18596835. Streaming Transfers and Ordering Model simplified abstract (Google LLC)
Contents
Streaming Transfers and Ordering Model
Organization Name
Inventor(s)
Rahul Nagarajan of San Jose CA (US)
Arpith Chacko Jacob of Los Altos CA (US)
Suvinay Subramanian of Sunnyvale CA (US)
Hema Hariharan of Cupertino CA (US)
Streaming Transfers and Ordering Model - A simplified explanation of the abstract
This abstract first appeared for US patent application 18596835 titled 'Streaming Transfers and Ordering Model
The abstract describes a hardware/software interface for asynchronous data movement between off-core memory and core-local memory, known as "stream transfers," along with a stream ordering model. Stream transfers enable more efficient expression of common data-movement patterns, particularly in sparse workloads. Direct stream instructions within a stream are processed in order, while indirect stream instructions process offset elements in an offset list in order. A sync flag is updated to indicate monotonic incremental progress for the stream.
- Hardware/software interface for asynchronous data movement
- Stream transfers between off-core memory and core-local memory
- Stream ordering model for efficient data-movement patterns
- In-order processing of direct stream instructions
- Processing of offset elements in an offset list in order for indirect stream instructions
- Update of sync flag to indicate incremental progress for the stream
Potential Applications: - High-performance computing - Data-intensive applications - Sparse workloads
Problems Solved: - Efficient data movement between different types of memory - Improved performance in sparse workloads
Benefits: - Faster data transfers - More efficient use of memory resources - Improved overall system performance
Commercial Applications: Title: "Enhanced Data Movement Interface for High-Performance Computing" This technology could be used in supercomputers, data centers, and other high-performance computing environments to optimize data movement and improve overall system performance.
Questions about the technology: 1. How does the hardware/software interface improve data movement efficiency? 2. What are the potential implications of using stream transfers in sparse workloads?
Original Abstract Submitted
Generally disclosed herein is a hardware/software interface for asynchronous data movement between an off-core memory and a core-local memory, referred to as “stream transfers”, and a stream ordering model. The stream transfers allow software to more efficiently express common data-movement patterns, specifically ones seen in sparse workloads. Direct stream instructions that belong to a stream are processed in-order. For indirect stream instructions, offset elements in an offset list are processed in order. A sync flag is updated to indicate monotonic incremental progress for the stream.
