https://github.com/ITensor/ITensor
Tip revision: 6d3b537acce3c1617e0e180b7c849a1d233a83dd authored by Matthew Fishman on 09 December 2019, 17:28:52 UTC
Use PWD instead of CURDIR to create this_dir.mk
Use PWD instead of CURDIR to create this_dir.mk
Tip revision: 6d3b537
itensor_impl.h
//
// Copyright 2018 The Simons Foundation, Inc. - All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#ifndef __ITENSOR_ITENSOR_INTERFACE_IMPL_H_
#define __ITENSOR_ITENSOR_INTERFACE_IMPL_H_
#include "itensor/itdata/task_types.h"
#include "itensor/tensor/contract.h"
//
// Template Method Implementations
//
namespace itensor {
namespace detail {
void
allocReal(ITensor& T);
void
allocReal(ITensor& T, IntArray const& inds);
void
allocCplx(ITensor& T);
} //namespace detail
template <typename... Inds>
ITensor::
ITensor(Index const& i1,
Inds const&... inds)
: is_(i1,inds...)
{
IF_USESCALE(scale_ = LogNum(1.);)
}
template <typename... Inds>
ITensor::
ITensor(QN q, Index const& i1,
Inds const&... inds)
{
*this = ITensor(q,IndexSet(i1,inds...));
}
template <class DataType>
ITensor::
ITensor(IndexSet iset,
DataType&& dat,
LogNum const& scale) :
is_(std::move(iset)),
store_(newITData<stdx::decay_t<DataType>>(std::move(dat)))
{
IF_USESCALE(scale_ = scale;)
static_assert(std::is_rvalue_reference<decltype(std::forward<DataType>(dat))>::value,
"Error: cannot pass lvalues to ITensor(...,DataType&& dat,...) constructor");
}
template<typename IV, typename... IVs>
auto ITensor::
eltC(IV const& iv1, IVs&&... ivs) const
-> stdx::if_compiles_return<Cplx,decltype(iv1.index),decltype(iv1.val)>
{
constexpr size_t size = sizeof...(ivs)+1;
auto vals = std::array<IndexVal,size>{{static_cast<IndexVal>(iv1),
static_cast<IndexVal>(ivs)...}};
if(size != size_t(inds().order()))
{
println("---------------------------------------------");
println("Tensor indices = \n",inds(),"\n");
println("---------------------------------------------");
println("Indices provided = ");
for(auto& iv : vals) println(iv.index);
println("---------------------------------------------");
Error(format("Wrong number of IndexVals passed to elt/eltC (expected %d, got %d)",
inds().order(),size));
}
auto inds = IntArray(size);
detail::permute_map(is_,vals,inds,
[](IndexVal const& iv) { return iv.val-1; });
auto z = itensor::doTask(GetElt{is_,inds},store_);
#ifndef USESCALE
return z;
#else
try {
return z*scale().real0();
}
catch(TooBigForReal const& e)
{
println("too big for real in eltC(...), scale = ",scale());
throw e;
}
catch(TooSmallForReal const&)
{
println("warning: too small for real in eltC(...)");
return Cplx(0.,0.);
}
return Cplx(NAN,NAN);
#endif
}
template<typename Int>
auto ITensor::
eltC(std::vector<Int> const& ints) const
-> stdx::enable_if_t<std::is_integral<Int>::value,Cplx>
{
if(!store()) Error("tensor storage unallocated");
auto size = ints.size();
if(size != size_t(inds().order()))
{
println("---------------------------------------------");
println("Tensor indices = \n",inds(),"\n");
println("---------------------------------------------");
print("Indices provided = ");
for(auto i : ints) print(" ",i);
println("\n---------------------------------------------");
Error(format("Wrong number of ints passed to elt/eltC (expected %d, got %d)",
inds().order(),size));
}
auto inds = IntArray(size);
for(auto i : range(size))
inds[i] = ints[i]-1;
auto z = itensor::doTask(GetElt{is_,inds},store_);
#ifndef USESCALE
return z;
#else
try {
return z*scale_.real0();
}
catch(TooBigForReal const& e)
{
println("too big for real in eltC(...), scale = ",scale());
throw e;
}
catch(TooSmallForReal const&)
{
println("warning: too small for real in eltC(...)");
return Cplx(0.,0.);
}
return Cplx(NAN,NAN);
#endif
}
template<typename Int, typename... Ints>
