pyvkfft - python interface to the CUDA and OpenCL backends of VkFFT (Vulkan Fast Fourier Transform library)

VkFFT is a GPU-accelerated Fast Fourier Transform library for Vulkan/CUDA/HIP/OpenCL.

pyvkfft offers a simple python interface to the CUDA and OpenCL backends of VkFFT, compatible with pyCUDA, CuPy and pyOpenCL.

Installation

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

Notes:

• the PyPI package includes vkfft.h and will automatically install pyopencl if opencl is available. However you should manually install either cupy or pycuda to use the cuda backend.
• if you want to specify the backend to be installed (which can be necessary e.g. if you have nvcc installed but cuda is not actually available), you can do that using e.g. VKFFT_BACKEND=opencl pip install pyvkfft. By default the opencl backend is always installed, and the cuda one if nvcc is found.

Requirements:

• pyopencl and the opencl libraries/development tools for the opencl backend
• pycuda or cupy and CUDA developments tools (nvcc) for the cuda backend
• numpy
• on Windows, this requires visual studio (c++ tools) and a cuda toolkit installation, with either CUDA_PATH or CUDA_HOME environment variable.
• Only when installing from source: vkfft.h installed in the usual include directories, or in the 'src' directory

This package can be installed from source using pip install ..

Note: python setup.py install is now disabled, to avoid messed up environments where both methods have been used.

Examples

The simplest way to use pyvkfft is to use the pyvkfft.fft interface, which will automatically create the VkFFTApp (the FFT plans) according to the type of GPU arrays (pycuda, pyopencl or cupy), and also cache these apps:

import pycuda.autoinit
import pycuda.gpuarray as cua
from pyvkfft.fft import fftn
import numpy as np

d0 = cua.to_gpu(np.random.uniform(0,1,(200,200)).astype(np.complex64))
# This will compute the fft to a new GPU array
d1 = fftn(d0)

# An in-place transform can also be done by specifying the destination
d0 = fftn(d0, d0)

# Or an out-of-place transform to an existing array (the destination array is always returned)
d1 = fftn(d0, d1)

See the scripts and notebooks in the examples directory. An example notebook is also available on google colab. Make sure to select a GPU for the runtime.

Features

• CUDA (using PyCUDA or CuPy) and OpenCL (using PyOpenCL) backends
• C2C, R2C/C2R for inplace and out-of-place transforms
• Direct Cosine Transform (DCT) of type 1, 2, 3 and 4 (EXPERIMENTAL, comparison with scipy DCT transforms are OK, but there are limitations on the array dimensions)
• single and double precision for all transforms (double precision requires device support)
• 1D, 2D and 3D transforms.
• array can be have more dimensions than the FFT (batch transforms).
• arbitrary array size, using Bluestein algorithm for prime numbers>13 (note that in this case the performance can be significantly lower, up to ~4x, depending on the transform size, see example performance plot below)
• transform along a given list of axes - this requires that after collapsing non-transformed axes, the last transformed axis is at most along the 3rd dimension, e.g. the following axes are allowed: (-2,-3), (-1,-3), (-1,-4), (-4,-5),... but not (-2, -4), (-1, -3, -4) or (-2, -3, -4). This is not allowed for R2C transforms.
• normalisation=0 (array L2 norm * array size on each transform) and 1 (the backward transform divides the L2 norm by the array size, so FFT*iFFT restores the original array)
• unit tests for all transforms: see test sub-directory. Note that these take a long time to finish due to the exhaustive number of sub-tests.
• Note that out-of-place C2R transform currently destroys the complex array for FFT dimensions >=2
• tested on macOS (10.13.6), Linux (Debian/Ubuntu, x86-64 and power9), and Windows 10 (Anaconda python 3.8 with Visual Studio 2019 and the CUDA toolkit 11.2)
• GPUs tested: mostly nVidia cards, but also some AMD cards and macOS with M1 GPUs.
• inplace transforms do not require an extra buffer or work area (as in cuFFT), unless the x size is larger than 8192, or if the y and z FFT size are larger than 2048. In that case a buffer of a size equal to the array is necessary. This makes larger FFT transforms possible based on memory requirements (even for R2C !) compared to cuFFT. For example you can compute the 3D FFT for a 1600**3 complex64 array with 32GB of memory.
• transforms can either be done by creating a VkFFTApp (a.k.a. the fft 'plan'), with the selected backend (pyvkfft.cuda for pycuda/cupy or pyvkfft.opencl for pyopencl) or by using the pyvkfft.fft interface with the fftn, ifftn, rfftn and irfftn functions which automatically detect the type of GPU array and cache the corresponding VkFFTApp (see the example notebook pyvkfft-fft.ipynb).
• the pyvkfft-test command-line script allows to test specifc transforms against expected accuracy values, for all types of transforms.
• pyvkfft results are now evaluated before any release with a comprehensive test suite, comparing transform results for all types of transforms: single and double precision, 1D, 2D and 3D, inplace and out-of-place, different norms, radix and Bluestein, etc... See pyvkfft/pyvkfft_test_suite.py to run the full suite, which takes 28 hours on a V100 GPU using up to 20 parallel process.

