arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2104.05755cs.AI
收藏

张量处理基元:深度学习和HPC工作负载中的效率和可移植性的编程抽象

Tensor Processing Primitives: A Programming Abstraction for Efficiency and Portability in Deep Learning & HPC Workloads

Intel Corporation(英特尔公司);Friedrich-Schiller–Universität Jena(耶拿弗里德里希·席勒大学)
Evangelos Georganas, Dhiraj Kalamkar, Sasikanth Avancha, Menachem Adelman, Deepti Aggarwal, Cristina Anderson, Alexander Breuer, Jeremy Bruestle, Narendra Chaud… 展开作者

Evangelos Georganas, Dhiraj Kalamkar, Sasikanth Avancha, Menachem Adelman, Deepti Aggarwal, Cristina Anderson, Alexander Breuer, Jeremy Bruestle, Narendra Chaudhary, Abhisek Kundu, Denise Kutnick, Frank Laub, Vasimuddin Md, Sanchit Misra, Ramanarayan Mohanty, Hans Pabst, Brian Retford, Barukh Ziv, Alexander Heinecke

摘要:

在过去的十年中,新的深度学习(DL)算法、工作负载和硬件已经被开发出来,以解决各种各样的问题。尽管工作负载和硬件生态系统取得了进步,但DL系统的编程方法仍然停滞不前。DL工作负载利用DL库中高度优化、但特定于平台且不灵活的内核,或者对于新型运算符,参考实现通过DL框架原语构建,性能不佳。这项工作引入了张量处理基元(TPP),这是一种编程抽象,致力于高效、可移植地实现DL工作负载并提高生产力。TPP定义了一组紧凑而通用的2D张量算子(或虚拟张量ISA),随后可以用作构建块来构建高维张量上的复杂算子。TPP规范是平台无关的,因此通过TPP表达的代码是可移植的,而TPP实现是高度优化的和特定于平台的。我们使用独立内核和完全通过TPP表达的端到端DL和HPC工作负载来证明我们方法的有效性和可行性,这些工作负载在多个平台上的性能优于最先进的实现。

英文摘要:

During the past decade, novel Deep Learning (DL) algorithms, workloads and hardware have been developed to tackle a wide range of problems. Despite the advances in workload and hardware ecosystems, the programming methodology of DL systems is stagnant. DL workloads leverage either highly-optimized, yet platform-specific and inflexible kernels from DL libraries, or in the case of novel operators, reference implementations are built via DL framework primitives with underwhelming performance. This work introduces the Tensor Processing Primitives (TPP), a programming abstraction striving for efficient, portable implementation of DL workloads with high-productivity. TPPs define a compact, yet versatile set of 2D-tensor operators (or a virtual Tensor ISA), which subsequently can be utilized as building-blocks to construct complex operators on high-dimensional tensors. The TPP specification is platform-agnostic, thus code expressed via TPPs is portable, whereas the TPP implementation is highly-optimized and platform-specific. We demonstrate the efficacy and viability of our approach using standalone kernels and end-to-end DL & HPC workloads expressed entirely via TPPs that outperform state-of-the-art implementations on multiple platforms.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