auto ITensor::
eltC(Int iv1, Ints... ivs) const
-> stdx::enable_if_t<std::is_integral<Int>::value
&& stdx::and_<std::is_integral<Ints>...>::value,Cplx>
{
return this->eltC(std::vector<Int>{{iv1,static_cast<int>(ivs)...}});
}
template <typename... IVals>
Real ITensor::
elt(IVals&&... ivs) const
{
if(itensor::isComplex(*this)) Error("Cannot call .elt(...) on an ITensor with complex storage. Please use .eltC(...) instead");
//TODO: make a specialized elt(...) version
auto z = eltC(std::forward<IVals>(ivs)...);
//if(fabs(z.imag()) > 1E-15 && fabs(z.imag()) > 1E-14*fabs(z.real()))
// {
// printfln("element = (%.5E,%.5E)",z.real(),z.imag());
// //Error("tensor is Complex valued, use .eltC(...) method");
// throw ITError("tensor is complex valued, use .eltC(...) method");
// }
return z.real();
}
template <typename... IVals>
Real ITensor::
real(IVals&&... ivs) const
{
return elt(std::forward<IVals>(ivs)...);
}
template <typename... IVals>
Cplx ITensor::
cplx(IVals&&... ivs) const
{
return eltC(std::forward<IVals>(ivs)...);
}
namespace detail {
template<typename IndexValT,
typename Iter,
typename IV>
auto
getVals(Iter it,
Cplx & z,
IV const& iv)
-> stdx::if_compiles_return<void,decltype(iv.index),decltype(iv.val)>
{
static_assert(stdx::false_regardless_of<IndexValT>::value,
"Last argument to .set method must be Real or Cplx scalar");
}
template<typename IndexValT, typename Iter>
void
getVals(Iter it,
Cplx & z,
Cplx const& w)
{
z = w;
}
template<typename IndexValT,
typename Iter,
typename IV,
typename... Rest>
auto
getVals(Iter it,
Cplx & z,
IV const& iv,
Rest&&... rest)
-> stdx::if_compiles_return<void,decltype(iv.index),decltype(iv.val)>
{
*it = static_cast<IndexValT>(iv);
getVals<IndexValT>(++it,z,std::forward<Rest&&>(rest)...);
}
template<typename IndexValT, typename Iter, typename... Rest>
bool
getVals(Iter it,
Cplx & z,
Cplx w,
Rest&&... rest)
{
static_assert(stdx::false_regardless_of<Iter>::value,
"New value passed to .set method must be last argument");
return false;
}
template<typename IntT,
typename Iter,
typename Arg>
auto
getInts(Iter it,
Cplx & z,
Arg const& iv)
-> stdx::enable_if_t<not std::is_convertible<Arg,Cplx>::value,void>
{
static_assert(stdx::false_regardless_of<IntT>::value,
"Last argument to .set method must be Real or Cplx scalar");
}
template<typename IntT,
typename Iter,
typename Arg>
auto
getInts(Iter it,
Cplx & z,
Arg const& w)
-> stdx::enable_if_t<std::is_convertible<Arg,Cplx>::value,void>
{
z = w;
}
template<typename IntT,
typename Iter,
typename Int,
typename... Rest>
auto
getInts(Iter it,
Cplx & z,
Int const& w,
Rest&&... rest)
-> stdx::enable_if_t<not std::is_integral<Int>::value,void>
{
static_assert(stdx::false_regardless_of<Iter>::value,
"New value passed to .set method must be last argument");
return false;
}
template<typename IntT,
typename Iter,
typename Int,
typename... Rest>
auto
getInts(Iter it,
Cplx & z,
Int w,
Rest&&... rest)
-> stdx::enable_if_t<std::is_integral<Int>::value,void>
{
*it = w-1;
getInts<IntT>(++it,z,std::forward<Rest&&>(rest)...);
}
template<typename IndexVals>
void
checkEltFlux(ITensor const& A, IndexVals const& ivs)
{
if(hasQNs(A))
{
QN elt_flux;
auto indsA = inds(A);
for(auto i : range1(order(A)))
{
auto iv = indsA(i)(ivs[i-1].val);
elt_flux += dir(iv)*qn(iv);
}
if(elt_flux != flux(A))
{
println("Trying to set element: ");
for(auto i : range1(order(A)))
println("Index: ", indsA(i), ", Val: ",ivs[i-1].val);
println("Element flux is: ",elt_flux);
println("ITensor flux is: ",flux(A));
Error("In .set, cannot set element with flux different from ITensor flux");
}
}
return;
}
template<typename Ints>
void
checkEltFluxInts(ITensor const& A, Ints const& ints)
{
if(hasQNs(A))
{
QN elt_flux;
auto indsA = inds(A);
