From: Peter Munch <peterrmuench@gmail.com>
Date: Sun, 4 Sep 2022 16:21:30 +0000 (+0200)
Subject: Add specializations of PreconditionRelaxation functions
X-Git-Tag: v9.5.0-rc1~990^2
X-Git-Url: https://gitweb.dealii.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=b4f2691240591e1f57a044119487e3b2fca18bcb;p=dealii.git

Add specializations of PreconditionRelaxation functions
---

diff --git a/include/deal.II/lac/precondition.h b/include/deal.II/lac/precondition.h
index caac658f8a..9162d1d863 100644
--- a/include/deal.II/lac/precondition.h
+++ b/include/deal.II/lac/precondition.h
@@ -1046,9 +1046,14 @@ namespace internal
       dst.add(relaxation, tmp);
     }
 
+    // 0) general implementation
     template <typename MatrixType,
               typename PreconditionerType,
-              typename VectorType>
+              typename VectorType,
+              std::enable_if_t<!has_vmult_with_std_functions_for_precondition<
+                                 PreconditionerType,
+                                 VectorType>,
+                               int> * = nullptr>
     void
     step_operations(const MatrixType &        A,
                     const PreconditionerType &preconditioner,
@@ -1079,6 +1084,94 @@ namespace internal
         }
     }
 
+    // 1) specialized implementation with a preconditioner that accepts
+    // ranges
+    template <typename MatrixType,
+              typename PreconditionerType,
+              typename VectorType,
+              std::enable_if_t<has_vmult_with_std_functions_for_precondition<
+                                 PreconditionerType,
+                                 VectorType>,
+                               int> * = nullptr>
+    void
+    step_operations(const MatrixType &        A,
+                    const PreconditionerType &preconditioner,
+                    VectorType &              dst,
+                    const VectorType &        src,
+                    const double              relaxation,
+                    VectorType &              tmp,
+                    VectorType &,
+                    const unsigned int i,
+                    const bool         transposed)
+    {
+      using Number = typename VectorType::value_type;
+
+      if (i == 0)
+        {
+          Number *      dst_ptr = dst.begin();
+          const Number *src_ptr = src.begin();
+
+          preconditioner.vmult(
+            dst,
+            src,
+            [&](const unsigned int start_range, const unsigned int end_range) {
+              // zero 'dst' before running the vmult operation
+              if (end_range > start_range)
+                std::memset(dst.begin() + start_range,
+                            0,
+                            sizeof(Number) * (end_range - start_range));
+            },
+            [&](const unsigned int start_range, const unsigned int end_range) {
+              if (relaxation == 1.0)
+                return; // nothing to do
+
+              const auto src_ptr = src.begin();
+              const auto dst_ptr = dst.begin();
+
+              DEAL_II_OPENMP_SIMD_PRAGMA
+              for (std::size_t i = start_range; i < end_range; ++i)
+                dst_ptr[i] *= relaxation;
+            });
+        }
+      else
+        {
+          tmp.reinit(src, true);
+
+          Assert(transposed == false, ExcNotImplemented());
+
+          A.vmult(tmp, src);
+
+          preconditioner.vmult(
+            dst,
+            tmp,
+            [&](const unsigned int start_range, const unsigned int end_range) {
+              const auto src_ptr = src.begin();
+              const auto tmp_ptr = tmp.begin();
+
+              if (relaxation == 1.0)
+                {
+                  DEAL_II_OPENMP_SIMD_PRAGMA
+                  for (std::size_t i = start_range; i < end_range; ++i)
+                    tmp_ptr[i] = src_ptr[i] - tmp_ptr[i];
+                }
+              else
+                {
+                  // note: we scale the residual here to be able to add into
+                  // the dst vector, which contains the solution from the last
+                  // iteration
+                  DEAL_II_OPENMP_SIMD_PRAGMA
+                  for (std::size_t i = start_range; i < end_range; ++i)
+                    tmp_ptr[i] = relaxation * (src_ptr[i] - tmp_ptr[i]);
+                }
+            },
+            [&](const unsigned int, const unsigned int) {
+              // nothing to do, since scaling by the relaxation factor
+              // has been done in the pre operation
+            });
+        }
+    }
+
+    // 2) specialized implementation for inverse-diagonal preconditioner
     template <typename MatrixType,
               typename VectorType,
               std::enable_if_t<
@@ -1149,6 +1242,8 @@ namespace internal
         }
     }
 
+    // 3) specialized implementation for inverse-diagonal preconditioner and
+    // matrix that accepts ranges
     template <typename MatrixType,
               typename VectorType,
               std::enable_if_t<
diff --git a/tests/lac/precondition_relaxation_01.cc b/tests/lac/precondition_relaxation_01.cc
index 884a4779a2..8c5a83c00a 100644
--- a/tests/lac/precondition_relaxation_01.cc
+++ b/tests/lac/precondition_relaxation_01.cc
@@ -63,6 +63,35 @@ private:
   DiagonalMatrix<VectorType> diagonal_matrix;
 };
 
+template <typename VectorType>
+class MyDiagonalMatrixWithPreAndPost
+{
+public:
+  void
+  vmult(VectorType &      dst,
+        const VectorType &src,
+        const std::function<void(const unsigned int, const unsigned int)>
+          &operation_before_matrix_vector_product,
+        const std::function<void(const unsigned int, const unsigned int)>
+          &operation_after_matrix_vector_product) const
+  {
+    operation_before_matrix_vector_product(0, src.size());
+
+    diagonal_matrix.vmult(dst, src);
+
+    operation_after_matrix_vector_product(0, src.size());
+  }
+
+  VectorType &
+  get_vector()
+  {
+    return diagonal_matrix.get_vector();
+  }
+
+private:
+  DiagonalMatrix<VectorType> diagonal_matrix;
+};
+
 template <typename SparseMatrixType>
 class MySparseMatrix : public Subscriptor
 {
@@ -259,6 +288,25 @@ main()
           results.emplace_back(test(preconditioner, src));
         }
 
+        {
+          // Test PreconditionRelaxation + wrapper around DiagonalMatrix with
+          // pre and post: alternative optimized path is taken
+          using PreconditionerType = MyDiagonalMatrixWithPreAndPost<VectorType>;
+
+          PreconditionRelaxation<MatrixType, PreconditionerType> preconditioner;
+
+          PreconditionRelaxation<MatrixType, PreconditionerType>::AdditionalData
+            ad;
+          ad.relaxation     = relaxation;
+          ad.n_iterations   = n_iterations;
+          ad.preconditioner = std::make_shared<PreconditionerType>();
+          ad.preconditioner->get_vector() = diagonal;
+
+          preconditioner.initialize(system_matrix, ad);
+
+          results.emplace_back(test(preconditioner, src, false));
+        }
+
         if (std::equal(results.begin(),
                        results.end(),
                        results.begin(),