Chip instance placement legalization method based on reinforcement learning

The invention discloses a legalization method for chip instance placement based on reinforcement learning, which can dynamically adapt to various placement rules and manufacturing limitations, including processing units with different heights and sizes. Therefore, the algorithm is not only suitable for the placement of the standard cell, but also can effectively process the placement problem of the macro block. According to the method, learning can be carried out from the past placement practice, and the placement strategy is continuously optimized. The learning ability enables the algorithm to continuously progress along with the development of the technology and the change of the design demand, and can better cope with the rapid iteration of the integrated circuit design. Moreover, when facing a large-scale chip, the reinforcement learning algorithm also can complete legalized placement of the chip instance within limited time. In a word, according to the method, the flexibility and accuracy of chip placeme