for(auto i : range1(order(A)))
{
auto iv = indsA(i)(ints[i-1]+1);
elt_flux += dir(iv)*qn(iv);
}
if(elt_flux != flux(A))
{
println("Trying to set element: ");
for(auto i : range1(order(A)))
println("Index: ", indsA(i), ", Val: ",ints[i-1]+1);
println("Element flux is: ",elt_flux);
println("ITensor flux is: ",flux(A));
Error("In .set, cannot set element with flux different from ITensor flux");
}
}
return;
}
} //namespace detail
template<typename IV, typename... VArgs>
auto ITensor::
set(IV const& iv1, VArgs&&... vargs)
-> stdx::if_compiles_return<void,decltype(iv1.index),decltype(iv1.val)>
{
static constexpr auto size = 1+(sizeof...(vargs)-1);
std::array<IndexVal,size> vals;
Cplx z;
detail::getVals<IndexVal>(vals.begin(),z,iv1,std::forward<VArgs&&>(vargs)...);
if(size != size_t(inds().order()))
{
println("---------------------------------------------");
println("Tensor indices = \n",inds(),"\n");
println("---------------------------------------------");
println("Indices provided = ");
for(auto& iv : vals) println(iv.index);
println("---------------------------------------------");
Error(format("Wrong number of IndexVals passed to set (expected %d, got %d)",
inds().order(),size));
}
auto inds = IntArray(is_.order(),0);
detail::permute_map(is_,vals,inds,
[](IndexVal const& iv) { return iv.val-1; });
//TODO: if !store_ and !is_real, call allocCplx instead
//and move this line after check for is_real
if(!store_) detail::allocReal(*this,inds);
scaleTo(1.);
detail::checkEltFlux(*this,vals);
if(z.imag()==0.0)
{
doTask(SetElt<Real>{z.real(),is_,inds},store_);
}
else
{
doTask(SetElt<Cplx>{z,is_,inds},store_);
}
}
template<typename Int, typename... VArgs>
auto ITensor::
set(Int iv1, VArgs&&... vargs)
-> stdx::enable_if_t<std::is_integral<Int>::value,void>
{
static constexpr auto size = 1+(sizeof...(vargs)-1);
auto ints = IntArray(size,0);
Cplx z;
detail::getInts<Int>(ints.begin(),z,iv1,std::forward<VArgs&&>(vargs)...);
if(size != size_t(inds().order()))
{
println("---------------------------------------------");
println("Tensor indices = \n",inds(),"\n");
println("---------------------------------------------");
print("Indices provided =");
for(auto& i : ints) print(" ",1+i);
println();
println("---------------------------------------------");
Error(format("Wrong number of ints passed to set (expected %d, got %d)",
inds().order(),size));
}
//TODO: if !store_ and !is_real, call allocCplx instead
//and move this line after check for is_real
if(!store_) detail::allocReal(*this,ints);
scaleTo(1.);
detail::checkEltFluxInts(*this,ints);
if(z.imag()==0.0)
{
doTask(SetElt<Real>{z.real(),is_,ints},store_);
}
else
{
doTask(SetElt<Cplx>{z,is_,ints},store_);
}
}
template <typename Func>
ITensor& ITensor::
generate(Func&& f)
{
if(not this->store())
{
using RetType = decltype(f());
if(std::is_same<RetType,Real>::value)
{
detail::allocReal(*this);
}
else if(std::is_same<RetType,Cplx>::value)
{
detail::allocCplx(*this);
}
else
{
Error("generate: generator function must return Real or Cplx scalar value");
}
}
fixBlockDeficient();
scaleTo(1);
doTask(GenerateIT<decltype(f)>{std::forward<Func>(f)},store_);
return *this;
}
template <typename Func>
ITensor& ITensor::
apply(Func&& f)
{
fixBlockDeficient();
scaleTo(1);
doTask(ApplyIT<decltype(f)>{std::forward<Func>(f)},store_);
return *this;
}
template <typename Func>
const ITensor& ITensor::
visit(Func&& f) const
{
doTask(VisitIT<decltype(f)>{std::forward<Func>(f),LogNum{scale().real0()}},store_);
return *this;
}
template <typename... IVals>
ITensor
setElt(IndexVal const& iv1,
IVals const&... rest)
{
const constexpr auto size = 1+sizeof...(rest);
auto ivs = stdx::make_array(iv1,rest...);
//TODO: try directly making inds as iv1.index,(rest.index)...
