In the area of AutoTuning some machine learning methods are used. The goal is to optimize throughput or latency for compute intensive workloads. The machine le

In the area of AutoTuning some machine learning methods are used. The goal is to optimize throughput or latency for compute intensive workloads. The machine learning model is trained with performance measurements on the hardware and then use it as proxy for real execution to try out different configurations. Notable works: AutoTVM - https://arxiv.org/abs/1805.08166 https://arxiv.org/abs/1805.08166 CHAMELEON - https://deepai.org/publication/chameleon-adaptive-code-optimization-for-expedited-deep-neural-network-compilation https://deepai.org/publication/chameleon-adaptive-code-optim... GPU-Warp Scheduling - https://deepai.org/publication/rlws-a-reinforcement-learning-based-gpu-warp-scheduler https://deepai.org/publication/rlws-a-reinforcement-learning...