Batch process files containing benchmarking data of computer programs, one file per run.

Determine the mean and standard deviation of benchmarking experiments with several runs.

Organize the benchmark statistics by program

Output scalability and speedup data, optionally generating associated figures.

Create publication ready benchmark comparison tables in LATEX.

These tools were originally developed to assess the performance of serial and parallel implementations of the PPHPC simulation model [

Performance analysis in

Although the

^{5}, 2 x 10^{5}, 3 x 10^{5} and 4 x 10^{5}. For each size, individual algorithms are executed ten times. Each run is benchmarked with GNU time, the output of which is redirected to a file with the following identifiers in its name: algorithm employed, run number and vector size.

In this context, a sorting algorithm is an

In the following paragraphs, implementations specs stipulating all the setups (i.e., vector sizes) for the Bubble sort, Selection sort, Merge sort and Quicksort algorithms are represented by the

Implementation specs for the remaining algorithms are defined in a similar fashion. Note that all implementations specs must have the same number of setups, and corresponding setups should have the same sname. Additionally, plotting with

The contents of the returned variables,

The

Scalability plot generated by the

The first parameter concerns the optional bar plot the function is able to generate. An absolute value of

Speedup plot generated by the

In a similar fashion to the mean and standard deviation matrices returned by

The

The first argument designates the references implementation or implementations, in a similar fashion to the second parameter of

The first argument,

Plain text table generated by

LATEX table generated by

The available functions are covered by unit tests in order to ensure their correct behavior. The

Any system capable of running MATLAB R2013a or GNU Octave 3.8.1, or higher.

MATLAB R2013a or GNU Octave 3.8.1, or higher.

There are no additional dependencies for the package tools. However, unit tests depend on the

The software was created by Nuno Fachada.

English.

These utilities can be used for analyzing any computational experiment. As described in ‘Implementation and architecture’, other benchmark data formats can be specified by implementing a custom function to replace

The authors declare that they have no competing interests.

This software is enhanced by the

e.g., files containing the output of GNU time.