Efficient utilization of synchronization primitives in multiprocessor computing system
By creating a dependency graph and automatically allocating synchronization primitives through the compiler, the problem of manual optimization difficulties in multiprocessor computing systems is solved, achieving more efficient asynchronous command execution and performance improvement.
Patent Information
- Application Number
- CN202480069191.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2023-11-30
- Filing Date
- 2024-10-17
- Publication Date
- 2026-06-02
AI Technical Summary
In multiprocessor computing systems, manually optimizing program code to coordinate asynchronous command execution using synchronization primitives is difficult and time-consuming, resulting in poor performance. Furthermore, existing technologies struggle to effectively utilize limited synchronization primitive resources.
The compiler creates a dependency graph and automatically allocates synchronization primitives, optimizes program code to coordinate asynchronous command execution, identifies dependencies using shared queue analysis, input/output analysis, and allocation overlap analysis, minimizes redundant edges, and adds back edges to enforce circular dependencies.
It enables more optimized utilization of synchronization primitives, improves the execution performance of multiprocessor computing systems, and eliminates the difficulty and time-consuming nature of manual optimization.