Chips & computingResearch findings
TDPNavigator-Placer · Thermal and Wirelength-Aware Chiplet Placement via Multi-Agent Reinforcement Learning
Original published 2026.01.30 18:04 KSTAdded to Blackdeck 2026.11.18 13:41 KST
What has changed?
Conflicting objectives of wirelength and thermal performance are separated into specialized agents with distinct reward functions.
Key points
- As 2.5D systems grow, balancing thermal and wirelength performance becomes essential.
- Existing placement methods combine multiple objectives into a single one, limiting practicality.
- TDPNavigator-Placer uses thermal design power-based multi-agent reinforcement learning to achieve balanced placement.
Connect it to your work
The method achieves a more balanced trade-off between wirelength and thermal performance in the Pareto front.
What remains unproven
This research note is based on an abstract; check the original for test conditions and application scope.
Originals used for this note
Links open the original publisher in a new tab.