2
0
mirror of https://github.com/boostorg/ublas.git synced 2026-01-19 16:52:13 +00:00
Files
ublas/examples/tensor/instantiate_tensor.cpp
Cem Bassoy 231ba5f730 refactor(core): simplify and eliminate auxiliary tensor types (#115)
Auxiliary functions for extents and strides were using different
functions. Additionally, many tags were used to distinguish between
different tensor types. This patch simplifies interfaces of different
core functions and unifies functions that can process different types of
extent and stride types.
2021-09-09 11:34:14 +02:00

122 lines
3.4 KiB
C++

//
// Copyright (c) 2018, Cem Bassoy, cem.bassoy@gmail.com
// Copyright (c) 2019, Amit Singh, amitsingh19975@gmail.com
//
// Distributed under the Boost Software License, Version 1.0. (See
// accompanying file LICENSE_1_0.txt or copy at
// http://www.boost.org/LICENSE_1_0.txt)
//
// The authors gratefully acknowledge the support of Google
//
#include <boost/numeric/ublas/tensor.hpp>
#include <vector>
#include <iostream>
#include <type_traits>
void instantiate_tensor_dynamic()
{
namespace ublas = boost::numeric::ublas;
using value = float;
using layout = ublas::layout::first_order; // storage format
using tensor = boost::numeric::ublas::tensor_dynamic<value,layout>;
constexpr auto ones = ublas::ones<value,layout>{};
try {
// tensor is resizable has dynamic dimensions
// elements are stored contiguously in memory using the 1st-format (column-major)
tensor t1 = ones(3,4,2);
std::cout << "t1 = " << t1 << std::endl;
tensor t2 = 2 * ones(3,4,2);
std::cout << "t2 = " << t2 << std::endl;
tensor t3 = 3*t2 + t1;
std::cout << "t3 = " << t3 << std::endl;
} catch (const std::exception& e) {
std::cerr << "Cought exception " << e.what();
std::cerr << "in the instantiate_tensor_dynamic function of instantiate-tensor." << std::endl;
throw;
}
}
void instantiate_tensor_dynamics_with_static_order()
{
namespace ublas = boost::numeric::ublas;
using value = float;
using layout = boost::numeric::ublas::layout::first_order; // storage format
using tensor = boost::numeric::ublas::tensor_static_rank<value,3u, layout>;
constexpr auto ones = ublas::ones_static_rank<value,layout>{};
try {
// tensor type has static order and dynamic dimensions
// elements are stored contiguously in memory using the 1st-format (column-major)
auto t1 = ones(3,4,2);
std::cout << "t1 = " << t1 << std::endl;
tensor t2 = 2*ones(3,4,2);
std::cout << "t2 = " << t2 << std::endl;
tensor t3 = 3*t2 + t1;
std::cout << "t3 = " << t3 << std::endl;
} catch (const std::exception& e) {
std::cerr << "Cought exception " << e.what();
std::cerr << "in the instantiate_tensor_dynamic function of instantiate-tensor." << std::endl;
throw;
}
}
void instantiate_tensor_static()
{
namespace ublas = boost::numeric::ublas;
using value = float;
using layout = ublas::layout::first_order; // storage format
using shape = ublas::extents<3,4,2>;
using tensor = ublas::tensor_static<value,shape,layout>;
constexpr auto ones = ublas::ones_static<value,layout>{};
try {
// tensor type has static order and static dimensions
// elements are stored contiguously in memory using the 1st-format (column-major)
auto t1 = tensor{};
std::cout << "t1 = " << t1 << std::endl;
tensor t2 = 2 * ones(shape{});
std::cout << "t2 = " << t2 << std::endl;
tensor t3 = 3*t2 + t1;
std::cout << "t3 = " << t3 << std::endl;
} catch (const std::exception& e) {
std::cerr << "Cought exception " << e.what();
std::cerr << "in the instantiate_tensor_dynamic function of instantiate-tensor." << std::endl;
throw;
}
}
int main()
{
try{
instantiate_tensor_dynamic();
instantiate_tensor_dynamics_with_static_order();
instantiate_tensor_static();
} catch (const std::exception& e) {
std::cerr << "Cought exception " << e.what();
std::cerr << "in the main function of instantiate-tensor." << std::endl;
}
}