Hi @DTolm , now that the release is out, I made some tests with on-the-fly convolution following Osamu's email exchange which picked my curiosity.

There is now a branch with convolution support

What I have seen (but I made so far a very limited number of tests):

• 2D and batched 2D convolution works for R2C and C2C (compared to convolution using numpy), both for out- and in-place transforms
• for batched 2D, I use config->coordinateFeatures = n_batch instead of numberBatches, I assume that's the proper way
• 1D transforms kernel compilations fail (see the notebook below)
• 3D transforms give incorrect results

The tests are all visible on the pyvkfft-convolve notebook.