Welcome to the libMD v0.0 documentation!
libMD
For information about a particular module, see its page. You will find the most up to date information about a module in its header file.
About
libMD is a library for common operations in machine learning methods applied to MD. libMD is presented as a SYCL library, written in C++20.
Important
Compiling and running any code containing libMD requires a working SYCL compiler. We recommend using OpenSYCL <https://github.com/OpenSYCL/OpenSYCL>
If this is the first time you encounter libMD and want a quick start guide check the QuickStart. The FAQ is also a good source of knowledge.
Look in Compiling if some Makefile gives you troubles. libMD is a header only framework and it is mostly self contained in terms of dependencies.
A quick example:
A brief example of how a code using libMD typically looks like.
auto q = md::get_default_queue();
int num_particles = 100;
float box_size = 128.f;
float cutoff = 1.5f;
auto positions = sycl::buffer<vec3<float>>(num_particles);
{
sycl::host_accessor positions_acc{positions, sycl::write_only, sycl::no_init};
//Positions are placed randomly inside a cubic box
std::mt19937 gen(0xBADA55D00D);
std::uniform_real_distribution<float> dis(0, 1);
for(int i = 0; i < num_particles; i++){
positions_acc[i] = vec3<float>(dis(gen), dis(gen), dis(gen))*box_size;
}
}
auto [neighbors, neighbor_indices, max_num_neighbors] =
md::computeNeighbors(q, positions, cutoff);
//The format of the neighbor list is as follows:
//neighbors[i]: The number of neighbors of particle i
//neighbor_indices[i * max_num_neighbors + j]: The index of the jth neighbor of particle i
//max_num_neighbors: The maximum number of neighbors of any particle
sycl::host_accessor neighbors_acc{neighbors, sycl::read_only};
std::cout<<"Number of neighbors of particle 0: "<<neighbors_acc[0]<<std::endl;