Line data    Source code

       1             : //* This file is part of the MOOSE framework
       2             : //* https://mooseframework.inl.gov
       3             : //*
       4             : //* All rights reserved, see COPYRIGHT for full restrictions
       5             : //* https://github.com/idaholab/moose/blob/master/COPYRIGHT
       6             : //*
       7             : //* Licensed under LGPL 2.1, please see LICENSE for details
       8             : //* https://www.gnu.org/licenses/lgpl-2.1.html
       9             : 
      10             : #pragma once
      11             : 
      12             : #include "MOOSEToNEML2.h"
      13             : #include "ElementUserObject.h"
      14             : 
      15             : /**
      16             :  * @brief Generic gatherer for collecting "batched" MOOSE data for NEML2
      17             :  *
      18             :  * It is generic in the sense that it can be used for most MOOSE data types that take the form of
      19             :  * MooseArray<T>.
      20             :  *
      21             :  * It is not so generic in the sense that the collected data is always a std::vector of
      22             :  * MooseArray<T>, where the vector size is generally the number of elements this ElementUserObject
      23             :  * operates on, and the MooseArray<T> size is generally the number of quadrature points in each
      24             :  * element.
      25             :  *
      26             :  * @tparam T Type of the underlying MOOSE data, e.g., Real, SymmetricRankTwoTensor, etc.
      27             :  */
      28             : template <typename T>
      29             : class MOOSEToNEML2Batched : public MOOSEToNEML2, public ElementUserObject
      30             : {
      31             : public:
      32             :   static InputParameters validParams();
      33             : 
      34             :   MOOSEToNEML2Batched(const InputParameters & params);
      35             : 
      36             : #ifndef NEML2_ENABLED
      37           0 :   void initialize() override {}
      38           0 :   void execute() override {}
      39           0 :   void finalize() override {}
      40           0 :   void threadJoin(const UserObject &) override {}
      41             : #else
      42             :   void initialize() override;
      43             :   void execute() override;
      44        2951 :   void finalize() override {}
      45             :   void threadJoin(const UserObject &) override;
      46             : 
      47             :   neml2::Tensor gatheredData() const override;
      48             : 
      49             :   // The number of batches
      50             :   std::size_t size() const { return _buffer.size(); }
      51             : 
      52             : protected:
      53             :   /// MOOSE data for the current element
      54             :   virtual const MooseArray<T> & elemMOOSEData() const = 0;
      55             : 
      56             :   /// Intermediate data buffer, filled during the element loop
      57             :   std::vector<T> _buffer;
      58             : #endif
      59             : };
      60             : 
      61             : template <typename T>
      62             : InputParameters
      63      142725 : MOOSEToNEML2Batched<T>::validParams()
      64             : {
      65      142725 :   auto params = MOOSEToNEML2::validParams();
      66      142725 :   params += ElementUserObject::validParams();
      67             : 
      68             :   // Since we use the NEML2 model to evaluate the residual AND the Jacobian at the same time, we
      69             :   // want to execute this user object only at execute_on = LINEAR (i.e. during residual evaluation).
      70             :   // The NONLINEAR exec flag below is for computing Jacobian during automatic scaling.
      71      142725 :   ExecFlagEnum execute_options = MooseUtils::getDefaultExecFlagEnum();
      72      570900 :   execute_options = {EXEC_INITIAL, EXEC_LINEAR, EXEC_NONLINEAR};
      73      142725 :   params.set<ExecFlagEnum>("execute_on") = execute_options;
      74             : 
      75      285450 :   return params;
      76      285450 : }
      77             : 
      78             : template <typename T>
      79          38 : MOOSEToNEML2Batched<T>::MOOSEToNEML2Batched(const InputParameters & params)
      80          38 :   : MOOSEToNEML2(params), ElementUserObject(params)
      81             : {
      82          38 : }
      83             : 
      84             : #ifdef NEML2_ENABLED
      85             : template <typename T>
      86             : void
      87        3303 : MOOSEToNEML2Batched<T>::initialize()
      88             : {
      89        3303 :   _buffer.clear();
      90        3303 : }
      91             : 
      92             : template <typename T>
      93             : void
      94       29530 : MOOSEToNEML2Batched<T>::execute()
      95             : {
      96       29530 :   const auto & elem_data = this->elemMOOSEData();
      97      105390 :   for (auto i : index_range(elem_data))
      98       75860 :     _buffer.push_back(elem_data[i]);
      99       29530 : }
     100             : 
     101             : template <typename T>
     102             : void
     103         352 : MOOSEToNEML2Batched<T>::threadJoin(const UserObject & uo)
     104             : {
     105             :   // append vectors
     106         352 :   const auto & m2n = static_cast<const MOOSEToNEML2Batched<T> &>(uo);
     107         352 :   _buffer.insert(_buffer.end(), m2n._buffer.begin(), m2n._buffer.end());
     108         352 : }
     109             : 
     110             : template <typename T>
     111             : neml2::Tensor
     112        1731 : MOOSEToNEML2Batched<T>::gatheredData() const
     113             : {
     114        1731 :   return NEML2Utils::fromBlob(_buffer);
     115             : }
     116             : #endif