Installation

Docker

This is the recommened method for new users. Accelergy and Timeloop are avialable in the same docker.

  • If you have already installed docker for Timeloop, you are all set for Accelergy.
  • Otherwise, please follow the docker installation instructions here.

Native Install

Prerequisites

  • Infrastructure tested on RedHat Linux, Ubuntu, MacOS
  • Required packages
    • Python >= 3.6
    • PyYAML >= 1.1 (dependency automatically handled at installation)
    • yamlordereddictloader >= 0.4 (dependency automatically handled at installation)

Install the Accelergy package

git clone git@github.com:Accelergy-Project/accelergy.git
cd accelergy
pip3 install .

Install Accelergy plug-ins

Accelergy communitcates with various energy/area estimation plug-ins to generate estimations for various technologies. We provide default plug-ins to get you started with estimating popular components in common architectures.

Install Cacti plug-in

Responsible for generating large SRAM energy/area for various technology nodes and DRAM energy. 

git clone https://github.com/HewlettPackard/cacti
cd cacti
make
cd ..
mv cacti ~/.local/bin/
git clone git@github.com:Accelergy-Project/accelergy-cacti-plug-in.git
cd accelergy-cacti-plug-in
pip3 install .

Install Aladdin plug-in

Responsible for generating the energy/area of various 45nm datapath components, e.g., intadder, intmultiplier, register files, etc.

git clone git@github.com:Accelergy-Project/accelergy-aladdin-plug-in.git
cd accelergy-aladdin-plug-in
pip3 install .

Install table-based plug-in

Allows easy integration of user-defined csv table-based plug-ins. More detailes will be provided in plug-in API explanation.

git clone git@github.com:Accelergy-Project/accelergy-table-based-plug-ins.git
cd accelergy-table-based-plug-ins
pip3 install .
accelergyTables

Test if the Accelergy package is installed

  • Please make sure your python bin, e.g.,~/.local/bin, is appropriately added to $PATH
  • A new command: accelergy should be available in your python bin
  • accelergy -h shows the help message for the command