This is one part of the course assignment for HKUST-GZ MICS 6000I Physical Design Automation of Digital Systems. This project is alive, maintained by [email protected]. Any discussion or suggestion would be greatly appreciated!

In this repo, the shelf floorplanning and skyline floorplanning methods are implemented, together with an optional optimization with the Guillotine algorithm.

• System Checking
  • Ubuntu OS: 20.04.5 LTS
  • Linux: 5.14.0-1054-oem
  • g++ 9.4.0
• To install Gurobi on Linux
  • Obtain an academic license at https://www.gurobi.com/downloads/end-user-license-agreement-academic/
  • Set the environment variables:

  # Gurobi Basics
  export GUROBI_HOME=<path-to-gurobi-home-directory>
  export PATH="${GUROBI_HOME}/bin:${PATH}"
  export LD_LIBRARY_PATH="${GUROBI_HOME}/lib:${LD_LIBRARY_PATH}"
  # For '#include "gurobi_c++.h"' in C++ source code
  export CPATH="${GUROBI_HOME}/include:${CPATH}"
  # License
  export GRB_LICENSE_FILE=<path-to-license-file-.lic>

• Python packages (for visualization)
  • matplotlib 3.5.1
• To enable #include <graphics.h> in C++ (poor visualization quality, Optional)
  1. sudo apt-get install build-essential
  2. sudo apt-get install libsdl-image1.2 libsdl-image1.2-dev guile-2.0 guile-2.0-dev libsdl1.2debian libart-2.0-dev libaudiofile-dev libdirectfb-dev libdirectfb-extra libfreetype6-dev libxext-dev x11proto-xext-dev libfreetype6 libaa1 libaa1-dev libslang2-dev libasound2 libasound2-dev
  3. sudo vim /etc/apt/sources.list, and paste the following sources

     deb http://us.archive.ubuntu.com/ubuntu/ xenial main universe
     deb-src http://us.archive.ubuntu.com/ubuntu/ xenial main universe
     

  4. sudo apt update
  5. sudo apt install libesd0-dev
  6. wget http://download.savannah.gnu.org/releases/libgraph/libgraph-1.0.2.tar.gz
  7. tar zxvf libgraph-1.0.2.tar.gz
  8. cd libgraph-1.0.2/
  9. CPPFLAGS="$CPPFLAGS $(pkg-config --cflags-only-I guile-2.0)" CFLAGS="$CFLAGS $(pkg-config --cflags-only-other guile-2.0)" LDFLAGS="$LDFLAGS $(pkg-config --libs guile-2.0)" ./configure
  10. make
  11. sudo make install
  12. sudo cp /usr/local/lib/libgraph.* /usr/lib

• To compile the C++ code:

  make

• To run all the testcases one-and-for-all
  • python run_all.py is an integrated script calling the C++ executable repetitively
  • Note that we only run MILP for 10-block case in this script because others lead to timeout (>2 hours)
• To generate floorplan for specific cases:

  ./floorplan ./testcase/50.txt 0 7000
  # 0 for MILP, 7000 means the program will terminate in 7000 seconds if not optimally solved yet
  
  # ./floorplan ./testcase/<testcase_name> <alg_type> <sub_alg_type> <sorting_type>, e.g.,
  ./floorplan ./testcase/500.txt 1 0 0

  • Standard output: the required floorplan report in terminal
  • Log file for visualization: *.log (without "verbose" keyword) in ./log folder
  • Detailed log (only for MILP): *.log with "verbose" keyword in ./log folder
• To visualize all logs and save images in the directory ./img:

  python vis_all.py

• To visualize certain log:

  python visualize.py ./log/<xxx>.log
  # output in base directory as "floorplan.png"

• To see the best solutions under each algorithm:

  python find_best.py
  # standard output in the terminal for each testcase, and for each type of algorithm

• How to compile C++ with Gurobi on Linux
• Gurobi Documents
  • Gurobi Optimizer Reference Manual
  • C++ API Example -- mip1_c++.cpp
• How to install "graphics.h" on Linux