Parallel programming on graphics processors and an application

Users' expectation to solve complex problems by computers is an everincreasingtrend. Algorithms developed to solve complex problems often have highalgorithmic complexity and can not run on general purpose microprocessors in realtime.Hardware manufacturers inspired by this demand design innovativemicroprocessors with many processing elements to run algorithms in parallel toreduce run time.A parallel programming technique, Compute Unified Device Architecture(CUDA) by Nvidia is introduced in this thesis. CUDA is then used to program analgorithm based on the fuzzy logic rules to identify skinned hazelnuts on a movingconveyor belt in real time to prove the effectiveness of the technique. Run times forboth parallel and serial implementation of the same algorithm is reported.