Text this: Efficient and Portable Support for Overdecomposition on Distributed Memory GPGPU Platforms