Analytical Model
The repository contains Python scripts for analytical model for distributed memory \(k\)-mer counting.
The default scripts will reproduce plots from our conference paper, based on hardware parameters and experiments performed using the Phoenix supercomputer at Georgia Tech.
Description
kcount.py
: Funtion definitions for the analytical model of distributed-memory k-mer counting.memory.py
: Seperate model to show memory overhead of multi-layered message aggregation.params.py
: Machine parametes used for the analytical model. Edit this file with parameters of your target machine.experiments.py
: Input and experimental values of k-mer counting for different synthetic datasets, as observed on the Phoenix machine. Edit this file with updated input and experimental results on your target machinedefaultplot.py
: Default plotting options.cachepred.py
: Analytically predicts the L3 cache misses for inputs mentioned inexperiments.py
and compares them against experimental results.hwresource.py
: Analytically predicts what percentage of k-mer counting is spent doing "memory access", "communication", and "computation".
How to execute
Edit the params.py
and experiments.py
file as required. Then simply run python <scriptname>.py
. The script will produce a figure inside the figures
directory.