3
0
mirror of https://github.com/triqs/dft_tools synced 2024-11-01 11:43:47 +01:00
dft_tools/triqs/arrays/vector.hpp
Olivier Parcollet 47cb8a03f7 [arrays] Important changes in implementation.
- Simplify group_indices
  - Only for C ordered, remove complex compile time.
  - Could be generalized to non C ordered, but no need.
- Fix slice for custom orders.
- Generalize the group_indices for the custom order.
- Add c_ordered_transposed_view (useful ?)
- Improve slice, special for ellipsis (quicker).
- Simplify TraversalOrder
- Assignement. Specialize one case for speed.
- use FORCEINLINE in foreach, according to speed test for clang
- add one speed test
- Modify iterators for better speed.
- along the lines decided for the foreach
- update doc.
2014-10-18 21:20:17 +02:00

354 lines
13 KiB
C++

/*******************************************************************************
*
* TRIQS: a Toolbox for Research in Interacting Quantum Systems
*
* Copyright (C) 2011-2014 by O. Parcollet
*
* 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
* Foundation, either version 3 of the License, or (at your option) any later
* version.
*
* TRIQS is distributed in the hope that it will be useful, but WITHOUT ANY
* WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
* FOR A PARTICULAR PURPOSE. See the GNU General Public License for more
* details.
*
* You should have received a copy of the GNU General Public License along with
* TRIQS. If not, see <http://www.gnu.org/licenses/>.
*
******************************************************************************/
#pragma once
#include "indexmaps/cuboid/map.hpp"
#include "indexmaps/cuboid/slice.hpp"
#include "impl/indexmap_storage_pair.hpp"
#include "impl/assignment.hpp"
#include <algorithm>
namespace triqs {
namespace arrays {
template <typename ValueType, bool Borrowed = false, bool IsConst = false> class vector_view;
template <typename ValueType> class vector;
// ---------------------- vector_view --------------------------------
#define IMPL_TYPE \
indexmap_storage_pair<indexmaps::cuboid::map<1>, storages::shared_block<ValueType, Borrowed>, void, IsConst, true, \
Tag::vector_view>
/** */
template <typename ValueType, bool Borrowed, bool IsConst>
class vector_view : Tag::vector_view, TRIQS_CONCEPT_TAG_NAME(MutableVector), public IMPL_TYPE {
public:
using regular_type = vector<ValueType>;
using view_type = vector_view<ValueType, false>;
using const_view_type = vector_view<ValueType, false, true>;
using weak_view_type = vector_view<ValueType, true>;
using indexmap_type = typename IMPL_TYPE::indexmap_type;
using storage_type = typename IMPL_TYPE::storage_type;
/// Build from an IndexMap and a storage
template <typename S> vector_view(indexmaps::cuboid::map<1> const& Ind, S const& Mem) : IMPL_TYPE(Ind, Mem) {}
/// Build from anything that has an indexmap and a storage compatible with this class
template <typename ISP> vector_view(const ISP& X) : IMPL_TYPE(X.indexmap(), X.storage()) {}
#ifdef TRIQS_WITH_PYTHON_SUPPORT
/// Build from a numpy.array : throws if X is not a numpy.array
explicit vector_view(PyObject* X) : IMPL_TYPE(X, false, "vector_view ") {}
#endif
/// Copy construction
vector_view(vector_view const& X) : IMPL_TYPE(X.indexmap(), X.storage()) {}
vector_view() = delete;
// Move
vector_view(vector_view&& X) { this->swap_me(X); }
/// Swap
friend void swap(vector_view& A, vector_view& B) { A.swap_me(B); }
/// Rebind the view
void rebind(vector_view const& X) {
this->indexmap_ = X.indexmap_;
this->storage_ = X.storage_;
}
// rebind the other view, iif this is const, and the other is not.
template <bool C = IsConst> ENABLE_IFC(C) rebind(vector_view<ValueType, Borrowed, !IsConst> const& X) {
this->indexmap_ = X.indexmap_;
this->storage_ = X.storage_;
}
/** Assignment. The size of the array MUST match exactly. */
template <typename RHS> vector_view& operator=(const RHS& X) {
triqs_arrays_assign_delegation(*this, X);
return *this;
}
vector_view& operator=(vector_view const& X) {
triqs_arrays_assign_delegation(*this, X);
return *this;
} // cf array_view class comment
// Move assignment not defined : will use the copy = since view must copy data
size_t size() const { return this->shape()[0]; }
std::ptrdiff_t stride() const { return this->indexmap().strides()[0]; }
TRIQS_DEFINE_COMPOUND_OPERATORS(vector_view);
// to make interface similar to std::vector : forward [] to ()
template <typename Arg> auto operator[](Arg&& arg) const DECL_AND_RETURN((*this)(std::forward<Arg>(arg)));
template <typename Arg> auto operator[](Arg&& arg)DECL_AND_RETURN((*this)(std::forward<Arg>(arg)));
};
#undef IMPL_TYPE
template <class V, int R, bool Borrowed, bool IsConst>
struct ISPViewType<V, R, void, Tag::vector_view, Borrowed, IsConst> {
using type = vector_view<V, Borrowed, IsConst>;
};
template <typename ValueType, bool Borrowed = false> using vector_const_view = vector_view<ValueType, Borrowed, true>;
// ---------------------- vector--------------------------------
#define IMPL_TYPE \
indexmap_storage_pair<indexmaps::cuboid::map<1>, storages::shared_block<ValueType>, void, false, false, Tag::vector_view>
template <typename ValueType> class vector : Tag::vector, TRIQS_CONCEPT_TAG_NAME(MutableVector), public IMPL_TYPE {
public:
using value_type = typename IMPL_TYPE::value_type;
using storage_type = typename IMPL_TYPE::storage_type;
using indexmap_type = typename IMPL_TYPE::indexmap_type;
using regular_type = vector<ValueType>;
using view_type = vector_view<ValueType>;
using const_view_type = vector_view<ValueType, false, true>;
using weak_view_type = vector_view<ValueType, true>;
/// Empty vector.
