Additional speed benchmarks¶

All methods are benchmarked using their default parameters, with the following exceptions:

basex(var) and daun(var) mean “variable regularization”, that is changing the regularization parameter for each transformed image.
direct_C and direct_Python correspond to the “direct” method using its C (Cython) and Python backends respectively.
linbasex and rbasex show whole-image (n × n) transforms, while all other methods show half-image (n rows, (n + 1)/2 columns) transforms.
rbasex(None) means no output-image creation (only the transformed radial distributions).

Intel i7-9700 (Linux)¶

CPU:	Intel Core i7-9700 (8 cores, 8 threads; 3.0 GHz base, 4.7 GHz max)
RAM:	32 GB DDR4-2666
OS:	Ubuntu 20.04 LTS
Libraries:	NumPy 1.18.1 SciPy 1.4.1 MKL 2020 daz

CPU:	Intel Core i7-6700 (4 cores, 8 threads; 3.4 GHz base, 4.0 GHz max)
RAM:	32 GB DDR4-2133
OS:	Ubuntu 19.10
Libraries:	NumPy 1.18.1 SciPy 1.4.1 MKL 2019 Update 5 daz

CPU:	AMD Ryzen 3 2200G (4 cores, 4 threads; 3.5 GHz base, 3.7 GHz max)
RAM:	16 GB DDR4-2400
OS:	Debian GNU/Linux 9
Libraries:	NumPy 1.12.1 SciPy 0.18.1 OpenBLAS 0.2.19

CPU:	AMD Ryzen 3 2200G (4 cores, 4 threads; 3.5 GHz base, 3.7 GHz max)
RAM:	16 GB DDR4-2400
OS:	Microsoft Windows 10
Libraries:	NumPy 1.18.4 SciPy 1.4.1 OpenBLAS 0.3.7