Alex Barnett

Dear Beliavsky (sorry, can't find your first name), Here are two useful tools for fortran to add to your list: * https://github.com/zgimbutas/mwrap which we use to wrap fort from matlab/octave...

use the correct eigenvector in evp.gsvdtension to output modes. Use 'ms' type search for gsvd. A lot of cleaning up to do here. This includes adding Dirichlet tension methods to...

periodic case: spectral diff non-periodic segment: high-order interpolation. accuracy will suffer a bit, but saves ppl a bunch of annoyance w/ Z, Zp, Zpp for Kress.

so Gary's codes can use it.

mathematical notation bugs in docs

1

Dear Tobias & crew, I'm very glad you have mathematical descriptions of the transforms - this is essential for a professional math library. But there are a bunch of typos...

The notation \nabla ( H_0 (k \| x - x_j \|) ) for a dipole source for Helmholtz is unclear, and implies \nabla_x which would be incorrect (the target-grad when...

Just lodging the request that we build some MATLAB/octave wrappers some day soon. Calling this from MATLAB would be really cool, as a drop-in replacement for FINUFFT.

Hmm, having to PR to master is annoying for docs :)

I might suggest that emfmm3d return U.Etarg, U.curlEtarg, U.divEtarg for consistency with the "targ" suffix for all other kernels. That way when source-eval is added, it won't break user code!...

faster zeroing of internal spreadinterp array.

1

Zero the output array - best way?  (1.5 s for 200M array on AMD laptop, too slow, 10% of a 1e8 1e8 1d1 transform @1e-6) switch to Calloc in makeplan,...