18091907. MODELING AND COMPILING TENSOR PROCESSING APPLICATIONS FOR A COMPUTING PLATFORM USING MULTI-LAYER ADAPTIVE DATA FLOW GRAPHS simplified abstract (XILINX, INC.)
MODELING AND COMPILING TENSOR PROCESSING APPLICATIONS FOR A COMPUTING PLATFORM USING MULTI-LAYER ADAPTIVE DATA FLOW GRAPHS
Organization Name
Inventor(s)
Chia-Jui Hsu of Santa Clara CA (US)
Mukund Sivaraman of Palo Alto CA (US)
Vinod Kathail of Palo Alto CA (US)
MODELING AND COMPILING TENSOR PROCESSING APPLICATIONS FOR A COMPUTING PLATFORM USING MULTI-LAYER ADAPTIVE DATA FLOW GRAPHS - A simplified explanation of the abstract
This abstract first appeared for US patent application 18091907 titled 'MODELING AND COMPILING TENSOR PROCESSING APPLICATIONS FOR A COMPUTING PLATFORM USING MULTI-LAYER ADAPTIVE DATA FLOW GRAPHS
Simplified Explanation: The patent application discusses modeling and compiling tensor processing applications using multi-layer adaptive data flow (ML-ADF) graphs, with a focus on optimizing resource sharing, kernel execution, memory reuse, and dataflow synchronization.
Key Features and Innovation:
- Modeling tensor processing applications using ML-ADF graphs
- Folding ML-ADF graphs for temporal sharing of platform resources
- Computing schedules for runtime orchestration of kernel execution
- Memory reuse, tensor movement, and dataflow synchronization
- Generating binary code for target computing platforms and re-targetable controller code
Potential Applications: This technology can be applied in various fields such as artificial intelligence, machine learning, image processing, and data analytics.
Problems Solved: The technology addresses challenges related to efficient resource utilization, kernel execution optimization, and dataflow management in tensor processing applications.
Benefits:
- Improved performance and efficiency in tensor processing applications
- Enhanced resource sharing and memory reuse
- Streamlined dataflow synchronization and kernel execution orchestration
Commercial Applications: Optimizing tensor processing applications for various industries such as healthcare, finance, autonomous vehicles, and robotics to improve efficiency and performance.
Prior Art: Prior research in the field of tensor processing, dataflow optimization, and resource sharing in computing platforms can provide valuable insights into the development of this technology.
Frequently Updated Research: Stay updated on the latest advancements in tensor processing, dataflow optimization, and resource sharing techniques to enhance the efficiency and performance of this technology.
Questions about Tensor Processing Technology: 1. How does the ML-ADF graph optimize resource sharing in tensor processing applications? 2. What are the key benefits of using multi-layer adaptive data flow graphs in kernel execution orchestration?
Original Abstract Submitted
Modeling and compiling tensor processing applications using multi-layer adaptive data flow (ML-ADF) graphs, including folding the ML-ADF graph for temporal sharing of platform resources, computing schedules for runtime orchestration of kernel execution, memory reuse, tensor and sub-volume movement, and dataflow synchronization, and generating binary code for processors of the target computing platform and re-targetable controller code. The ML-ADF graph may represent: tensor processing of a layer of a neural network as data flow through the data nodes and distribution to compute tiles across memory hierarchy; data flow amongst layers of the neural network using connections amongst data nodes of the respective layers; and multi-dimension data partitioning and distribution using tiling parameters associated with ports of the data nodes.