yomichi / mptensor

Parallel Library for Tensor Network Methods

mptensor v0.2

"mptensor" is parallel C++ libarary for tensor calculations. It provides similar interfaces as Numpy and Scipy in Python.

Requirements

• MPI Library
• ScaLAPACK
• (For RSVD) C++11 std::random, Boost C++ library, or dSFMT

How to Use

1. Compile mptensor
   • Modify Makefile.option as your environment and then make.
   • or use cmake
2. Include src/mptensor.hpp in your codes.
3. Complie your applications with src/libmptensor.a .

Documents

The HTML documents are available in here.

By make doc, HTML and LaTeX documents are generated in doc/doxygen.

Examples

#include <mptensor.hpp>
using namespace mptensor;
typedef Tensor<scalapack::Matrix,double> ptensor;
ptensor A(Shape(3,4,5));

Example codes of TRG and HOTRG for the 2D Ising model are in examples/Ising_2D.

License

GNU Lesser General Public License v3.0 (see LICENSE)

Links

• Tensordot: Code generator for tensor contraction
• cuscalapack: pdgemm and pzgemm with cuBLAS