accelergy-table-based-plug-ins

An energy estimation plug-in for Accelergy framework. This plug-in provides an entry point for easy creation of various energy tables.

Get started

Install Accelergy framework (if you have not yet).
Run pip install . with the same options you used for installing Accelergy.
The installation installs a command accelergyTables to your system.

Try the provided example set of table templates

Populate the provided table templates (can skip if don't want to add more data to templates)

At least one row of example filled cells are provided in the templates. Populate the cells with your own data. Run pip install . to reinstall the updated data files

Add root

Add the root to the provided set of tables by running accelergyTables.

Run accelergy

You should see the following in the printed log：

The table-based-plug-ins is found by accelergy.
The table-based-plug-ins identifies a set of tables called test_tables.

Note that most of the provided energy tables do not contain numerical data.

Create your own set of tables

Add the root to your sets of tables

Add the root to your sets of tables to the Accelergy config file by running

accelergyTables -r $myRoot

You can also manually add the root in the ~/.config/accelergy/accelergy_config.yaml file.

The plug-in will recursively search the subdirectories of the provided root to locate various sets of tables

YAML identifier file

A YAML identifier file that describes the basic information of your set of tables needs to be created.
set_of_table_templates contains an example YAML identifier file test.table.yaml.

The name of the YAML file must end with table.yaml and following information needs to be provided in the file:

The name of the set of tables
The supported technology
The accuracy of your set of tables
The path to the directory that contains all your csv data files

CSV data tables

You can create your csv tables anywhere in you file structure as long as the path pointer in its identifier file is correct.

set_of_table_templates/data folder provides the necessary headers for all default primitive component classes.
counter.csv and regfile.csv are populated with random data to show what should be filled in the cells.

There are several rules for creating these data files:

Data tables must be in csv format.
The name of the file should be <primitive component class name>.csv, e.g., SRAM.csv for SRAM energy data.
The necessary hardware attribute names, action names, argument names are the headers(the first row) of the csv, and their corresponding values are filled in the cells.
The energy column must have a header called "energy".

tanner-andrulis/accelergy-table-based-plug-ins