Tensor computation plays a paramount role in a broad range of domains, including machine learning, data analytics, and scientific computing. The wide adoption of tensor computation and its huge computation cost has led to high demand for flexible, portable, and high-performance library implementation on heterogeneous hardware accelerators such as GPUs and FPGAs. However, the current tensor library implementation mainly requires programmers to manually design low-level implementation and optimize from the algorithm, architecture, and compilation perspectives. Such a manual development process often takes months or even years, which falls far behind the rapid evolution of the application algorithms.
To address these challenges, we introduce FlexTensor, which is a schedule exploration and optimization framework for tensor computation on heterogeneous systems. FlexTensor can optimize tensor computation programs without human interference, allowing programmers to only work on high-level programming abstraction without considering the hardware platform details. FlexTensor systematically explores the optimization design spaces that are composed of many different schedules for different hardware. Then, FlexTensor combines different exploration techniques, including heuristic method and machine learning method to find the optimized schedule configuration. Finally, based on the results of exploration, customized schedules are automatically generated for different hardware. In the experiments, we test 12 different kinds of tensor computations with totally hundreds of test cases and FlexTensor achieves average 1.83x performance speedup on NVIDIA V100 GPU compared to cuDNN; 1.72x performance speedup on Intel Xeon CPU compared to MKL-DNN for 2D convolution; 1.5x performance speedup on Xilinx VU9P FPGA compared to OpenCL baselines; 2.21x speedup on NVIDIA V100 GPU compared to the state-of-the-art.