Work

From CUDA to OpenCL: Towards a performance-portable solution for multi-platform GPU programming