Hardware thread reordering to boost OpenCL throughput on FPGAs