Skip to content

Commit b4f91c0

Browse files
committed
Update doc, prepare for release.
1 parent 7295dae commit b4f91c0

File tree

2 files changed

+16
-8
lines changed

2 files changed

+16
-8
lines changed

CHANGELOG.rst

Lines changed: 7 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -2,11 +2,13 @@ Version 2022.1.0 (2022-01-XX)
22
-----------------------------
33
* Added accuracy unit tests, which can be used systematically
44
using the 'pyvkfft-test' installed script
5-
* Update to VkFFT 1.2.X, with support for DCT types 1, 2, 3 and 4,
6-
also fixing DCT issues (see https://github.com/DTolm/VkFFT/issues/48).
7-
DCT calculations now match scipy.
8-
This also fixes a bug for some Bluestein transforms,
9-
e.g. for 2D iFFT with a size of 808x808 or 1010x1010
5+
* An extensive testing is now made before official releases,
6+
evaluating all type of transforms (c2c, r2c, dct, 1, 2 and 3D,
7+
in and out-of-place, norm 0 and 1), different GPUs, both OpenCL
8+
and CUDA, etc... Comparison is made against pyfftw, scipy or numpy.
9+
* Update to VkFFT 1.2.20, with support for DCT types 1, 2, 3 and 4,
10+
also fixing a number of issues (see closed issues at
11+
https://github.com/DTolm/VkFFT/issues)
1012
* Raise a RuntimeError if the VkFFTApp initialisation or the
1113
GPU kernel launch fails, with the corresponding VkFFT error.
1214
* [BUG] Correct inverse FFT calculation using pyvkfft.fft.ifftn()

README.rst

Lines changed: 9 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -12,9 +12,15 @@ Installation
1212

1313
Install using ``pip install pyvkfft`` (works on macOS, Linux and Windows).
1414

15-
Note that the PyPI archive includes ``vkfft.h`` and will automatically install ``pyopencl``
16-
if opencl is available. However you should manually install either ``cupy`` or ``pycuda``
17-
to use the cuda backend.
15+
Notes:
16+
17+
- the PyPI package includes ``vkfft.h`` and will automatically install ``pyopencl``
18+
if opencl is available. However you should manually install either ``cupy`` or ``pycuda``
19+
to use the cuda backend.
20+
- if you want to specify the backend to be installed (which can be necessary e.g.
21+
if you have ``nvcc`` installed but cuda is not actually available), you can do
22+
that using e.g. ``VKFFT_BACKEND=opencl pip install pyvkfft``. By default the opencl
23+
backend is always installed, and the cuda one if nvcc is found.
1824

1925
Requirements:
2026

0 commit comments

Comments
 (0)