forked from opendacs/EMSim_plus
Update README.md
This commit is contained in:
parent
7c19bc9f05
commit
f757dd0272
47
README.md
47
README.md
|
@ -9,6 +9,13 @@ EMSim+是一款基于生成对抗网络(GAN)的版图级芯片电磁仿真
|
|||
<img src="./docs/EMSim+ vs. EMSim.jpg" width=395px>
|
||||
|
||||
|
||||
# 环境配置
|
||||
|
||||
- 操作系统:Linux or Windows
|
||||
|
||||
- 依赖项:Python 3.8+ with PIP, TensorFlow2.4 with GPU
|
||||
|
||||
|
||||
|
||||
# Table of contents
|
||||
- [Prerequisites](#prerequisites)
|
||||
|
@ -21,15 +28,9 @@ EMSim+是一款基于生成对抗网络(GAN)的版图级芯片电磁仿真
|
|||
- [Contributors](#contributors)
|
||||
- [Copyright](#copyright)
|
||||
|
||||
# Prerequisites
|
||||
At a minimum:
|
||||
# 实现框架
|
||||
|
||||
- Python 3.8+ with PIP
|
||||
- TensorFlow2.4 with GPU
|
||||
- Linux or Windows
|
||||
|
||||
# Running EMSim
|
||||
EMSim+ consists of three main steps: feature extraction, GAN model training and EM prediction.
|
||||
EMSim+的实现框架包含特征提取,GAN模型训练和电磁预测三部分。
|
||||
|
||||
<table>
|
||||
<tr>
|
||||
|
@ -40,14 +41,14 @@ EMSim+ consists of three main steps: feature extraction, GAN model training and
|
|||
|
||||
## Current Analysis and Electromagnetic Computation
|
||||
|
||||
These two steps are from EMSim and are intended to subsequently produce training data for GANs.
|
||||
这两部分是使用EMSim完成的,参考链接:
|
||||
由Current Analysis and Electromagnetic Computation获得的单元电流信息、电源位置信息和电磁计算结果。
|
||||
|
||||
## Feature Extraction
|
||||
|
||||
Feature extraction aims to extract cell current, power grid and EM information from the database of the chip physical layout.
|
||||
Then, we convert them into feature maps.
|
||||
特征提取的目的是从芯片物理布局数据库中提取单元电流、电源网格和电磁信息,并将其转换为特征图。
|
||||
|
||||
### Cell Current Map
|
||||
### 单元电流地图
|
||||
|
||||
```
|
||||
create_current_map.py
|
||||
|
@ -68,7 +69,7 @@ optional arguments:
|
|||
|
||||
```
|
||||
|
||||
### Power Grid Map
|
||||
### 电源网格地图
|
||||
|
||||
```
|
||||
create_grid_map(2-pad).py
|
||||
|
@ -95,14 +96,16 @@ optional arguments:
|
|||
[ --layout_min_y ] Reference coordinate in y axial direction
|
||||
|
||||
```
|
||||
There is only one power grid map of one layout design, which is used in both the model training and EM prediction phases.
|
||||
- Note:
|
||||
- You need to get the coordinates of VDD and VSS in the layout design.
|
||||
- You should choose different scripts for 2-pad and 4-pad power supply designs.
|
||||
|
||||
|
||||
对于一个版图级设计,只有一个电源网格地图,在模型训练和电磁预测阶段均使用该电网图。
|
||||
- 注意
|
||||
- 需要在版图级设计中获取 VDD 和 VSS 的坐标。
|
||||
- 对于 2-pad 和 4-pad 的不同电源设计需要选择不同的脚本。
|
||||
|
||||
|
||||
|
||||
### EM Map
|
||||
### 电磁地图
|
||||
|
||||
```
|
||||
create_em_map.py
|
||||
|
@ -121,11 +124,7 @@ optional arguments:
|
|||
## GAN Model Training
|
||||
|
||||
|
||||
We aim to design and train a GAN for EM prediction.
|
||||
- Note:
|
||||
- The generator G accepts cell current maps, power grid maps and time sequence.
|
||||
- Both the EM maps predicted by G and the real EM maps, together with the input maps of G, are alternatively fed to the discriminator D for determination.
|
||||
- The results of D are further fed back to G to enhance the quality of the predicted EM maps
|
||||
GAN网络采用U-net结构搭建生成器,采用PatchGAN结构搭建判别器。
|
||||
|
||||
|
||||
```
|
||||
|
@ -148,7 +147,7 @@ optional arguments:
|
|||
```
|
||||
## EM Prediction
|
||||
|
||||
The generator G is preserved and serves as an inference model for EM prediction.
|
||||
在电磁预测阶段,训练好的生成器被保留下来,用做推理模型。
|
||||
|
||||
```
|
||||
GAN4EM_prediction.py
|
||||
|
|
Loading…
Reference in New Issue