https://hgpu.org/?p=4418
Performance evaluation of the multi-device OpenCL FDTD solver