IC Placement Routing Incremental Optimization

The patent segments the violation cells into different groups based on their spatial locations and routing relationships. Instead of treating all violation cells uniformly, the method divides them into multiple batches, where each batch contains cells that can be optimized independently without affecting each other's routing environment. This segmentation allows the optimization to proceed in a controlled manner that maintains timing stability while improving convergence.

The patent implements a dynamic optimization approach where the placement and routing environment is updated incrementally after each violation cell is optimized, rather than maintaining a static default environment throughout. The method dynamically adjusts the routing environment to reflect the actual changes made during optimization, ensuring that subsequent violations are evaluated in the context of the updated environment. This dynamic adaptation prevents timing jumps and improves convergence reliability.

The patent performs preliminary actions by pre-calculating and storing the routing environment state before each optimization batch. Before optimizing a group of violation cells, the method captures the current routing environment as a reference state. This preliminary action allows the system to evaluate whether subsequent violations are genuine or artifacts of environmental changes, thereby maintaining measurement precision while still adapting to real environment changes.

The patent implements a feedback mechanism where the timing analysis results and violation states are continuously monitored and fed back into the optimization process. After each optimization step, the method re-evaluates the placement and routing environment to determine if new violations have occurred or if existing violations have been resolved. This feedback loop ensures that the optimization adapts to real environment changes while maintaining accurate knowledge of the violation state, preventing timing jumps through iterative refinement.

The patent segments violation cells into multiple independent batches based on spatial and routing relationship analysis. Cells that are spatially close or share routing resources are grouped together and optimized in the same batch, while cells that are far apart or have independent routing paths are placed in separate batches. This segmentation enables parallel optimization of independent cells (maintaining productivity) while ensuring that cells affecting the same routing environment are optimized sequentially (maintaining accuracy).

The patent applies partial optimization by focusing on optimizing only the necessary number of violation cells in each batch, rather than attempting to optimize all violation cells simultaneously. The method determines the optimal batch size based on the complexity of the routing environment and the number of violations, performing partial optimization iterations until convergence. This approach balances productivity (by optimizing multiple cells) with accuracy (by limiting batch size to maintain environment stability).