Talks and presentations

AMOS @ISCA 2022 (online)

June 19, 2022

Talk, AMOS 2022, New York City, USA

Abstract

Hardware specialization is a promising trend to sustain performance growth. Spatial hardware accelerators that employ specialized and hierarchical computation and memory resources have recently shown high performance gains for tensor applications such as deep learning, scientific computing, and data mining. To harness the power of these hardware accelerators, programmers have to use specialized instructions with certain hardware constraints. However, these hardware accelerators and instructions are quite new and there is a lack of understanding of the hardware abstraction, performance optimization space, and automatic methodologies to explore the space. Existing compilers use hand-tuned computation implementations and optimization templates, resulting in sub-optimal performance and heavy development costs.

In this paper, we propose AMOS, which is an automatic compilation framework for spatial hardware accelerators. Central to this framework is the hardware abstraction that not only clearly specifies the behavior of spatial hardware instructions, but also formally defines the mapping problem from software to hardware. Based on the abstraction, we develop algorithms and performance models to explore various mappings automatically. Finally, we build a compilation framework that uses the hardware abstraction as compiler intermediate representation (IR), explores both compute mappings and memory mappings, and generates high-performance code for different hardware backends. Our experiments show that AMOS achieves more than $2.50\times$ speedup to hand-optimized libraries on Tensor Core, $1.37\times$ speedup to TVM on vector units of Intel CPU for AVX-512, and up to $25.04\times$ speedup to AutoTVM on dot units of Mali GPU. The source code of AMOS is publicly available.

AHS: An Agile Framework for Hardware Specialization and Software Mapping

October 18, 2021

Conference proceedings talk, MICRO 2021, Global online

Overview

As Moore’s law is approaching to the end, designing specialized hardware along with the software that map the applications onto the specialized hardware is a promising solution. The hardware design determines the peak performance, while the software is also important as it determines the actual performance. Hardware/software (HW/SW) co-design can optimize the hardware acceleration and software mapping in concert and improve overall performance. However, the current flow designs hardware and software in isolation. More importantly, both hardware and software are difficult to design and optimize due to the low level programming and huge design space.

FleXTensor @ASPLOS 2020 (online)

March 16, 2020

Talk, ASPLOS 2020 (online), Lausanne, Switzerland

Abstract

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. In this paper, 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.