https://hgpu.org/?p=6295
Scope for performance enhancement of CMU Sphinx by parallelising with OpenCL