vector() {}
// Move
explicit vector(vector&& X) { this->swap_me(X); }
///
vector(size_t dim) : IMPL_TYPE(indexmap_type(mini_vector<size_t, 1>(dim))) {}
/// to mimic std vector
template <typename Arg> vector(size_t dim, Arg&& arg) : IMPL_TYPE(indexmap_type(mini_vector<size_t, 1>(dim))) {
(*this)() = std::forward<Arg>(arg);
}
/** Makes a true (deep) copy of the data. */
vector(const vector& X) : IMPL_TYPE(X.indexmap(), X.storage().clone()) {}
/**
* Build a new vector from X.domain() and fill it with by evaluating X. X can be :
* - another type of array, array_view, matrix,.... (any <IndexMap, Storage> pair)
* - a expression : e.g. array<int, IndexOrder::C<1> > A( B+ 2*C);
* - ml : useless directly, since there only one ml, but used in generic code it maintains the same constructor as array,
* matrix
*/
template <typename T>
// vector(const T & X, std14::enable_if_t< ImmutableCuboidArray<T> ::value> *dummy =0):
vector(const T& X, TYPE_ENABLE_IF(memory_layout<1>, ImmutableCuboidArray<T>) ml = memory_layout<1>{})
: IMPL_TYPE(indexmap_type(X.domain(), ml)) {
triqs_arrays_assign_delegation(*this, X);
}
#ifdef TRIQS_WITH_PYTHON_SUPPORT
/// Build from a numpy.array X (or any object from which numpy can make a numpy.array). Makes a copy.
explicit vector(PyObject* X) : IMPL_TYPE(X, true, "vector ") {}
#endif
// build from a init_list
template <typename T>
vector(std::initializer_list<T> const& l)
: IMPL_TYPE(indexmap_type(mini_vector<size_t, 1>(l.size()), memory_layout<1>{})) {
size_t i = 0;
for (auto const& x : l) (*this)(i++) = x;
}
/**
* Resizes the vector. NB : all references to the storage is invalidated.
* Does not initialize the vector by default: to resize and init, do resize(IND).init()
*/
vector& resize(size_t L) {
IMPL_TYPE::resize(typename IMPL_TYPE::domain_type(mini_vector<size_t, 1>(L)));
return *this;
}
/**
* Resizes the vector. NB : all references to the storage is invalidated.
* Does not initialize the vector by default: to resize and init, do resize(IND).init()
*/
vector& resize(const indexmaps::cuboid::domain_t<IMPL_TYPE::rank>& l) {
IMPL_TYPE::resize(l);
return *this;
}
/// Assignement resizes the vector. All references to the storage are therefore invalidated.
vector& operator=(const vector& X) {
IMPL_TYPE::resize_and_clone_data(X);
return *this;
}
/**
* Assignement resizes the vector. All references to the storage are therefore invalidated.
