Kernel Configuration Compilation for Valid Streaming Link Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Compilation of compute kernels to implement target functionality is complicated and time-consuming due to the complexity of managing compatible and valid kernel configurations, especially when logical inputs are spread across independent physical streaming links.
Innovation Solution
A kernel library that stores definitions mapping logical inputs and outputs to physical inputs and outputs, allowing an automatic compiler to identify valid kernel configurations and optimize them for a target computing device.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If manual compilation methods are used to select and verify compute kernels, then flexibility in implementing various compute kernels is improved, but compilation time and complexity increase significantly
Solution Approach 1:
The patent applies preliminary action by pre-defining kernel definitions that specify the relationship between logical inputs/outputs and physical inputs/outputs before compilation. These definitions are stored in a library and automatically retrieved during compilation, eliminating the need for manual verification of kernel compatibility and reducing compilation time while maintaining flexibility.
Solution Approach 2:
The patent introduces kernel definitions as an intermediary layer between the compute kernel library and the compilation process. These definitions act as mediators that automatically encode compatibility rules and validation logic, allowing the compiler to efficiently verify kernel configurations without manual intervention while preserving the ability to implement diverse kernels.
2Reliability
If comprehensive validation of kernel configurations is performed to ensure validity and compatibility, then reliability of kernel configuration is improved, but compilation complexity increases
Solution Approach 1:
The patent reduces compilation complexity by performing validation work in advance through pre-defined kernel definitions. These definitions contain all necessary compatibility rules and validation logic, allowing the compiler to automatically verify kernel configurations without complex manual checks, thus maintaining high reliability while simplifying the compilation process.
Solution Approach 2:
The kernel definitions serve as intermediaries that encapsulate complex validation rules. By introducing this intermediary layer, the patent separates the complexity of validation logic from the compilation process itself, allowing the compiler to rely on pre-prepared definitions rather than implementing complex validation algorithms, thereby reducing compilation complexity while ensuring reliability.
3Ease of operation
If detailed mapping definitions are stored for each compute kernel to manage physical and logical inputs/outputs, then ease of automatic compilation is improved, but memory requirements and library size increase
Solution Approach 1:
The patent applies universality by designing kernel definitions with a standardized structure that can be applied across different compute kernels. This multi-functional definition format allows the same template to describe various kernels, reducing the overall library size while maintaining ease of automatic compilation through consistent, reusable definition patterns.
Solution Approach 2:
The patent manages library size by using parameterized kernel definitions that can adapt to different kernels through parameter substitution rather than storing complete unique definitions for each kernel. This approach maintains the ease of automatic compilation by preserving detailed mapping information while reducing redundancy in the library storage.
Data Source
AI summary
An example method includes: obtaining a logical netlist comprising a plurality of nodes and node connections representing a network; storing a library of compute kernels, each compute kernel configured to implement one or more of the nodes in the logical netlist; storing a kernel definition associated with each of the compute kernels, the kernel definition mapping (i) a physical input source to a logical input of the compute kernel and (ii) a logical output of the compute kernel to a physical output; and selecting a kernel configuration satisfying a compilation condition, the kernel configuration comprising (a) a subset of the compute kernels, such that each node is covered by at least one compute kernel, and (b) links between the compute kernels in the subset, wherein the links are defined based on the kernel definitions; and compiling the selected kernel configuration to implement the network.