Compute Unified Device Architecture (CUDA) is mature tool for developing high performance scientific applications on the GPU. OpenCL is open standard for cross-platform, parallel programming of heterogeneous processing systems which proposes a model which maps well to contemporary GPU architectures and a GPU programming model that bears close similarities to CUDA. OpenCL has attracted vendor support, with implementations available from NVIDIA, AMD , Apple and IBM. A tool for porting CUDA program to OpenCL called SWAN is already developed.
This task comprises porting of OpenCL program to CUDA and to test performance of application on OpenCL and on CUDA platform after porting. The design has to be optimized in terms of throughput and latency requirements of the system.
Prerequisites: C programming.
Prerequisites: C programming.
Contact: Muhammad Kashif Hanif
No comments:
Post a Comment