add a resource adapter to align on a specified size

include/rmm/detail/aligned.hpp

@@ -31,6 +31,12 @@ namespace detail {
  */
 static constexpr std::size_t RMM_DEFAULT_HOST_ALIGNMENT{alignof(std::max_align_t)};
 
+/**
+ * @brief Default alignment used for CUDA memory allocation.
+ *
+ */
+static constexpr std::size_t CUDA_ALLOCATION_ALIGNMENT{256};
+
 /**
  * @brief Returns whether or not `n` is a power of 2.
  *

include/rmm/mr/device/aligned_resource_adaptor.hpp

@@ -0,0 +1,223 @@
+/*
+ * Copyright (c) 2021, NVIDIA CORPORATION.
+ *
+ * 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.
+ */
+#pragma once
+
+#include <mutex>
+#include <optional>
+#include <unordered_map>
+
+#include <rmm/cuda_stream_view.hpp>
+#include <rmm/detail/aligned.hpp>
+#include <rmm/detail/error.hpp>
+#include <rmm/mr/device/device_memory_resource.hpp>
+
+namespace rmm::mr {
+/**
+ * @brief Resource that adapts `Upstream` memory resource to allocate memory in a specified
+ * alignment size.
+ *
+ * An instance of this resource can be constructed with an existing, upstream resource in order
+ * to satisfy allocation requests. This adaptor wraps allocations and deallocations from Upstream
+ * using the given alignment size.
+ *
+ * By default, any address returned by one of the memory allocation routines from the CUDA driver or
+ * runtime API is always aligned to at least 256 bytes. For some use cases, such as GPUDirect
+ * Storage (GDS), allocations need to be aligned to a larger size (4 KiB for GDS) in order to avoid
+ * additional copies to bounce buffers.
+ *
+ * Since a larger alignment size has some additional overhead, the user can specify a threshold
+ * size. If an allocation's size falls below the threshold, it is aligned to the default size. Only
+ * allocations with a size above the threshold are aligned to the custom alignment size.
+ *
+ * @tparam Upstream Type of the upstream resource used for allocation/deallocation.
+ */
+template <typename Upstream>
+class aligned_resource_adaptor final : public device_memory_resource {
+ public:
+  /**
+   * @brief Construct an aligned resource adaptor using `upstream` to satisfy allocation requests.
+   *
+   * @throws `rmm::logic_error` if `upstream == nullptr`
+   * @throws `rmm::logic_error` if `allocation_alignment` is not a power of 2
+   *
+   * @param upstream The resource used for allocating/deallocating device memory.
+   * @param allocation_alignment The size used for allocation alignment.
+   * @param alignment_threshold Only allocations with a size larger than or equal to this threshold
+   * are aligned.
+   */
+  explicit aligned_resource_adaptor(
+    Upstream* upstream,
+    std::size_t allocation_alignment = rmm::detail::CUDA_ALLOCATION_ALIGNMENT,
+    std::size_t alignment_threshold  = default_alignment_threshold)
+    : upstream_{upstream},
+      allocation_alignment_{allocation_alignment},
+      alignment_threshold_{alignment_threshold}
+  {
+    RMM_EXPECTS(nullptr != upstream, "Unexpected null upstream resource pointer.");
+    RMM_EXPECTS(rmm::detail::is_supported_alignment(allocation_alignment),
+                "Allocation alignment is not a power of 2.");
+  }
+
+  aligned_resource_adaptor()                                = delete;
+  ~aligned_resource_adaptor() override                      = default;
+  aligned_resource_adaptor(aligned_resource_adaptor const&) = delete;
+  aligned_resource_adaptor(aligned_resource_adaptor&&)      = delete;
+  aligned_resource_adaptor& operator=(aligned_resource_adaptor const&) = delete;
+  aligned_resource_adaptor& operator=(aligned_resource_adaptor&&) = delete;
+
+  /**
+   * @brief Get the upstream memory resource.
+   *
+   * @return Upstream* pointer to a memory resource object.
+   */
+  Upstream* get_upstream() const noexcept { return upstream_; }
+
+  /**
+   * @copydoc rmm::mr::device_memory_resource::supports_streams()
+   */
+  [[nodiscard]] bool supports_streams() const noexcept override
+  {
+    return upstream_->supports_streams();
+  }
+
+  /**
+   * @brief Query whether the resource supports the get_mem_info API.
+   *
+   * @return bool true if the upstream resource supports get_mem_info, false otherwise.
+   */
+  [[nodiscard]] bool supports_get_mem_info() const noexcept override
+  {
+    return upstream_->supports_get_mem_info();
+  }
+
+ private:
+  static constexpr std::size_t default_alignment_threshold = 0;
