_evaluate_uniform_cubic_floored_parallel

ell1fit.templates._evaluate_uniform_cubic_floored_parallel(phases, coefficients, x0, dx, n_intervals, floor)

Multi-threaded twin of _evaluate_uniform_cubic_floored().

Each iteration writes one independent output element, so there is no reduction whose order could change: the result is bitwise identical to the serial kernel, verified in the tests. Only the wall-clock time differs.