Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Threading in evaluation of tight binding object on a Brillouin zone mesh. #712

Open
wants to merge 2 commits into
base: 2.1.x
Choose a base branch
from
Open

Threading in evaluation of tight binding object on a Brillouin zone mesh. #712

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
0bb235a
[lattice] openmp parallell eval of tight_binding dispersion
HugoStrand Aug 30, 2018
7677a38
[lattice] tight binding renaming fix
HugoStrand Nov 15, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
35 changes: 33 additions & 2 deletions triqs/lattice/tight_binding.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,8 @@
* TRIQS: a Toolbox for Research in Interacting Quantum Systems
*
* Copyright (C) 2011 by M. Ferrero, O. Parcollet
* Copyright (C) 2018, The Simons Foundation
* Author: H. U.R. Strand
*
* TRIQS is free software: you can redistribute it and/or modify it under the
* terms of the GNU General Public License as published by the Free Software
Expand Down Expand Up @@ -32,9 +34,12 @@ namespace triqs {
*/
class tight_binding {

using displs_t = std::vector<std::vector<long>>;
using matrices_t = std::vector<matrix<dcomplex>>;

bravais_lattice bl_;
std::vector<std::vector<long>> displ_vec_;
std::vector<matrix<dcomplex>> overlap_mat_vec_;
displs_t displ_vec_;
matrices_t overlap_mat_vec_;

public:
///
Expand All @@ -52,6 +57,9 @@ namespace triqs {
for (int i = 0; i < n; ++i) f(tb.displ_vec_[i], tb.overlap_mat_vec_[i]);
}

const displs_t & displacements() const { return displ_vec_; }
const matrices_t & matrices() const { return overlap_mat_vec_; }

// a simple function on the domain brillouin_zone
struct fourier_impl {
tight_binding const &tb;
Expand All @@ -65,6 +73,10 @@ namespace triqs {
template <typename K> std14::enable_if_t<!clef::is_clef_expression<K>::value, matrix<dcomplex>> operator()(K const &k) const {
matrix<dcomplex> res(nb, nb);
res() = 0;

/*

// Previous Implementation
foreach (tb, [&](std::vector<long> const &displ, matrix<dcomplex> const &m) {
double dot_prod = 0;
int imax = displ.size();
Expand All @@ -73,6 +85,25 @@ namespace triqs {
res += m * exp(2_j * M_PI * dot_prod);
})
;
*/

#pragma omp declare reduction (+ : triqs::arrays::matrix<dcomplex> : omp_out += omp_in ) \
initializer(omp_priv = triqs::arrays::matrix<dcomplex>(omp_orig.shape()))

#pragma omp parallel for reduction(+:res)
for(int idx = 0; idx < tb.displacements().size(); idx++) {

auto const & displ = tb.displacements()[idx];
auto const & m = tb.matrices()[idx];

double dot_prod = 0;
int imax = displ.size();
for (int i = 0; i < imax; ++i) dot_prod += k(i) * displ[i];
matrix<dcomplex> tmp = m * exp(2_j * M_PI * dot_prod);

res += tmp;
}

return res;
}
};
Expand Down