Domain-Specific Hardware Accelerators: DSAs gain efficiency from specialization and performance from parallelism.

The article discusses hardware computing engines known as domain-specific accelerators (DSAs). It discusses Moore's Law, the design of the DSA, and various sources of acceleration for DSAs. It also discusses algorithms, total cost of ownership (TCO), and memory in the accelerators.

Bibliographic Details
Published in:Communications of the ACM Vol. 63; no. 7; pp. 48 - 58
Main Authors: DALLY, WILLIAM J., TURAKHIA, YATISH, SONG HAN
Format: Article
Published: Association for Computing Machinery Jul2020
Subjects:
Online Access:View this record in EBSCOhost
Description
Summary:The article discusses hardware computing engines known as domain-specific accelerators (DSAs). It discusses Moore's Law, the design of the DSA, and various sources of acceleration for DSAs. It also discusses algorithms, total cost of ownership (TCO), and memory in the accelerators.