Readme.md

Requirements

• Install the Intel Power Gadget Tool In order to have access to the power measurements of the machine

• install the python libraries pandas, and tabulate

Usage

Ijoules can be used in 3 different ways

Basic usage

from ijoules import IJoules

def foo(): 
    energy_meter= IJoules()
    energy_meter.configure() 
    # ...
    energy_meter.begin() 
    # Code that we want to measure the nergy consumption
    energy_meter.record("first part of the code ")
    # Another measurement 
    energy_meter.end()
    data = energy_meter.report # a dataframe containing all the energy 

Use it as a context

With the instruction with

from ijoules import IJoules

def foo() :
    with Ijoules() as energy_meter : 
        # Work that we want to measure 
        energy_meter.record("First part of the work")
        # Second part of the work 
        energy_meter.record("Second part of the work")
        # Second part of the work 

    # This will print the energies as a table 

Use it as a decorator

from ijoules import  measureit

@measureit()
def foo(): 
    # Code that we want to measure 


def main() 
    foo() 
    # This will print the energy of the function as a table 

Example

	pkg_j	core_j	dram_j	duration_s	Tag
1	13.5483	11.4319	0.408875	0.610369	worker1
2	13.5273	11.413	0.367126	0.60813	second time
3	10.4964	3.47644	1.36255	2.00226	end

IJoules allow you to measure different parts of your CPU based on their availablity. To know the available parts that can be measured you can call the methdoe get_available_domains

energy_meter= IJoules()
energy_meter.configure()
energy_meter.get_available_domains()

The possible domains (CPU parts) are :

• core : The energy consumption of the core units -in Joules-
• dram : The energy consumption of DRAM -in Joules-
• gpu : The energy consumption of intergated GPU -in Joules-
• pkg : The whole energy consumption of the whole CPU (Core+GPU+other parts) -in Joules-

The figure below will illustrate more those parts:

References :

• https://blog.chih.me/read-cpu-power-with-RAPL.html
• https://software.intel.com/content/www/us/en/develop/articles/intel-power-gadget.html