smoothing kernels in specfem

[carltape]

Smoothing volumetric kernels can be computationally expensive. Many groups—salvus, U. Toronto, Princeton—are using PDE-based approaches (like diffusion equation) to smooth, rather than using convolution. UT and Princeton, are these tools in a state where they can be added into both packages, or are they still under development? It looks like the currently available approach for smoothing (and the one employed within seisflows/pyatoa) is convolution, as shown in smooth_sem.F90. Please chime in here if you are using these more efficient smoothing operations with specfem volumetric files. Thanks.

[rdno]

SPECFEM3D_GLOBE has the laplacian_smoothing_sem.F90 implemented by Stephen Beller (@1sbeller) and
Congyue Cui (@icui) . I am not sure if there was a plan to port it to Cartesian.