Performance

See the benchmark notebook, which allows to plot OpenCL and CUDA backend throughput, as well as compare with cuFFT (using scikit-cuda) and clFFT (using gpyfft).

Example result for batched 2D FFT with array dimensions of batch x N x N using a Titan V:

Notes regarding this plot:

• the computed throughput is theoretical, as if each transform axis for the couple (FFT, iFFT) required exactly one read and one write. This is obviously not true, and explains the drop after N=1024 for cuFFT and (in a smaller extent) vkFFT.
• the batch size is adapted for each N so the transform takes long enough, in practice the transformed array is at around 600MB. Transforms on small arrays with small batch sizes could produce smaller performances, or better ones when fully cached.
• a number of blue + (CuFFT) are actually performed as radix-N transforms with 7<N<127 (e.g. 11) -hence the performance similar to the blue dots- but the list of supported radix transforms is undocumented so they are not correctly labeled.

The general results are:

• vkFFT throughput is similar to cuFFT up to N=1024. For N>1024 vkFFT is much more efficient than cuFFT due to the smaller number of read and write per FFT axis (apart from isolated radix-2 3 sizes)
• the OpenCL and CUDA backends of vkFFT perform similarly, though there are ranges where CUDA performs better, due to different cache . [Note that if the card is also used for display, then difference can increase, e.g. for nvidia cards opencl performance is more affected when being used for display than the cuda backend]
• clFFT (via gpyfft) generally performs much worse than the other transforms, though this was tested using nVidia cards. (Note that the clFFT/gpyfft benchmark tries all FFT axis permutations to find the fastest combination)

Accuracy

See the accuracy notebook, which allows to compare the accuracy for different FFT libraries (pyvkfft with different options and backend, scikit-cuda (cuFFT), pyfftw), using pyfftw long-double precision as a reference.

Example results for 1D transforms (radix 2,3,5 and 7) using a Titan V:

Analysis:

• in single precision on the nVidia Titan V card, the VkFFT computed accuracy is about 3 times larger (worse) than pyfftw (also computed in single precision), e.g. 6e-7 vs 2e-7, which can be pretty negligible for most applications. However when using a lookup-table for trigonometric values instead of hardware functions (useLUT=1 in VkFFTApp), the accuracy is identical to pyfftw, and better than cuFFT.
• accuracy is the same for cuda and opencl, though this can depend on the card and drivers used (e.g. it's different on a GTX 1080)

You can easily test a transform using the pyvkfft-test command line script, e.g.:

pyvkfft-test --systematic --backend pycuda --nproc 8 --range 2 4500 --radix --ndim 2

Use pyvkfft-test --help to list available options.

You can use the pyvkfft/pyvkfft_test_suite.py script to run the comprehensive test suite which is used to evaluate pyvkfft before a new release.

TODO

• access to the other backends:
  • for vulkan and rocm this only makes sense combined to a pycuda/cupy/pyopencl equivalent.
• out-of-place C2R transform without modifying the C array ? This would require using a R array padded with two wolumns, as for the inplace transform
• half precision ?
• convolution ?
• zero-padding ?
• access to tweaking parameters in VkFFTConfiguration ?
• access to the code of the generated kernels ?