compilation fixes

- assertions weren't compiled before
- MPI was missing from refactor compile_commands.json

Author: upsj

core/distributed/vector.cpp

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: 2017 - 2025 The Ginkgo authors
+// SPDX-FileCopyrightText: 2017 - 2026 The Ginkgo authors
 //
 // SPDX-License-Identifier: BSD-3-Clause
 
@@ -204,7 +204,8 @@ void Vector<ValueType>::read_distributed_impl(
     auto rank = this->get_communicator().rank();
     local_.fill(zero<ValueType>());
     exec->run(vector::make_build_local(
-        data, make_temporary_clone(exec, partition).get(), rank, &local_));
+        data, make_temporary_clone(exec, partition).get(), rank,
+        local_.get_device_view()));
 }
 
 
@@ -345,8 +346,9 @@ Vector<ValueType>::compute_absolute() const
         absolute_type::create(exec, this->get_communicator(), this->get_size(),
                               this->get_local_vector()->get_size());
 
-    exec->run(vector::make_outplace_absolute_dense(this->get_local_vector(),
-                                                   &result->local_));
+    exec->run(vector::make_outplace_absolute_dense(
+        this->get_local_vector()->get_const_device_view(),
+        result->local_.get_device_view()));
 
     return result;
 }
@@ -551,7 +553,7 @@ void Vector<ValueType>::compute_norm2(ptr_param<LinOp> result,
     const auto comm = this->get_communicator();
     auto dense_res = make_temporary_clone(exec, as<NormVector>(result));
     this->compute_squared_norm2(dense_res.get(), tmp);
-    exec->run(vector::make_compute_sqrt(dense_res.get()));
+    exec->run(vector::make_compute_sqrt(dense_res->get_device_view()));
 }
 
 
@@ -607,8 +609,9 @@ void Vector<ValueType>::compute_squared_norm2(ptr_param<LinOp> result,
     auto exec = this->get_executor();
     const auto comm = this->get_communicator();
     auto dense_res = make_temporary_clone(exec, as<NormVector>(result));
-    exec->run(vector::make_compute_squared_norm2(this->get_local_vector(),
-                                                 dense_res.get(), tmp));
+    exec->run(vector::make_compute_squared_norm2(
+        this->get_local_vector()->get_const_device_view(),
+        dense_res->get_device_view(), tmp));
     exec->synchronize();
     auto norm_sum_op = gko::experimental::mpi::sum<remove_complex<ValueType>>();
     if (mpi::requires_host_buffer(exec, comm)) {

omp/matrix/coo_kernels.cpp

@@ -68,7 +68,7 @@ void spmv2_blocked(std::shared_ptr<const OmpExecutor> exec,
                    matrix::view::dense<const ValueType> b,
                    matrix::view::dense<ValueType> c, ValueType scale)
 {
-    GKO_ASSERT(b->get_size()[1] > block_size);
+    GKO_ASSERT(b.size[1] > block_size);
     const auto coo_val = a->get_const_values();
     const auto coo_col = a->get_const_col_idxs();
     const auto coo_row = a->get_const_row_idxs();
@@ -198,7 +198,7 @@ void spmv2_small_rhs(std::shared_ptr<const OmpExecutor> exec,
                      matrix::view::dense<const ValueType> b,
                      matrix::view::dense<ValueType> c, ValueType scale)
 {
-    GKO_ASSERT(b->get_size()[1] == num_rhs);
+    GKO_ASSERT(b.size[1] == num_rhs);
     const auto coo_val = a->get_const_values();
     const auto coo_col = a->get_const_col_idxs();
     const auto coo_row = a->get_const_row_idxs();

omp/matrix/ell_kernels.cpp

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: 2017 - 2025 The Ginkgo authors
+// SPDX-FileCopyrightText: 2017 - 2026 The Ginkgo authors
 //
 // SPDX-License-Identifier: BSD-3-Clause
 
@@ -35,7 +35,7 @@ void spmv_small_rhs(std::shared_ptr<const OmpExecutor> exec,
                     matrix::view::dense<const InputValueType> b,
                     matrix::view::dense<OutputValueType> c, OutFn out)
 {
-    GKO_ASSERT(b->get_size()[1] == num_rhs);
+    GKO_ASSERT(b.size[1] == num_rhs);
     using arithmetic_type =
         highest_precision<InputValueType, OutputValueType, MatrixValueType>;
     using a_accessor =
@@ -85,7 +85,7 @@ void spmv_blocked(std::shared_ptr<const OmpExecutor> exec,
                   matrix::view::dense<const InputValueType> b,
                   matrix::view::dense<OutputValueType> c, OutFn out)
 {
-    GKO_ASSERT(b->get_size()[1] > block_size);
+    GKO_ASSERT(b.size[1] > block_size);
     using arithmetic_type =
         highest_precision<InputValueType, OutputValueType, MatrixValueType>;
     using a_accessor =

omp/matrix/sellp_kernels.cpp

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: 2017 - 2025 The Ginkgo authors
+// SPDX-FileCopyrightText: 2017 - 2026 The Ginkgo authors
 //
 // SPDX-License-Identifier: BSD-3-Clause
 
@@ -28,7 +28,7 @@ void spmv_small_rhs(std::shared_ptr<const OmpExecutor> exec,
                     matrix::view::dense<const ValueType> b,
                     matrix::view::dense<ValueType> c, OutFn out)
 {
-    GKO_ASSERT(b->get_size()[1] == num_rhs);
+    GKO_ASSERT(b.size[1] == num_rhs);
     auto slice_lengths = a->get_const_slice_lengths();
     auto slice_sets = a->get_const_slice_sets();
     auto slice_size = a->get_slice_size();