Version 0.1.2 -- 8/22/22
- Installing the simulator
- Running the simulator
- Plot Settings
- Energy Grid Mix File
- Markers File
- Sliders File
- Publications to cite
- Authors and Thanks
Step 1: First, make sure you have python 3 installed on your machine (Ubuntu 22.04 should have it by default). Next, install python 3's package manager, called pip, which will be used to install other python 3 packages, by using the following command:
sudo apt-get install -y python3-pip
Step 2: Next, install git by using the following command:
sudo apt-get install -y git
Step 3: Navigate to your desired directory for Greenchip, and run the following command to download it:
git clone https://github.com/Pitt-Jones-Lab/Greenchip.git
Step 4: Run "cd Greenchip" to navigate to the downloaded tool's directory, and then run the following command to install some of the needed python 3 packages:
pip3 install -r requirements.txt
Step 5: Run the following command to install the last necessary python 3 packages.
sudo apt-get install -y python3-tk idle3 python3-pil python3-pil.imagetk
Step 1: Install python 3 and pip, if you don't have them already. There are many resources available online detailing how to do this.
Step 2: Follow the instructions for windows installation for git from this link:
https://git-scm.com/book/en/v2/Getting-Started-Installing-Git
Step 3: Navigate to your desired directory for Greenchip, and run the following command to download it:
git clone https://github.com/Pitt-Jones-Lab/Greenchip.git
Step 4: Run "cd Greenchip" to navigate to the downloaded tool's directory, and then run the following command to install the needed python 3 packages:
pip3 install -r windows-requirements.txt
Step 1: Make sure you have python3 installed on your machine. From the command line type which python3. If you don't have it the easiest way is to install it via homebrew. Instructions can be found here: https://www.geeksforgeeks.org/how-to-download-and-install-python-latest-version-on-macos-mac-os-x/
Step 1a: Make sure you have pip3 installed. Be careful, you might have two versions of python3 and pip3 installed if you installed using homebrew.
Step 2: Navigate to your desired directory for Greenchip, and run the following command to download it:
git clone https://github.com/Pitt-Jones-Lab/Greenchip.git
Step 3: Install python-tk (note some commands may require sudo)
If you have the regular system python3 and pip3 (which pip3 gives /usr/bin/pip3) then you can install using pip, but it is not recommended. If you installed Homebrew skip to the Homebrew instructions
First upgrade pip3:
pip3 install --upgrade pip
then install python-tk:
pip3 install tk
pip3 install -r $PATHTOGREENCHIP/requirements.txt (from the Greenchip directory)
You MAY need to upgrade tk
curl https://files.pythonhosted.org/packages/a0/81/742b342fd642e672fbedecde725ba44db44e800dc4c936216c3c6729885a/tk-0.1.0.tar.gz > tk.tar.gz
tar -xzvf tk.tar.gz
cd tk-0.1.0
python3 setup.py install
run python3 and then type:
import tk
If you have the homebrew installation (preferred)
brew install python3
brew install python-tk
python3 -m pip install -r $PATHTOGREENCHIP/requirements.txt
python3 SimVis.py
GreenChip SimVis currently has 6 different analysis modes, and an informational plot:
a) Raw Data Entry
This mode allows the user to input the raw data directly into the gui. This avoids requiring the use of a detailed simulator. This mode possess the ability to generate energy and carbon plots.
b) Raw Data Entry Sliders
Similar to Raw Data Entry, except it uses sliders, and it also possesses the ability to push data from single sniper output runs to the sliders, and can generate energy and carbon plots.
c) Indifference
This mode takes an input folder structure, and finds all instances of McPAT and Sniper output in subdirectories, and allows the user to generate specific indifference comparisons they want to see. It will prompt the user for the DRAM specifications.
d) Breakeven
This mode takes an input folder structure, and finds all instances of McPAT and Sniper output in subdirectories, and allows the user to generate specific breakeven comparisons they want to see. It will prompt the user for the DRAM specifications.
e) All Indifference
This mode takes a specific input folder structure defined at the bottom of this section, and outputs all valid indifference comparisons between all data within that folder structure. Runs immediately after hitting "chose analysis mode".
f) All Breakeven
This mode takes a specific input folder structure defined at the bottom of this section, and outputs all valid breakeven comparisons between all data within that folder structure. Runs immediately after hitting "chose analysis mode".
Currently, All_Indifference, All_Breakeven expect a specific format.