auto inds = std::array<Index,size>{};
for(size_t j = 0; j < size; ++j) inds[j] = ivs[j].index;
auto D = ITensor{IndexSet(inds)};
D.set(iv1,rest...,1.);
return D;
}
template<typename... VarArgs>
Real
elt(ITensor A,
VarArgs&&... vargs)
{
return A.elt(std::forward<VarArgs>(vargs)...);
}
template<typename... VarArgs>
Cplx
eltC(ITensor A,
VarArgs&&... vargs)
{
return A.eltC(std::forward<VarArgs>(vargs)...);
}
//Apply x = f(x) for each element x of T
//and return the resulting tensor
template<typename F>
ITensor
apply(ITensor T, F&& f)
{
T.apply(std::forward<F>(f));
return T;
}
template<typename T, typename... CtrArgs>
ITensor::storage_ptr
readType(std::istream& s, CtrArgs&&... args)
{
T t(std::forward<CtrArgs>(args)...);
read(s,t);
return newITData<T>(std::move(t));
}
struct Write
{
std::ostream& s;
Write(std::ostream& s_) : s(s_) { }
};
inline const char*
typeNameOf(Write const&) { return "Write"; }
template<typename D>
auto
doTask(Write & W, D const& d)
-> stdx::if_compiles_return<void,decltype(itensor::write(W.s,d))>
{
write(W.s,d);
}
template<typename Container, class>
ITensor
diagITensor(Container const& C,
IndexSet const& is)
{
if( not hasQNs(is) )
{
#ifdef DEBUG
using size_type = decltype(C.size());
//Compute min of all index dimensions
auto mindim = dim(is[0]);
for(const auto& ind : is)
if(dim(ind) < mindim) mindim = dim(ind);
if(C.size() != size_type(mindim))
{
println("mindim = ",mindim);
println("C.size() = ",C.size());
Error("Wrong size of data in diagonal ITensor constructor");
}
#endif
using value_type = typename Container::value_type;
return ITensor(std::move(is),Diag<value_type>(C.begin(),C.end()));
}
else
{
Error("diagITensor constructor not yet implemented for QNs");
return ITensor();
}
}
template<typename V>
TenRef<Range,V>
getBlock(ITensor & T,
IntArray block_ind)
{
if(block_ind.size() != size_t(T.order())) Error("Mismatched number of indices and ITensor order");
if(not T.store())
{
QN q;
for(auto n : range(block_ind))
{
auto& I = inds(T)[n];
q += qn(I,block_ind[n])*dir(I);
}
T = ITensor(inds(T),QDense<V>(inds(T),q));
}
//Interface is 1-indexed; switch to 0-indexed
for(auto& i : block_ind) { i -= 1; }
auto G = GetBlock<V>(inds(T),block_ind);
return doTask(G,T.store());
}
//
// Deprecated
//
template<typename Container, typename... Inds, class>
ITensor
diagTensor(Container const& C,
Index const& i1,
Inds&&... inds)
{
Global::warnDeprecated("diagTensor(Container,Index,...) is deprecated in favor of diagITensor(Container,Index,...)");
return diagITensor(C,IndexSet(i1,std::forward<Inds>(inds)...));
}
template <typename... Inds>
ITensor
randomTensor(Index const& i1, Inds&&... inds)
{
Global::warnDeprecated("randomTensor(Index,...) is deprecated in favor of randomITensor(Index,...)");
return randomITensor(i1,std::forward<Inds>(inds)...);
}
template <typename... Inds>
ITensor
randomTensorC(Index const& i1, Inds&&... inds)
{
Global::warnDeprecated("randomTensorC(Index,...) is deprecated in favor of randomITensorC(Index,...)");
return randomITensorC(i1,std::forward<Inds>(inds)...);
}
template <typename... Inds>
ITensor
randomTensor(QN q, Index const& i1, Inds&&... inds)
{
Global::warnDeprecated("randomTensor(QN,Index,...) is deprecated in favor of randomITensor(QN,Index,...)");
return randomITensor(q,i1,std::forward<Inds>(inds)...);
}
template <typename... Inds>
ITensor
randomTensorC(QN q, Index const& i1, Inds&&... inds)
{
Global::warnDeprecated("randomTensorC(QN,Index,...) is deprecated in favor of randomITensorC(QN,Index,...)");
return randomITensorC(q,i1,std::forward<Inds>(inds)...);
}
template<typename... Inds>
ITensor
reindex(ITensor const& cT,
Index o1, Index n1,
Inds... indxs)
{
Error("Error: reindex(ITensor,Index,Index,...) is deprecated in favor of replaceInds(ITensor,Index,Index,...)");
return ITensor();
}
} // namespace itensor
#endif