+  using lock_guard                                         = std::lock_guard<std::mutex>;
+
+  /**
+   * @brief Allocates memory of size at least `bytes` using the upstream resource with the specified
+   * alignment.
+   *
+   * @throws `rmm::bad_alloc` if the requested allocation could not be fulfilled
+   * by the upstream resource.
+   *
+   * @param bytes The size, in bytes, of the allocation
+   * @param stream Stream on which to perform the allocation
+   * @return void* Pointer to the newly allocated memory
+   */
+  void* do_allocate(std::size_t bytes, cuda_stream_view stream) override
+  {
+    if (allocation_alignment_ == rmm::detail::CUDA_ALLOCATION_ALIGNMENT ||
+        bytes < alignment_threshold_) {
+      return upstream_->allocate(bytes, stream);
+    } else {
+      auto const size            = upstream_allocation_size(bytes);
+      void* pointer              = upstream_->allocate(size, stream);
+      auto const address         = reinterpret_cast<std::size_t>(pointer);
+      auto const aligned_address = rmm::detail::align_up(address, allocation_alignment_);
+      void* aligned_pointer      = reinterpret_cast<void*>(aligned_address);
+      if (pointer != aligned_pointer) {
+        lock_guard lock(mtx_);
+        pointers_.emplace(aligned_pointer, pointer);
+      }
+      return aligned_pointer;
+    }
+  }
+
+  /**
+   * @brief Free allocation of size `bytes` pointed to to by `p` and log the deallocation.
+   *
+   * @throws Nothing.
+   *
+   * @param p Pointer to be deallocated
+   * @param bytes Size of the allocation
+   * @param stream Stream on which to perform the deallocation
+   */
+  void do_deallocate(void* p, std::size_t bytes, cuda_stream_view stream) override
+  {
+    if (allocation_alignment_ == rmm::detail::CUDA_ALLOCATION_ALIGNMENT ||
+        bytes < alignment_threshold_) {
+      upstream_->deallocate(p, bytes, stream);
+    } else {
+      {
+        lock_guard lock(mtx_);
+        auto const i = pointers_.find(p);
+        if (i != pointers_.end()) {
+          p = i->second;
+          pointers_.erase(i);
+        }
+      }
+      upstream_->deallocate(p, upstream_allocation_size(bytes), stream);
+    }
+  }
+
+  /**
+   * @brief Compare this resource to another.
+   *
+   * @throws Nothing.
+   *
+   * @param other The other resource to compare to
+   * @return true If the two resources are equivalent
+   * @return false If the two resources are not equivalent
+   */
+  [[nodiscard]] bool do_is_equal(device_memory_resource const& other) const noexcept override
+  {
+    if (this == &other)
+      return true;
+    else {
+      auto cast = dynamic_cast<aligned_resource_adaptor<Upstream> const*>(&other);
+      return cast != nullptr && upstream_->is_equal(*cast->get_upstream()) &&
+             allocation_alignment_ == cast->allocation_alignment_ &&
+             alignment_threshold_ == cast->alignment_threshold_;
+    }
+  }
+
+  /**
+   * @brief Get free and available memory from upstream resource.
+   *
+   * The free size may not be fully allocatable because of alignment requirements.
+   *
+   * @throws `rmm::cuda_error` if unable to retrieve memory info.
+   *
+   * @param stream Stream on which to get the mem info.
+   * @return std::pair containing free_size and total_size of memory
+   */
+  [[nodiscard]] std::pair<size_t, size_t> do_get_mem_info(cuda_stream_view stream) const override
+  {
+    return upstream_->get_mem_info(stream);
+  }
+
+  /**
+   * @brief Calculate the allocation size needed from upstream to account for alignments of both the
+   * size and the base pointer.
+   *
+   * @param bytes The requested allocation size.
+   * @return Allocation size needed from upstream to align both the size and the base pointer.
+   */
+  std::size_t upstream_allocation_size(std::size_t bytes) const
+  {
+    auto const aligned_size = detail::align_up(bytes, allocation_alignment_);
+    return aligned_size + allocation_alignment_ - rmm::detail::CUDA_ALLOCATION_ALIGNMENT;
+  }
+
+  Upstream* upstream_;  ///< The upstream resource used for satisfying allocation requests
+  std::unordered_map<void*, void*> pointers_;  ///< Map of aligned pointers to upstream pointers.
+  std::size_t allocation_alignment_;           ///< The size used for allocation alignment
+  std::size_t alignment_threshold_;  ///< The size above which allocations should be aligned
+  mutable std::mutex mtx_;           ///< Mutex for exclusive lock.
+};
+
+}  // namespace rmm::mr