Folder Levels 1 through (N-1): Various Parameters/labels (can be anything) Folder Level N: Requires Sniper output files, and DramSim2 output files are preferred for greater accuracy for DRAM power values: sim.out, power.xml, power.py, run_out.txt (*)
An example:
Folder level 1- Number of Cores: 4cores 8cores
Folder level 2- Technology Node: 28 45 90
Folder level 3- Cache size: 1mb 512kb 4mb
Folder level 4- benchmark name: blackscholes canneal x264
Folder level 5- power.py power.xml sim.cfg sim.out sim.stats.sqlite3 run_out.txt (*)
(*) Dram2Sim output, but if it is not found, Greenchip will use the power.py energy specifications for the DRAM power.
As it currently stands, there exists no information about the DRAM technology node or manufacturing and use locations in any of the above files, and while there is a spot for information on the DRAM chip area, it is usually 0. This option allows the user to select a directory, a desired chip area, a desired technology node, and desired locations for manufacturing and usage. A Greenchip.txt file containing this is created and put into each subdirectory containing a power.py (simulator output).
Both Raw Data Entry and Raw Data Entry Slider are able to use importance, which involves perturbing certain System 2 configuration parameters by 1%, and seeing how the breakeven time at a specified sleep and activito ratio reacts. From there it normalizes all of the changes such that they add up to 100, and reports the values back to the user. This lets the user know which values that would be the best for tweaking. There are three methods for selecting sleep percentages and activity ratios to calculate over:
a) Single Point Analysis: This prompts the user for a sleep percentage and an activity ratio, and does all calculations at that point.
b) Average Analysis: This prompts the user for a sleep percentage, an activity ratio, and a radius, and averages all of the pertubations for all of the points within the circle.
c) All Point Analysis: This averages over all of the pertubations for all combinations of sleep percentage and activity ratio.
Both Raw Data Entry and Raw Data Entry Slider modes can generate CSV output, where every value in the plot (either breakeven or indifference) will be outputted, along with its corresponding sleep percentage and activity ratio, to a CSV file in a directory and filename selected by the user.
The plot settings option contains various settings for customizing the plots generated by the tool. Currently, there are two sections, a Range section for setting the minimum and maximum days and years for the colorbars and an Increment section that allows the user to customize the plot resolution for each of the modes.
Note: Increasing the resolution can have a noticable effect on the time it takes to generate a plot.
The EnergyGridMix.csv file contains information about the energy generation percentages for various states and countries. This file can be modified in cases where the user desires values for a different year or wants to add an additional coutry/province. A few example rows of the CSV file are shown:
| Country | State/Province | Coal | Nat. Gas | Geo | Hydro | Solar | Wind | Nuclear | Bio | Other | Petro | Coal % | Nat. Gas % | Geo % | Hydro % | Solar % | Wind % | Nuclear % | Bio % | Other % | Petro % |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| China | 980 | 465 | 0 | 24 | 65 | 11 | 27 | 0 | 81 | 0 | 63.63 | 3.16 | 0 | 15.34 | 3.91 | 7.26 | 4.82 | 0 | 1.75 | 0 | |
| United States | |||||||||||||||||||||
| Alabama | 980 | 465 | 27 | 24 | 65 | 11 | 27 | 54 | 0 | 0 | 16 | 40.5 | 0 | 8.8 | 0.3 | 0 | 32 | 2.5 | 0 | 0 | |
| Alaska | 980 | 465 | 27 | 24 | 65 | 11 | 27 | 54 | 0 | 0 | 12.6 | 37.6 | 0 | 30.5 | 0 | 2.8 | 0 | 0.6 | 0 | 0 |
Each of the power production options have two values: a value for the equivalent gCO2 production per Kilowatt hour (Coal, Geo, ect.) and the percentage of a country's/province's power that comes from that option (Coal %, Geo %, ect.)
When adding a new country with provinces or adding provinces to an existing country, the row containing the country name is left blank as it cannot be selected inside of the tool.
The user also has the option to create entirely new CSV files for Energy Grid Mix data, but it must follow the above format for the columns.
The marker script operates on a line-by-line basis. Each lie is read as a marker command (unless it starts with #, in which case it is assumed to be a comment and thus ignored), and should be structured in the following way
MarkerSpecifier Sleep Activity Radius Label Color
MarkerSpecifier - This should be either "M" or "WM" (not case sensitive.) "M" specifies a regular marker, and "WM" specifies a wide marker.
