Database for On-Line Function Approximation (DOLFA) is a storage/retrieval chemistry-acceleration scheme that is based on Pope's ISAT method [S.B. Pope Combust. Theory Modell. 1:41-63, 1997]. The interface and nomenclature of DOLFA are similar to those of ISAT. The performance of DOLFA/ISAT greatly depends on the specification of their controlling parameters. This work examines the impacts of key input parameters on accuracy and computational efficiency in multidimensional CFD for incylinder combustion in IC engines. Highly nonhomogeneous systems with direct in-cylinder liquid fuel injection are considered, and results with a Lagrangian particle probability density function (PDF) method are presented. Numerical experiments show that the choice of independent tabulation variables, error tolerances, and scale factors greatly affect accuracy and performance.