The three-dimensional KIVA code has been used to study the effects of injection pressure and split injections on diesel engine performance and soot and NOx emissions. The code has been updated with state-of-the-art submodels including: a wave breakup atomization model, drop drag with drop distortion, spray/wall interaction with sliding, rebounding, and breaking-up drops, multistep kinetics ignition and laminar-turbulent characteristic time combustion, wall heat transfer with unsteadiness and compressibility, Zeldovich NOx formation, and soot formation with Nagle Strickland-Constable oxidation. The computational results are compared with experimental data from a single-cylinder Caterpillar research engine equipped with a high-pressure, electronically-controlled fuel injection system, a full-dilution tunnel for soot measurements, and gaseous emissions instrumentation. The results show the use of the updated version of KIVA gives good agreement between measured and predicted engine cylinder pressures and heat release data for single injection cases. This level of agreement was found to be necessary so that the predicted and measured soot and NOx emissions trends would also agree well. Soot and NOx emissions were also found to be very sensitive to factors that influence the chamber gas temperatures such as crevice flow. The results show that the combustion is controlled by the details of the spray model. In particular, drop breakup and drop drag effects govern the penetration and mixing of vaporizing sprays under diesel engine conditions. The effects of spray-wall impingement were also found to be important. Consistent with measured data, computations showed that soot and NOx emissions increase at low injection pressures. Predictions made with ultra-high injection pressures suggest that there is a limit to the usefulness of high injection pressures for NOx emissions reduction. Good agreement with measured data was also found for split injections with relatively short dwells between injections. For split injections with larger dwell, the agreement deteriorated, indicating that refinements in spray, ignition and combustion models are still needed. Consistent with experiments, the computations show that split injections have a significant effect on the overall rate of pressure rise. The good agreement between experiments and model predictions indicates that computer models are now available for use by the engine industry to provide directions for engine design and to gain insight into in-cylinder events.