Post-process div U* in spectral space for the non-periodic x case.
After the x-y transpose applied by fft_forward_100_omp, dim1 holds the periodic y r2c modes and dim2 holds the non-periodic x modes. process_spectral_100 handles that arrangement directly.
The CUDA backend stages the equivalent work across three kernels so that a mode can fetch its pairing partner from another GPU. On a single rank every partner is local, so the staging collapses back into this one fused pass.
| Type | Intent | Optional | Attributes | Name | ||
|---|---|---|---|---|---|---|
| class(omp_poisson_fft_t) | :: | self |