include/rmm/mr/device/binning_memory_resource.hpp

@@ -39,9 +39,6 @@ namespace mr {
 template <typename Upstream>
 class binning_memory_resource final : public device_memory_resource {
  public:
-  // The required alignment of this allocator
-  static constexpr std::size_t allocation_alignment = 256;
-
   /**
    * @brief Construct a new binning memory resource object.
    *
@@ -136,7 +133,8 @@ class binning_memory_resource final : public device_memory_resource {
    */
   void add_bin(std::size_t allocation_size, device_memory_resource* bin_resource = nullptr)
   {
-    allocation_size = rmm::detail::align_up(allocation_size, allocation_alignment);
+    allocation_size =
+      rmm::detail::align_up(allocation_size, rmm::detail::CUDA_ALLOCATION_ALIGNMENT);
 
     if (nullptr != bin_resource)
       resource_bins_.insert({allocation_size, bin_resource});

include/rmm/mr/device/detail/arena.hpp

@@ -132,9 +132,6 @@ class block {
   std::size_t size_{};  ///< Size in bytes.
 };
 
-/// The required allocation alignment.
-constexpr std::size_t allocation_alignment = 256;
-
 /**
  * @brief Align up to the allocation alignment.
  *
@@ -143,7 +140,7 @@ constexpr std::size_t allocation_alignment = 256;
  */
 constexpr std::size_t align_up(std::size_t v) noexcept
 {
-  return rmm::detail::align_up(v, allocation_alignment);
+  return rmm::detail::align_up(v, rmm::detail::CUDA_ALLOCATION_ALIGNMENT);
 }
 
 /**
@@ -154,7 +151,7 @@ constexpr std::size_t align_up(std::size_t v) noexcept
  */
 constexpr std::size_t align_down(std::size_t v) noexcept
 {
-  return rmm::detail::align_down(v, allocation_alignment);
+  return rmm::detail::align_down(v, rmm::detail::CUDA_ALLOCATION_ALIGNMENT);
 }
 
 /**

include/rmm/mr/device/detail/stream_ordered_memory_resource.hpp

@@ -74,9 +74,6 @@ struct crtp {
 template <typename PoolResource, typename FreeListType>
 class stream_ordered_memory_resource : public crtp<PoolResource>, public device_memory_resource {
  public:
-  // TODO use rmm-level def of this.
-  static constexpr size_t allocation_alignment = 256;
-
   ~stream_ordered_memory_resource() { release(); }
 
   stream_ordered_memory_resource()                                      = default;
@@ -211,7 +208,7 @@ class stream_ordered_memory_resource : public crtp<PoolResource>, public device_
 
     auto stream_event = get_event(stream);
 
-    bytes = rmm::detail::align_up(bytes, allocation_alignment);
+    bytes = rmm::detail::align_up(bytes, rmm::detail::CUDA_ALLOCATION_ALIGNMENT);
     RMM_EXPECTS(bytes <= this->underlying().get_maximum_allocation_size(),
                 rmm::bad_alloc,
                 "Maximum allocation size exceeded");
@@ -241,7 +238,7 @@ class stream_ordered_memory_resource : public crtp<PoolResource>, public device_
     auto stream_event = get_event(stream);
     RMM_LOG_TRACE("[D][stream {:p}][{}B][{:p}]", fmt::ptr(stream_event.stream), bytes, p);
 
-    bytes        = rmm::detail::align_up(bytes, allocation_alignment);
+    bytes        = rmm::detail::align_up(bytes, rmm::detail::CUDA_ALLOCATION_ALIGNMENT);
     auto const b = this->underlying().free_block(p, bytes);
 
     // TODO: cudaEventRecord has significant overhead on deallocations. For the non-PTDS case

include/rmm/mr/device/fixed_size_memory_resource.hpp

@@ -54,8 +54,6 @@ class fixed_size_memory_resource
   // This is the number of blocks that the pool starts out with, and also the number of
   // blocks by which the pool grows when all of its current blocks are allocated
   static constexpr std::size_t default_blocks_to_preallocate = 128;
-  // The required alignment of this allocator
-  static constexpr std::size_t allocation_alignment = 256;
 
   /**
    * @brief Construct a new `fixed_size_memory_resource` that allocates memory from
@@ -73,7 +71,7 @@ class fixed_size_memory_resource
     std::size_t block_size            = default_block_size,
     std::size_t blocks_to_preallocate = default_blocks_to_preallocate)
     : upstream_mr_{upstream_mr},
-      block_size_{rmm::detail::align_up(block_size, allocation_alignment)},
+      block_size_{rmm::detail::align_up(block_size, rmm::detail::CUDA_ALLOCATION_ALIGNMENT)},
       upstream_chunk_size_{block_size * blocks_to_preallocate}
   {
     // allocate initial blocks and insert into free list
@@ -201,7 +199,8 @@ class fixed_size_memory_resource
   {
     // Deallocating a fixed-size block just inserts it in the free list, which is
     // handled by the parent class
-    RMM_LOGGING_ASSERT(rmm::detail::align_up(size, allocation_alignment) <= block_size_);
+    RMM_LOGGING_ASSERT(rmm::detail::align_up(size, rmm::detail::CUDA_ALLOCATION_ALIGNMENT) <=
+                       block_size_);
     return block_type{p};
   }
 

include/rmm/mr/device/limiting_resource_adaptor.hpp

@@ -15,6 +15,7 @@
  */
 #pragma once
 
+#include <rmm/detail/aligned.hpp>
 #include <rmm/detail/error.hpp>
 #include <rmm/mr/device/device_memory_resource.hpp>
 
@@ -44,9 +45,10 @@ class limiting_resource_adaptor final : public device_memory_resource {
    * @param upstream The resource used for allocating/deallocating device memory
    * @param allocation_limit Maximum memory allowed for this allocator.
    */
-  limiting_resource_adaptor(Upstream* upstream,
-                            std::size_t allocation_limit,
-                            std::size_t allocation_alignment = 256)
+  limiting_resource_adaptor(
+    Upstream* upstream,
+    std::size_t allocation_limit,
+    std::size_t allocation_alignment = rmm::detail::CUDA_ALLOCATION_ALIGNMENT)
     : allocation_limit_{allocation_limit},
       allocated_bytes_(0),
       allocation_alignment_(allocation_alignment),