This repository includes files to perform a speed test between named pipes and shared memory. This was made for Advanced Robot Programming course of UniGe.

Choose a directory to work in, then clone the repo by typing

git clone https://github.com/EbruBaglan/IPC_benchmark

After cloning the repo, just hit

cd IPC_benchmark && chmod +x *.sh

to make .sh files executable.

Then, specify the folder to install the files into by typing . install.sh <pathname>. For instance

. install.sh ./fileshere

To see the performance of named pipe for 100 MB data transfer time, type

./run.sh 2 102400

Number 2 is specified to select 'named pipe'. (In later implementations 1 is used for unnamed pipes, 3 is for sockets)

For the shared memory scenario, the circular buffer size should also be specified as the fourth argument. For a size of 2KB circular buffer, type

./run.sh 4 102400 2

The results are somehow perplexing as the shared memory seems slower than the named pipe. In reality, it should be the other way around. The investigation is still going on.