We propose development of new algorithms specifically designed to exploit the highly parallel structure of graphics processing units (GPUs) for performing the following most expensive, but parallelizable computations in combustion CFD: (1) Chemical kinetics source term (including Jacobian matrix) evaluation; (2) Transport property evaluations; and (3) Matrix factorizations and inversions. The algorithms developed in this work will be implemented as software modules that can be easily interfaced with arbitrary CFD solvers for rapid computations using GPUs. A user guide will be delivered with directions for coupling the provided algorithms with users' CFD programs. Phase I work will demonstrate the computational acceleration achieved using the preliminary algorithms; and Phase II work will optimize the algorithms for improved performance and implement the algorithms as well-documented, distributable software modules as described above. This work will significantly increase the predictive capability of combustion CFD simulations by enabling efficient application of much larger chemistry models (which is essential, but currently prohibitively expensive) for accurately modeling the combustion of practical fuels.