1 paper
Josse Van Delm, Anton Lydike, Joren Dumoulin +6
Contemporary compute platforms increasingly offload compute kernels from CPU to integrated hardware accelerators to reach maximum performance per Watt. Unfortunately, the time the…