solprop.backends.numba_cpu#

The CPU backend: fused numba kernels over FFTW plans.

The execution chosen by a bake-off of candidate backends (docs/dev/backends.md section 5): the seven numbered passes of architecture.md section 8 collapsed into a handful of compiled kernels, the transforms planned once by FFTW on buffers that never move, and all five thread pools owned rather than inherited.

Modules

backend

The CPU backend: the seam, the stage driver and the statistics.

batch

Execute CPU member stages with one compiled batch update and segmented norms.

coupled_flow

The coupled flow of a coherent wall or a split block, on the CPU backend.

dispersive

SEP 6's frequency-dependent electronic response on the fused node path.

dispersive_general

SEP 6's frequency-dependent electronic response on the general modal route.

dispersive_spatial

SEP 6's frequency-dependent electronic response on the Cartesian and planar routes.

envelope_plasma

Own compiled envelope-current workspaces and their forward transforms.

fft

FFTW plans, wisdom, and the five thread pools the backend owns.

field_plasma

Own the held field-current streams, transforms and frequency assembly.

general_modal

The general modal guide's nonlinear source, compiled, on the CPU backend.

kernels

The fused stage kernels: seven numbered passes in a handful of traversals.

low_storage

The low-storage RK4IP march's register kernels and resident engine on numba_cpu.

multistep

The fused Adams step's passes, compiled (SEP 14 phase 2, SEP 16 phase 3).

openmp

The pyFFTW import, made so that its OpenMP threads sleep when they are idle.

pass_timers

Opt-in wall-clock timers for the passes of a free-space evaluation (SEP 28 phase 0).

planar

Execute the hybrid planar transform's compiled passes.

spatial

Execute separable spatial coefficients and bounded Cartesian column transfers.

tables

Evaluate the reference z-table representation into persistent host stage arrays.