# Throughput (1D FDTD)

**Throughput** measures how many grid cells the simulation advances per second. FDTD OpenMP reports it in millions of cells per second, or Mcells/s. For a grid of width $K$ advanced through $N$ steps in $t$ seconds:

$$
Q=\frac{KN}{10^6t}
$$

One cell-step represents the work of advancing a cell through a complete simulation step, including the magnetic and electric update phases. The plotting script uses the grid width rather than separately counting the slightly different loop lengths at the boundaries.

For example, advancing 100,000 cells through 500 steps in 0.1 seconds gives 500 Mcells/s. This makes runs with different amounts of work easier to compare than elapsed time alone.

Throughput can change with grid size even when the update equations stay the same. Small grids expose synchronization overhead, while larger working sets can exceed cache capacity and increase memory traffic. The benchmark plots throughput against both grid width and thread count. Use it alongside [[fdtd-openmp-speedup|speedup]] to distinguish high absolute performance from improvement over a particular serial baseline.