Abstract:The improvement of chip integration leads to the increase of power density of system chips, which leads to the overheating of system chips. When dispatching the power density of system chips, some working modules are selectively closed to avoid all modules on the chip being turned on at the same time and to solve the problem of overheating. Taking 2D grid-on-chip network as the research object, an optimal scheduling algorithm of system-on-chip power density based on network-on-chip (NoC) is proposed. Under the constraints of thermal design power (TDP) and system, dynamic programming algorithm is used to solve the optimal application set throughput allocation from bottom to top by dynamic programming for the number and frequency level of each application configuration processor under the given application set of network-on-chip. On this basis, the simulated annealing algorithm is used to complete the application mapping aiming at heat dissipation effect and communication delay. The open and closed processor layout is determined. After obtaining the layout results, the TDP is adjusted. The maximum TDP constraint is iteratively searched according to the feedback loop of the system over-hot spots, and the power density scheduling performance of the system chip is maximized under this constraint, so as to ensure the system core. At the same time, chip throughput can effectively solve the problem of chip overheating. The experimental results show that the proposed algorithm increases the system chip throughput by about 11%, improves the system throughput loss, and achieves a balance between the system chip power consumption and scheduling time.

A genetic algorithm based method for the uniformity of power in SoC during dynamic burn-in

Power-and thermal-aware mapping for NoC under latency constraint

Power-aware Mapping Based-on Genetic Algorithm for Network-on-Chip

Self-heating Burn-in Pattern Generation Based on the Genetic Algorithm Incorporated with a BACK-like Procedure.

A test pattern selection method of dynamic burn-in for logic circuits

A Test Pattern Selection Method for Dynamic Burn-In of Logic Circuits Based on Atpg Technique

Design Time Body Bias Selection for Parametric Yield Improvement.

Test Scheduling for IP Cores Optimization Based on the Genetic Algorithm

Test pattern generation for static burn-in based on equivalent fault model

A Test Pattern Selection Method for Static Burn-In of ICs

Niche Genetic Algorithm Based Optimization of Test Access Mechanism for System-on-chip

Low-power Weighted Random Pattern Testing

A Layout Optimization Method to Reduce Commutation Inductance of Multi-Chip Power Module Based on Genetic Algorithm

Genetic Algorithm Based Partition Design For Test (DFT)

A Test Scheduling Scheme For Core-Based Socs Using Genetic Algorithm

Genetic Algorithm Applied to Programmable Current Adder

SoC Test Scheduling with Hot-Spot Avoidance and Even Heat Distribution

Optimal Scheduling Algorithms of System Chip Power Density Based on Network on Chip

3D NoC Low-Power Mapping Optimization Based on Improved Genetic Algorithm

A Generation Method for Test Patterns with Low Power and Low Cost

CA optimization based on simulation annealing in BIST