Generating integrated circuit placements using neural networks

Patent metadata

Field	Value
Publication number	US-12248745-B2
Application number	US-202318395251-A
Country	US
Kind code	B2
Filing date	Dec 22, 2023
Priority date	Apr 22, 2020
Publication date	Mar 11, 2025
Grant date	Mar 11, 2025

Abstract

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating a computer chip placement. One of the methods includes obtaining netlist data for a computer chip; and generating a computer chip placement, comprising placing a respective macro node at each time step in a sequence comprising a plurality of time steps, the placing comprising, for each time step: generating an input representation for the time step; processing the input representation using a node placement neural network having a plurality of network parameters, wherein the node placement neural network is configured to process the input representation in accordance with current values of the network parameters to generate a score distribution over a plurality of positions on the surface of the computer chip; and assigning the macro node to be placed at the time step to a position from the plurality of positions using the score distribution.

First claim

What is claimed is:

1. A method of training a node placement neural network that comprises:
an encoder neural network that is configured to, at each of a plurality of time steps, receive an input representation comprising data representing a current state of a placement of a netlist of nodes on a surface of an integrated circuit chip as of the time step and process the input representation to generate an encoder output, and
a policy neural network configured to, at each of the plurality of time steps, receive an encoded representation generated from the encoder output generated by the encoder neural network and process the encoded representation to generate a score distribution over a plurality of positions on the surface of the integrated circuit chip, the method comprising:
obtaining supervised training data comprising:
a plurality of training input representations, each training input representation representing a respective placement of a respective netlist of nodes, and
for each training input representation, a respective target value of a reward function that measures a quality of the respective placement of the respective netlist of nodes; and

training at least the encoder neural network on the plurality of training input representations using the target values of the reward function through supervised learning;
after the training through supervised learning:
receiving a new netlist of nodes; and
training the node placement neural network on the new netlist of nodes through reinforcement learning.

2. The method of claim 1 , wherein training the node placement neural network through reinforcement learning comprises training the policy neural network through reinforcement learning to generate score distributions that result in placements for the new netlist of nodes that maximize the reward function.

3. The method of claim 2 , wherein training the policy neural network through reinforcement learning comprises holding values of parameters of the encoder neural network fixed during the training of the policy neural network through reinforcement learning.

4. The method of claim 1 , further comprising:
after training through reinforcement learning, generating an integrated circuit placement for the new netlist data using the node placement neural network, comprising placing a respective node from the new netlist data at each of a plurality of time steps using score distributions generated by the node placement neural network.

5. The method of claim 1 , wherein the reward function includes a wire length term that measures a wire length of wires on the surface of the integrated circuit chip.

6. The method of claim 1 , wherein the reward function includes a congestion term that measures congestion on the surface of the integrated circuit chip.

7. The method of claim 1 , wherein the reward function includes a timing term that measures a timing performance of the integrated circuit chip.

8. The method of claim 1 , further comprising:
generating the supervised training data, comprising:
obtaining data specifying a training accelerator netlist;
generating a plurality of placements for the training accelerator netlist using a different policy neural network; and
determining a respective value of the reward function for each of the plurality of placements for the training accelerator netlist.

9. A system comprising one or more computers and one or more storage devices storing instructions that when executed by the one or more computers cause the one or more computers to perform operations for training a node placement neural network that comprises:
an encoder neural network that is configured to, at each of a plurality of time steps, receive an input representation comprising data representing a current state of a placement of a netlist of nodes on a surface of an integrated circuit chip as of the time step and process the input representation to generate an encoder output, and
a policy neural network configured to, at each of the plurality of time steps, receive an encoded representation generated from the encoder output generated by the encoder neural network and process the encoded representation to generate a score distribution over a plurality of positions on the surface of the integrated circuit chip, the method comprising:
obtaining supervised training data comprising:
a plurality of training input representations, each training input representation representing a respective placement of a respective netlist of nodes, and
for each training input representation, a respective target value of a reward function that measures a quality of the respective placement of the respective netlist of nodes; and

training at least the encoder neural network on the plurality of training input representations using the target values of the reward function through supervised learning;
after the training through supervised learning:
receiving a new netlist of nodes, and
training the node placement neural network on the new netlist of nodes through reinforcement learning.

10. The system of claim 9 , wherein training the node placement neural network through reinforcement learning comprises training the policy neural network through reinforcement learning to generate score distributions that result in placements for the new netlist of nodes that maximize the reward function.

11. The system of claim 10 , wherein training the policy neural network through reinforcement learning comprises holding values of parameters of the encoder neural network fixed during the training of the policy neural network through reinforcement learning.

12. The system of claim 9 , the operations further comprising:
after training through reinforcement learning, generating an integrated circuit placement for the new netlist data using the node placement neural network, comprising placing a respective node from the new netlist data at each of a plurality of time steps using score distributions generated by the node placement neural network.

13. The system of claim 9 , wherein the reward function includes a wire length term that measures a wire length of wires on the surface of the integrated circuit chip.

14. The system of claim 9 , wherein the reward function includes a congestion term that measures congestion on the surface of the integrated circuit chip.

15. The system of claim 9 , wherein the reward function includes a timing term that measures a timing performance of the integrated circuit chip.

16. The system of claim 9 , the operations further comprising:
generating the supervised training data, comprising:
obtaining data specifying a training accelerator netlist;
generating a plurality of placements for the training accelerator netlist using a different policy neural network; and
determining a respective value of the reward function for each of the plurality of placements for the training accelerator netlist.

17. One or more non-transitory computer-readable storage media storing instructions that when executed by the one or more computers cause the one or more computers to perform operations for training a node placement neural network that comprises:
an encoder neural network that is configured to, at each of a plurality of time steps, receive an input representation comprising data representing a current state of a placement of a netlist of nodes on a surface of an integrated circuit chip as of the time step and process the input representation to generate an encoder output, and
a policy neural network configured to, at each of the plurality of time steps, receive an encoded r

Assignees

• Google Llc

Inventors

• Goldie Anna Darling
• Mirhoseini Azalia
• Songhori Ebrahim
• Jiang Wenjie
• Wang Shen
• Carpenter Roger David
• Lee Young-Joon
• Yazgan Mustafa Nazim
• Ho Chian-Min Richard
• Le Quoc V
• Laudon James
• Dean Jeffrey Adgate
• Srinivasa Setty Kavya
• Pathak Omkar

Classifications

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/092

  Reinforcement learning · CPC title

• G06N3/098

  Distributed learning, e.g. federated learning · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title