Semiconductor Ring Chip Token Power Budget Management

The patent applies preliminary action by requiring chips to collect tokens in advance before performing peak zone operations. Tokens are accumulated during non-peak periods, and only when sufficient tokens are collected can a chip execute peak zone operations. This pre-preparation mechanism ensures that simultaneous operations by multiple chips do not exceed the power budget, as tokens represent pre-allocated energy permissions that constrain total current consumption while still enabling high-performance operations when resources are available.

The patent implements feedback through the token circulation mechanism where tokens are distributed and collected based on power budget status. The system monitors token inventory and adjusts token distribution to chips accordingly - when power budget is tight, fewer tokens are circulated; when budget allows, more tokens are available. This feedback loop dynamically controls operation timing to balance performance with power constraints, preventing simultaneous peak operations that would exceed current limits.

The patent applies dynamics by making the token circulation direction adaptive rather than fixed. While the basic structure uses a unidirectional ring for simplicity, the system dynamically adjusts which chips can proceed with operations based on real-time token availability and power budget status. Chips that have collected sufficient tokens can execute operations independently even if upstream chips are waiting, creating a dynamic execution flow that prevents deadlock while preserving the simple unidirectional token passing structure.

The patent uses preliminary action to prevent deadlock by requiring chips to complete token collection and operation permission acquisition before executing peak zone operations. This pre-validation ensures that chips only proceed when they have the necessary permissions, preventing situations where chips would block each other waiting for tokens. The preliminary token accumulation phase separates the permission-granting process from the actual operation execution, eliminating circular waiting conditions that cause deadlock in simpler unidirectional systems.

The patent applies continuity of useful action by enabling chips to perform non-peak zone operations continuously without waiting for tokens. Peak zone operations, which are more power-intensive, are the only operations requiring token permissions. This distinction allows the system to maintain continuous productive activity at lower power levels while still enabling high-performance peak operations when tokens are available, thus minimizing the impact of waiting on overall system productivity while ensuring power budget compliance.

The patent uses partial action by allowing chips to perform subsets of operations without full token permissions. Non-peak zone operations can be executed partially or continuously without waiting for complete token accumulation, while peak zone operations require full token permissions. This partial execution approach maintains system productivity by keeping chips actively engaged in permissible operations rather than forcing complete idle waiting, while still enforcing power budget constraints on the more intensive peak operations.