A persistent parallel region keeps the thread team active across all time steps in one kernel call. The repository places omp parallel outside the time loop:

#pragma omp parallel \
    default(none) \
    firstprivate(Hy, Ez, Ch1, Ch2, Ce1, Ce2, width, nsteps) \
    shared(sim)
{
    for (size_t n = 0; n < nsteps; n++) {
 
        // Update Hy
        #pragma omp for schedule(static)
        for (size_t k = 0; k < width - 1; k++) {
            Hy[k] = Ch1[k] * Hy[k] 
                + Ch2[k] * (Ez[k + 1] - Ez[k]);
        }
 
        // Update Ez
        #pragma omp for schedule(static)
        for (size_t k = 1; k < width; k++) {
            Ez[k] = Ce1[k] * Ez[k]
                + Ce2[k] * (Hy[k] - Hy[k - 1]);
        }
 
        // Inject Ez source
        #pragma omp single
        {
            Fdtd_EzSourcesApply(sim->fsim_sources, sim);
            Fdtd_EzStateNextStep(sim->fsim_state);
        }
    }
}

Every thread executes the outer loop, while the two omp for directives distribute spatial work. Their barriers separate field phases; single applies sources and advances the counter once.

The team joins only after the time loop finishes. This avoids entering a new parallel region for every field phase, but does not remove the per-step synchronization costs. A later Fdtd_EzSimRun() call enters another region; the runtime may reuse its worker threads.

Source: fdtd_ez_kernels.c.