Sleep - This should be a number that specifies the marker's sleep ratio (should be between 0 and 100)
Activity - This should be a number that specifies the marker's activity ratio (should be between 0 and 100)
Radius - This is a number tat is only used in the "WM" mode to specify the marker's radius. If you are in the "M" mode, don't put anything here (don't put an extra space either)
Label - This specifies the marker's label
Color(optional) - This specifies the marker's color (not case sensitive). It ports the string directly to matplotlib, so check matplotlib's color list to see which one you want to use. If nothing is specified here, then the label will be black with a white bounding box
Some examples are given below:
M 20 40 Phone2 red
This creates a marker with sleep ratio of 20, activity ratio of 40, and a red label of Phone2.
WM 30 60 15 Tablet
This creates a wide marker with sleep ratio of 30, activity ratio of 60, a radius of 15, and a label of Tablet. It uses the default label color scheme of black with a white bounding box
M 30 10 Laptop
This creates a marker with sleep ratio of 30, activity ratio of 10, and a label of Laptop. It again uses the default label color scheme.
Each of the sliders that are used in Raw_Data_Entry_Sliders are customizable through a text file that is read line by line. Each slider is adjusted using the following line:
SliderName min max resolution
min - Value at the far left of the slider. Sliders start at this value
max - Value at the far right of the slider.
resolution - The minimum amount the slider can increment. Clicking the slider bar moves it by this amount.
min, max, and resolution can all be specified as decimals
A list of SliderNames is provided below:
CHIPAREA1, CHIPAREA2, DYNAMICPOWER1, DYNAMICPOWER2, STATICPOWER1, STATICPOWER2, DYNAMICMEMORYPOWER1, DYNAMICMEMORYPOWER2, STATICMEMORYPOWER1, STATICMEMORYPOWER2, DRAMCHIPAREA1, DRAMCHIPAREA2, IPC1, IPC2, FREQUENCY1, FREQUENCY2
The names are NOT case sensitive
Each new slider requires a new line and a new command call
Any sliders not specified default the following values:
CHIPAREA1, CHIPAREA2: min: 0, max: 5000, resolution: 0.01
DYNAMICPOWER1, DYNAMICPOWER2: min: 0, max: 200, resolution: 0.01
STATICPOWER1, STATICPOWER2: min: 0, max: 100, resolution: 0.01
DYNAMICMEMORYPOWER1, DYNAMICMEMORYPOWER2: min: 0, max: 30, resolution: 0.01
STATICMEMORYPOWER1, STATICMEMORYPOWER2: min: 0, max: 30, resolution: 0.01
IPC1, IPC2: min: 0, max: 1000, resolution: 0.01
FREQUENCY1, FREQUENCY: min: 0, max: 5, resolution: 0.01
Some examples are given below:
CHIPAREA1 5 1000 1
The slider ChipArea1 has a minimum value of 5, max of 1000, and the slider increments by 1
DYNAMICPOWER2 200 700 0.1
The slider DynamicPower2 has a minimum value of 200, max of 700, and the slider increments by 0.1
If you publish work using GreenChip, please cite:
Donald Kline, Jr., Nikolas Parshook, Xiaoyu Ge, Erik Brunvand, Rami Melhem, Panos K. Chrysanthis, Alex K. Jones, GreenChip: A Tool for Evaluating Holistic Sustainability of Modern Computing Systems, In Sustainable Computing: Informatics and Systems, 2017.
http://www.sciencedirect.com/science/article/pii/S2210537917300823
This is an extension to our previous work, in IGSC 2016:
Donald Kline, Jr., Nikolas Parshook, Xiaoyu Ge, Erik Brunvand, Rami Melhem, Panos K. Chrysanthis, Alex K. Jones, Holistically evaluating the environmental impacts in modern computing systems, In International Green and Sustainable Computing Conference, 2016.
http://ieeexplore.ieee.org/abstract/document/7892605/
Primary Authors, version 0.1: Nikolas Parshook
Primary Authors, versions 0.2-0.5: Donald Kline, Jr, Nikolas Parshook
Primary Authors, versions 0.6-0.9: Donald Kline, Jr
Primary Authors, version 0.1.0-Current: Ryan Caginalp, Sebastien Ollivier, Stephen Cahoon, Chayanika Choudhuri
Guidance: Dr. Alex K. Jones (University of Pittsburgh) Dr. Panos Chrysanthis (University of Pittsburgh)
Thanks: Eric Cheung, for his work on the indifference plots