Optimizing tensor contractions for embedded devices with racetrack memory scratch-pads
Research output: Contribution to book/conference proceedings/anthology/report › Conference contribution › Contributed › peer-review
Contributors
Abstract
Tensor contraction is a fundamental operation in many algorithms with a plethora of applications ranging from quantum chemistry over fluid dynamics and image processing to machine learning. The performance of tensor computations critically depends on the efficient utilization of on-chip memories. In the context of low-power embedded devices, efficient management of the memory space becomes even more crucial, in order to meet energy constraints. This work aims at investigating strategies for performance- and energy-efficient tensor contractions on embedded systems, using racetrack memory (RTM)-based scratch-pad memory (SPM). Compiler optimizations such as the loop access order and data layout transformations paired with architectural optimizations such as prefetching and preshifting are employed to reduce the shifting overhead in RTMs. Experimental results demonstrate that the proposed optimizations improve the SPM performance and energy consumption by 24% and 74% respectively compared to an iso-capacity SRAM.
Details
Original language | English |
---|---|
Title of host publication | LCTES 2019: Proceedings of the 20th ACM SIGPLAN/SIGBED International Conference on Languages, Compilers, and Tools for Embedded Systems |
Publisher | Association for Computing Machinery (ACM), New York |
Pages | 5-18 |
Number of pages | 14 |
ISBN (electronic) | 978-1-4503-6693-9 |
Publication status | Published - 23 Jun 2019 |
Peer-reviewed | Yes |
Publication series
Series | CPSWeek: Cyber-physical Systems (SIGPLAN/SIGBED) |
---|
Conference
Title | 20th ACM SIGPLAN/SIGBED International Conference on Languages, Compilers, and Tools for Embedded Systems, LCTES 2019, co-located with PLDI 2019 |
---|---|
Duration | 23 June 2019 |
City | Phoenix |
Country | United States of America |
External IDs
ORCID | /0000-0002-5007-445X/work/141545623 |
---|
Keywords
Research priority areas of TU Dresden
Sustainable Development Goals
ASJC Scopus subject areas
Keywords
- Compiler optimization, Data transformation, Embedded systems, Matrix multiplication, Prefetching, Preshifting, Racetrack memory, Tensor contraction, Tensors