User Guide¶

To use the library, simply include/import cytnx.

In Python, using import:

import cytnx

In C++, using the include header:

#include "cytnx.hpp";

Note

In C++, there is a namespace cytnx.

Aliases in Python modules and C++ namespaces can be used equivalently, for example if we want to alias cytnx as cy,

In Python:

import cytnx as cy

This is equivalent in C++ to:

In C++:

#include "cytnx.hpp";
namespace cy=cytnx;

Now we are ready to start using cytnx!

Overview¶

This is a comprehensive guide to explain the usage and features of Cytnx. It covers fundamental concepts, basic structures, and high-level objects. While this guide can be read from start to end for a comprehensive understanding, most users will be interested in specific topics to start with:

Start reading¶

Beginners are advised to start with the following chapters:

Conventions explains the basic concepts of the library such as naming conventions and the unified C++/Python framework. This chapter is therefore essential for all users.

Tensor notation introduces the abstract objects that Cytnx builds on, tensors. A graphical notation is introduced for convenience.

UniTensor concerns the object that users will most frequently use: a ‘UniTensor’, which contains not only the tensor elements but also further metadata and methods to characterize and manipulate the tensor. Typical tensor network implementations will be based on this object.

Tensor decomposition explains how a tensor can be split into a product of several tensors by a singular value decompositon, eigenvalue decomposition, or QR decomposition.

Iterative solver concerns efficient iterative methods, for example for obtaining the smallest eigenvalue.

Linear algebra contains a list of further linear algebra functions that can be applied to a tensor.

Advanced topics¶

The remaining sections cover specific topics or the backend of Cytnx:

Device is the object that defines where the data is stored, and is important for parallel and distributed computing on GPUs and CPUs.

Tensor contains the data together with basic information about the indices, such as the shape of the tensor. This object is similar to numpy.arrays or torch.Tensor.

Storage concerns the basic memory access.

Scalar is an elementary object that refers to a number, which can have different data types.

Common APIs lists methods that several objects in Cytnx have in common.

User Guide¶

Overview¶

Start reading¶

Advanced topics¶

Contents¶