多任务物理映射方法及装置
By generating task sequences and using neural networks to determine target physical kernels for multi-task physical mapping, the problem of system performance degradation caused by physical kernel fragmentation is solved, achieving more efficient physical kernel utilization and improved system performance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TSINGHUA UNIVERSITY
- Filing Date
- 2023-04-18
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies tend to exacerbate the fragmentation of physical cores in chips during multi-task physical mapping, leading to a decline in system performance and making it impossible to perform multi-task physical mapping efficiently and reasonably.
By generating task sequences, utilizing the policy network in the neural network and the current physical core deployment state, the target physical core is determined from multiple physical cores, the target task is physically mapped to the target physical core, and the current physical core deployment state is updated. The policy network parameters are adjusted by combining the reward function and the value network to optimize the physical mapping process.
It improves the speed and efficiency of multi-task physical mapping, reduces physical kernel fragmentation, and enhances system performance.
Smart Images
Figure CN116467095B_ABST