transpose() [1/4]

template<Rank rank, typename Scalar , TransposeMethod method = TransposeMethod::Default, typename Indices = std::make_index_sequence<(rank < 2 ? 0 : (rank - 2))>>

void syten::transpose	(	ConstSpan< Scalar > const	inp,
		MutSpan< Scalar >	out,
		[[maybe_unused] ] ConstSpan< Index >	in_perm,
		[[maybe_unused] ] ConstSpan< Index >	in_dim,
		[[maybe_unused] ] Conj	do_conj = Conj::n(),
		[[maybe_unused] ] bool	do_checks = true
	)

Entry point for transposition of a continuous array into another array.

Template Parameters

rank	maximal number of indices
Scalar	underlying scalar type
method	transposition method, default is heuristic based on tensor size
Indices	helper argument to generate an index sequence (see syten::transpose_impl).

Parameters

inp	source array, row-major
out	output array, row-major
in_perm	permutation array, perm[i] = j says that the old leg j should be placed at position i+1 (all 1-indexed), at most size rank
in_dim	input dimensions array, at most size rank
do_conj	if Conj::y(), conjugate every element
do_checks	if true, remove 1-dimensional indices and combine runs

References syten::SpanImpl::MutSpan< T >::begin(), syten::Cuda::conj_copy(), std::copy(), if_constexpr, rank(), syten::SpanImpl::MutSpan< T >::size(), SYTEN_ASSERT_MSG, and syten::YesNo< struct _prod_conj >::y().

Referenced by syten::Delinearise::cols_only(), syten::Eigen::eigenGen(), syten::Eigen::eigenSym(), exponentialTensorSym(), syten::Krylov::SolverPolicies::Exponential::get_coeffs(), syten::IdentityDenseTensorImpl::merge(), syten::DenseProduct::one_times_many(), pseudo_inverse_svd(), syten::Delinearise::rows(), syten::MPS::DMRG::LBOWorker::singleUpdate(), syten::Krylov::ClassicImpl::ExponentialSolver< Solver >::solve(), syten::DenseSolve::solve(), syten::SVD::svd(), and syten::MPS::TDVP::LBOWorker::sweep_single().

Here is the call graph for this function:

Here is the caller graph for this function: