GPU optimized computation of stencil based algorithms

The paper describes an optimized GPU based approach for stencil based algorithms. The simulations have been performed for a two dimensional steady state heat conduction problem, which has been solved through the red black point successive over relaxation method. Two kernels have been developed and their performance has been greatly improved through coalesced memory accesses and special shared memory approaches. The approach described in the paper does not only represent a step forward for the steady state heat conduction problem but also for any other algorithm which performs the numerical solution of partial differential equations or which is stencil based. The paper not only describes the various code versions but also the process which has lead to these improvements. Also the optimized GPU code version has been compared with the corresponding CPU version. The testing results show that the GPU algorithm always leads to an improvement. The value of the improvement though greatly depends on the number of grid points on which the computations are performed.