

- Use `find_packages` in `setup.py` (3436)

Documentation

- Fix sphinx version for travis (3417)
- Document `cupy.fromfile` (3447, thanks jakirkham!)
- Fix typos in `cupy.linalg.det docstring` (3458, thanks grlee77!)
- Fix docstring of `tofile()` (3471, thanks leofang!)

Installation

- Remove theano for doctest requirement (3463)


Tests

- Add `__init__.py` to allow importing test packages (3409)

Others

- Bump version to v7.6.0 (3480)

This is the release note of v7.5.0. See [here](https://github.com/cupy/cupy/milestone/73?closed=1) for the complete list of solved issues and merged PRs.

Enhancements

- Show versions of CUB and cuTENSOR on `cupy.show_config` (3353)
- Support sorting complex arrays (3336, thanks leofang!)

Bug Fixes

- Fix byte buffer handling to support PyPy (3227)
- Fix `put` when using scalars (3332)
- Remove some `xfail`s in sorting tests (3345)
- Fix `linalg.svd` for 0-sized matrices (3355)
- Assign a workpace to `ormqr` functions in `_solve` (3356)
- Fix windows build issue with CUDA 8.0 (3379)

Documentation

- Remove upper restrictions for numpy and scipy in doc build (3338)
- Add PFN to the README (3352)

Others

- Bump version to v7.5.0 (3377)

This is the release note of v7.4.0. See [here](https://github.com/cupy/cupy/milestone/71?closed=1) for the complete list of solved issues and merged PRs.

Enhancements
- Add CUDA 10.2 support (3218, thanks jakirkham!)

Bug Fixes

- Fix: n-dimensional FFTs must preserve array contiguity when copying a view (3249, thanks grlee77!)
- Use larger type to represent index range in `cupy.take` (3265)

Documentation

- Update installation guide for conda-forge (3198, thanks leofang!)
- Small typo/spelling fixes (3248, thanks svlandeg!)

Tests
- Fix overfow in `matmul` test (3245)

Others
- Bump version to v7.4.0 (3300)