Enhancing manageability of execution and data for GPGPU computing