arianajdari / llamba

LLAMBA

LLAMBA is a c++ header only library leveraging multithreading in order to help robotics engineers develop robotics applications.

Running benchmarks

The easiest way to run the benchmarks and check the results is to run provided python script benchmarks.py. In order to run this script, do the following:

 $ cd python_bindings

Before executing the file, check the comments inside benchmarks.py and decide whether you would like to print the results or save them to a file. After this, do the following:

 $ python3 benchmarks.py

Viewing of Benchmarks

If you would like to view the benchmarks without running the tests, head to measurements folder.

Usage of LLAMBA

Since LLAMBA is a header only library, it is enough to include its headers in the project. After including them, create a .cpp file and paste the following code:

#include  <llamba/base/multiply_parallel.hpp>
#include  <llamba/generators/single_generator.hpp>

#define DATA_SIZE 200

int main()
{
	auto matrix_a = llamba::single_generator::generate_input<double>(DATA_SIZE);
	auto matrix_b = llamba::single_generator::generate_input<double>(DATA_SIZE);
	auto result   = llamba::single_generator::generate_input<double>(DATA_SIZE);

	llamba::base::multiply_parallel<double> multiply = llamba::base::multiply_parallel<double>(matrix_a, matrix_b, result, settings_serial);
	
	return 0;
}

Now you can compile the file with g++ compiler.

If you would like to use Benchmark Component, than you can use the code below (multiplication example):

#include  <vector>
#include  <llamba/generators/generator.hpp>
#include  <llamba/base/multiply_parallel.hpp>
#include  <llamba/benchmark/benchmark.hpp>

#define DATA_TYPE double

int main()
{
	auto matrices_a = llamba::generator::generate_inputs<DATA_TYPE>({100, 200});
	auto matrices_b = llamba::generator::generate_inputs<DATA_TYPE>({100, 200});
	
	llamba::settings input_settings_one = llamba::settings(4_THREADS, PARALLELIZATION_STRATEGY::PTHREADS);

	llamba::settings input_settings_two = llamba::settings(PARALLELIZATION_STRATEGY::SERIAL);

	llamba::benchmark<DATA_TYPE> benchmark_multiplication(new llamba::base::multiply_parallel<DATA_TYPE>(), matrices_a, matrices_b, std::vector<llamba::settings>{input_settings_one, input_settings_two});

	benchmark_multiplication.execute_all<TIME_FORMAT::MILLI>();
	return  0;
}