Task Scheduling

Progress

• Implemented Largest Task First Scheduling
• Implemented Threshold Based List Scheduling
• Implemented Uniform Scaling
• Implemented Critical Path based Speed Scaling

Contents

This archive includes:

• C++ Source Code used to generate the schedules and compute energy
• Python Scripts to plot the graph
• Task Data of a sample task with different total utilizations (100 in total, each the average of 1000 different random splits (with fixed sum))

Task Data generation takes a long time using TGFF(~30 mins), which is why it is included

How to execute

• Build the executables using GNU Make.

make run.o plot_data.o

Plotting Individual Schedules

• run.o is the executable responsible for generating individual schedules
• To execute it run ./run.o <tgff-file>. For example,

./run.o task-graphs/example.tgff
# Runs the scheduler on the sample testcase

• By default, it generates 6 files:

ltf-schedule
ltf-schedule-cs
ltf-schedule-us

tbls-schedule
tbls-schedule-cs
tbls-schedule-us

Here, cs is an abbreviation for Contingency Schedule and us for Uniform Scaling. No hyphen '-' means it is the default schedule generated.

You should rename these files if you need them as running run.o again will overwrite them.

• All of these files can be visualized using the python script plot_schedule.py. For example,

python3 plot_schedule.py < ltf-schedule-us

Plotting Energy Curves

• plot_data.o is the executable responsible for generating energy information for each schedule
• To execute it run ./plot_data.o <list of tgff-files>. For example,

./plot_data.o task-graphs/graphs/*.tgff
# Runs the scheduler on the generated testcases

Note, this will take some time (~10 seconds) and you might have to increase your stack limit temporarily using ulimit in Linux. I have included the generated files in case you are unable to.

• By default, it generates 1 file:

energy-plot

• To visualize this, use the script plot_energy_avg.py. For example,

python3 plot_energy_avg.py energy-plot