The water-driven AquaCrop model is used extensively for simulating crop growth and water use. A three-year field experiment (2015–2017) of foxtail millet (Setaria italica) that was grown using plastic film mulching (PM) and no mulching (NM) was conducted in a rain-fed region of China to simulate canopy cover (CC), biomass, soil water content (SWC), yield, evapotranspiration (ETc), and water use efficiency (WUE). The year 2015 was much drier and warmer than the two other years. The model was calibrated using field data from 2016 and validated using the data from 2015 and 2017. Simulations of CC, biomass, and yield achieved favorable performance for both PM and NM in all years, as indicated by the high determination coefficient (R²), model efficiency (EF), small root mean square error (RMSE), normalized root mean square error (NRMSE), and deviations < 10%. Simulations of SWC, ETc, and WUE gave acceptable results for both PM and NM in the normal year (2017). However, low R² and EF, and large NRMSE, RMSE, and deviations were observed in the predictions of PM and NM for SWC, ETc, and WUE in the dry year (2015) with a severe drought stress, indicating that the model performed unsatisfactorily under severe drought stress condition that was caused by the adverse weather. In addition, the simulation performance of NM was more favorable than that of PM for most crop growth and water use indexes under no drought stress condition.