* NB : to avoid that, do make_view(A) = X instead of A = X
*/
template <typename RHS> vector& operator=(const RHS& X) {
static_assert(ImmutableCuboidArray<RHS>::value, "Assignment : RHS not supported");
IMPL_TYPE::resize(X.domain());
triqs_arrays_assign_delegation(*this, X);
return *this;
}
/// Move assignment
vector& operator=(vector&& X) {
this->swap_me(X);
return *this;
}
friend void swap(vector& A, vector& B) { A.swap_me(B); }
size_t size() const { return this->shape()[0]; }
std::ptrdiff_t stride() const { return this->indexmap().strides()[0]; }
TRIQS_DEFINE_COMPOUND_OPERATORS(vector);
// to make interface similar to std::vector : forward [] to ()
template <typename Arg> auto operator[](Arg&& arg) const DECL_AND_RETURN((*this)(std::forward<Arg>(arg)));
template <typename Arg> auto operator[](Arg&& arg)DECL_AND_RETURN((*this)(std::forward<Arg>(arg)));
}; // vector class
}
} // namespace triqs::arrays
#undef IMPL_TYPE
#include "./blas_lapack/scal.hpp"
#include "./blas_lapack/copy.hpp"
#include "./blas_lapack/swap.hpp"
#include "./blas_lapack/axpy.hpp"
namespace triqs {
namespace arrays {
// norm2 squared
template <typename V> std14::enable_if_t<ImmutableVector<V>::value, typename V::value_type> norm2_sqr(V const& a) {
int dim = a.size();
auto r = typename V::value_type{};
for (int i = 0; i < dim; ++i) r += a(i) * a(i);
return r;
}
// norm2
template <typename V> std14::enable_if_t<ImmutableVector<V>::value, typename V::value_type> norm2(V const& a) {
using std::sqrt;
return sqrt(norm2(a));
}
// lexicographical comparison operators
template <typename V1, typename V2>
std14::enable_if_t<ImmutableVector<V1>::value&& ImmutableVector<V2>::value, bool> operator<(V1 const& a, V2 const& b) {
return std::lexicographical_compare(a.begin(), a.end(), b.begin(), b.end());
}
template <typename V1, typename V2>
std14::enable_if_t<ImmutableVector<V1>::value&& ImmutableVector<V2>::value, bool> operator>(V1 const& a, V2 const& b) {
return (b < a);
}
template <typename RHS, typename T>
std14::enable_if_t<is_vector_or_view<RHS>::value> triqs_arrays_assign_delegation(vector<T>& lhs, RHS const& rhs) {
blas::copy(rhs, lhs);
}
template <typename RHS, typename T>
std14::enable_if_t<is_vector_or_view<RHS>::value> triqs_arrays_compound_assign_delegation(vector<T>& lhs, RHS const& rhs,
char_<'A'>) {
T a = 1.0;
blas::axpy(a, rhs, lhs);
}
template <typename RHS, typename T>
std14::enable_if_t<is_vector_or_view<RHS>::value> triqs_arrays_compound_assign_delegation(vector<T>& lhs, RHS const& rhs,
char_<'S'>) {
T a = -1.0;
blas::axpy(a, rhs, lhs);
}
template <typename RHS, typename T>
std14::enable_if_t<is_scalar_for<RHS, vector<T>>::value> triqs_arrays_compound_assign_delegation(vector<T>& lhs, RHS const& rhs,
char_<'M'>) {
T a = rhs;
blas::scal(a, lhs);
}
template <typename RHS, typename T>
std14::enable_if_t<is_scalar_for<RHS, vector<T>>::value> triqs_arrays_compound_assign_delegation(vector<T>& lhs, RHS const& rhs,
char_<'D'>) {
T a = 1 / rhs;
blas::scal(a, lhs);
}
template <typename RHS, typename T>
std14::enable_if_t<is_vector_or_view<RHS>::value> triqs_arrays_assign_delegation(vector_view<T>& lhs, RHS const& rhs) {
blas::copy(rhs, lhs);
}
template <typename RHS, typename T>
std14::enable_if_t<is_vector_or_view<RHS>::value> triqs_arrays_compound_assign_delegation(vector_view<T>& lhs, RHS const& rhs,
char_<'A'>) {
T a = 1.0;
blas::axpy(a, rhs, lhs);
}
template <typename RHS, typename T>
std14::enable_if_t<is_vector_or_view<RHS>::value> triqs_arrays_compound_assign_delegation(vector_view<T>& lhs, RHS const& rhs,
char_<'S'>) {
T a = -1.0;
blas::axpy(a, rhs, lhs);
}
template <typename RHS, typename T>
std14::enable_if_t<is_scalar_for<RHS, vector_view<T>>::value>
triqs_arrays_compound_assign_delegation(vector_view<T>& lhs, RHS const& rhs, char_<'M'>) {
T a = rhs;
blas::scal(a, lhs);
}
template <typename RHS, typename T>
std14::enable_if_t<is_scalar_for<RHS, vector_view<T>>::value>
triqs_arrays_compound_assign_delegation(vector_view<T>& lhs, RHS const& rhs, char_<'D'>) {
T a = 1 / rhs;
blas::scal(a, lhs);
}
template <typename T, bool Borrowed>
void triqs_arrays_assign_delegation(vector_view<T, Borrowed>& av, std::vector<T> const& vec) {
std::size_t size = vec.size();
for (std::size_t n = 0; n < size; ++n) av(n) = vec[n];
}
// swapping 2 vector
template <typename V, bool B1, bool B2, bool B3, bool B4> void deep_swap(vector_view<V, B1, B2> x, vector_view<V, B3, B4> y) {
blas::swap(x, y);
}
template <typename V> void deep_swap(vector<V>& x, vector<V>& y) { blas::swap(x, y); }
}
}
// The std::swap is WRONG for a view because of the copy/move semantics of view.
// Use swap instead (the correct one, found by ADL).
namespace std {
template <typename V, bool B1, bool B2, bool C1, bool C2>
void swap(triqs::arrays::vector_view<V, B1, C1>& a, triqs::arrays::vector_view<V, B2, C2>& b) = delete;
}
#include "./expression_template/vector_algebra.hpp"