Add prototype for shared allocator

Author: hmenke

c++/mpi/allocator.hpp

@@ -0,0 +1,54 @@
+// Copyright (c) 2023 Simons Foundation
+//
+// 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.txt
+//
+// 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.
+//
+// Authors: Philipp Dumitrescu, Olivier Parcollet, Nils Wentzell
+
+#pragma once
+#include "mpi.hpp"
+#include <algorithm>
+#include <memory>
+
+namespace mpi {
+
+  template <class T>
+  class shared_allocator {
+    shared_communicator shm = communicator{}.split_shared();
+    std::shared_ptr<std::vector<shared_window<T>>> blocks = std::make_shared<std::vector<shared_window<T>>>();
+  public:
+    using value_type = T;
+
+    shared_allocator() = default;
+    explicit shared_allocator(shared_communicator const &shm) noexcept : shm{shm} {}
+
+    [[nodiscard]] shared_communicator get_communicator() { return shm; }
+
+    [[nodiscard]] auto get_window(T *p) {
+      return std::find_if(blocks->begin(), blocks->end(), [p](shared_window<T> const &win) {
+        return win.base(0) == p;
+      });
+    }
+
+    [[nodiscard]] T* allocate(std::size_t n) {
+      shared_window<T> &win = blocks->emplace_back(shm, shm.rank() == 0 ? n : 0);
+      return win.base(0);
+    }
+
+    void deallocate(T *p, std::size_t) {
+      blocks->erase(std::remove_if(blocks->begin(), blocks->end(), [p](shared_window<T> const &win) {
+        return win.base(0) == p;
+      }), blocks->end());
+    }
+  };
+
+} // namespace mpi

c++/mpi/mpi.hpp

@@ -23,6 +23,7 @@
 #include <complex>
 #include <type_traits>
 #include <algorithm>
+#include <utility>
 #include <unistd.h>
 
 /// Library namespace
@@ -364,9 +365,15 @@ namespace mpi {
   public:
     window() = default;
     window(window const&) = delete;
-    window(window &&) = delete;
+    window(window &&other) noexcept : win{std::exchange(other.win, MPI_WIN_NULL)} {}
     window& operator=(window const&) = delete;
-    window& operator=(window &&) = delete;
+    window& operator=(window &&rhs) noexcept {
+      if (this != std::addressof(rhs)) {
+        this->free();
+        this->win = std::exchange(rhs.win, MPI_WIN_NULL);
+      }
+      return *this;
+    }
 
     /// Create a window over an existing local memory buffer
     explicit window(communicator &c, BaseType *base, MPI_Aint size = 0) noexcept {
@@ -379,15 +386,17 @@ namespace mpi {
       MPI_Win_allocate(size * sizeof(BaseType), alignof(BaseType), MPI_INFO_NULL, c.get(), &baseptr, &win);
     }
 
-    ~window() {
+    ~window() { free(); }
+
+    explicit operator MPI_Win() const noexcept { return win; };
+    explicit operator MPI_Win*() noexcept { return &win; };
+
+    void free() noexcept {
       if (win != MPI_WIN_NULL) {
         MPI_Win_free(&win);
       }
     }
 
-    explicit operator MPI_Win() const noexcept { return win; };
-    explicit operator MPI_Win*() noexcept { return &win; };
-
     /// Synchronization routine in active target RMA. It opens and closes an access epoch.
     void fence(int assert = 0) const noexcept {
       MPI_Win_fence(assert, win);
@@ -474,8 +483,10 @@ namespace mpi {
   template <class BaseType>
   class shared_window : public window<BaseType> {
   public:
+    shared_window() = default;
+
     /// Create a window and allocate memory for a shared memory buffer
-    shared_window(shared_communicator& c, MPI_Aint size) noexcept {
+    explicit shared_window(shared_communicator& c, MPI_Aint size) noexcept {
       void* baseptr = nullptr;
       MPI_Win_allocate_shared(size * sizeof(BaseType), alignof(BaseType), MPI_INFO_NULL, c.get(), &baseptr, &(this->win));
     }

test/c++/CMakeLists.txt

@@ -10,7 +10,7 @@ file(GLOB_RECURSE all_tests RELATIVE ${CMAKE_CURRENT_SOURCE_DIR} *.cpp)
 # List of all no mpi tests
 file(GLOB_RECURSE nompi_tests RELATIVE ${CMAKE_CURRENT_SOURCE_DIR} *.cpp)
 # remove custom mpi test, as this one explicitly uses MPI
-list(REMOVE_ITEM nompi_tests mpi_custom.cpp mpi_monitor.cpp mpi_window.cpp)
+list(REMOVE_ITEM nompi_tests mpi_custom.cpp mpi_monitor.cpp mpi_window.cpp mpi_allocator.cpp)
 
 # ========= OpenMP Dependency ==========
 

test/c++/mpi_allocator.cpp

@@ -0,0 +1,45 @@
+// Copyright (c) 2023 Simons Foundation
+//
+// 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.txt
+//
+// 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.
+//
+// Authors: Philipp Dumitrescu, Olivier Parcollet, Nils Wentzell
+
+#include <mpi/mpi.hpp>
+#include <mpi/allocator.hpp>
+#include <mpi/vector.hpp>
+#include <gtest/gtest.h>
+#include <numeric>
+
+TEST(MPI_Allocator, SharedAllocator) {
+    mpi::shared_allocator<int> alloc;
+    int *p = alloc.allocate(1);
+    alloc.deallocate(p, 1);
+}
+
+TEST(MPI_Allocator, SharedAllocatorVector) {
+    std::vector<int, mpi::shared_allocator<int>> v(128);
+    auto shm = v.get_allocator().get_communicator();
+
+    // Fill the vector in parallel
+    v.get_allocator().get_window(v.data())->fence();
+    auto slice = itertools::chunk_range(0, v.size(), shm.size(), shm.rank());
+    for (auto i = slice.first; i < slice.second; ++i) {
+        v.at(i) = i;
+    }
+    v.get_allocator().get_window(v.data())->fence();
+
+    int const sum = std::accumulate(v.begin(), v.end(), int{0});
+    EXPECT_EQ(sum, v.size() * (v.size() - 1) / 2);
+}
+
+MPI_TEST_MAIN;