[Performance] Redirect `AllocWorkspace` to PyTorch's allocator if available (#4199)

CMakeLists.txt

@@ -304,6 +304,7 @@ if(BUILD_TORCH)
       ${CMAKE_COMMAND} -E env
       CMAKE_COMMAND=${CMAKE_CMD}
       CUDA_TOOLKIT_ROOT_DIR=${CUDA_TOOLKIT_ROOT_DIR}
+      USE_CUDA=${USE_CUDA}
       BINDIR=${BINDIR}
       cmd /e:on /c ${BUILD_SCRIPT} ${TORCH_PYTHON_INTERPS}
       DEPENDS ${BUILD_SCRIPT}
@@ -315,6 +316,7 @@ if(BUILD_TORCH)
       ${CMAKE_COMMAND} -E env
       CMAKE_COMMAND=${CMAKE_CMD}
       CUDA_TOOLKIT_ROOT_DIR=${CUDA_TOOLKIT_ROOT_DIR}
+      USE_CUDA=${USE_CUDA}
       BINDIR=${CMAKE_CURRENT_BINARY_DIR}
       bash ${BUILD_SCRIPT} ${TORCH_PYTHON_INTERPS}
       DEPENDS ${BUILD_SCRIPT}

include/dgl/runtime/tensordispatch.h

@@ -69,8 +69,12 @@ class TensorDispatcher {
 
   /*!
    * \brief Allocate an empty tensor.
-   *
    * Used in NDArray::Empty().
+
+   * \param shape The shape
+   * \param dtype The data type
+   * \param ctx The device
+   * \return An empty NDArray.
    */
   inline NDArray Empty(std::vector<int64_t> shape, DLDataType dtype, DLContext ctx) const {
     auto entry = entrypoints_[Op::kEmpty];
@@ -78,6 +82,36 @@ class TensorDispatcher {
     return NDArray::FromDLPack(result);
   }
 
+#ifdef DGL_USE_CUDA
+  /*!
+  * \brief Allocate a piece of GPU memory via
+  * PyTorch's THCCachingAllocator.
+  * Used in CUDADeviceAPI::AllocWorkspace().
+  * 
+  * \note THCCachingAllocator specify the device to allocate on
+  * via cudaGetDevice(). Make sure to call cudaSetDevice()
+  * before invoking this function.
+  *
+  * \param nbytes The size to be allocated.
+  * \return Pointer to the allocated memory.
+  */
+  inline void* AllocWorkspace(size_t nbytes) {
+    auto entry = entrypoints_[Op::kRawAlloc];
+    return FUNCCAST(tensoradapter::RawAlloc, entry)(nbytes);
+  }
+
+  /*!
+  * \brief Free the GPU memory.
+  * Used in CUDADeviceAPI::FreeWorkspace().
+  *
+  * \param ptr Pointer to the memory to be freed.
+  */
+  inline void FreeWorkspace(void* ptr) {
+    auto entry = entrypoints_[Op::kRawDelete];
+    FUNCCAST(tensoradapter::RawDelete, entry)(ptr);
+  }
+#endif  // DGL_USE_CUDA
+
  private:
   /*! \brief ctor */
   TensorDispatcher() = default;
@@ -91,19 +125,33 @@ class TensorDispatcher {
    */
   static constexpr const char *names_[] = {
     "TAempty",
+#ifdef DGL_USE_CUDA
+    "RawAlloc",
+    "RawDelete",
+#endif  // DGL_USE_CUDA
   };
 
   /*! \brief Index of each function to the symbol list */
   class Op {
    public:
     static constexpr int kEmpty = 0;
+#ifdef DGL_USE_CUDA
+    static constexpr int kRawAlloc = 1;
+    static constexpr int kRawDelete = 2;
+#endif  // DGL_USE_CUDA
   };
 
   /*! \brief Number of functions */
   static constexpr int num_entries_ = sizeof(names_) / sizeof(names_[0]);
 
   /*! \brief Entrypoints of each function */
-  void* entrypoints_[num_entries_] = {nullptr};
+  void* entrypoints_[num_entries_] = {
+    nullptr,
+#ifdef DGL_USE_CUDA
+    nullptr,
+    nullptr,
+#endif  // DGL_USE_CUDA
+  };
 
   bool available_ = false;
 #if defined(WIN32) || defined(_WIN32)

src/runtime/cuda/cuda_device_api.cc

@@ -4,7 +4,7 @@
  * \brief GPU specific API
  */
 #include <dgl/runtime/device_api.h>
-
+#include <dgl/runtime/tensordispatch.h>
 #include <dmlc/thread_local.h>
 #include <dgl/runtime/registry.h>
 #include <cuda_runtime.h>
@@ -224,10 +224,20 @@ class CUDADeviceAPI final : public DeviceAPI {
   }
 
   void* AllocWorkspace(DGLContext ctx, size_t size, DGLType type_hint) final {
+    // Redirect to PyTorch's allocator when available.
+    SetDevice(ctx);
+    TensorDispatcher* td = TensorDispatcher::Global();
+    if (td->IsAvailable())
+      return td->AllocWorkspace(size);
+    else
       return CUDAThreadEntry::ThreadLocal()->pool.AllocWorkspace(ctx, size);
   }
 
   void FreeWorkspace(DGLContext ctx, void* data) final {
+    TensorDispatcher* td = TensorDispatcher::Global();
+    if (td->IsAvailable())
+      td->FreeWorkspace(data);
+    else
       CUDAThreadEntry::ThreadLocal()->pool.FreeWorkspace(ctx, data);
   }
 

tensoradapter/include/tensoradapter.h

@@ -18,7 +18,7 @@ namespace tensoradapter {
 extern "C" {
 
 /*!
- * \brief Allocate an empty tensor
+ * \brief Allocate an empty tensor.
  *
  * \param shape The shape
  * \param dtype The data type
@@ -28,6 +28,24 @@ extern "C" {
 DLManagedTensor* TAempty(
     std::vector<int64_t> shape, DLDataType dtype, DLContext ctx);
 
+#ifdef DGL_USE_CUDA
+/*!
+ * \brief Allocate a piece of GPU memory via
+ * PyTorch's THCCachingAllocator.
+ *
+ * \param nbytes The size to be allocated.
+ * \return Pointer to the allocated memory.
+ */
+void* RawAlloc(size_t nbytes);
+
+/*!
+ * \brief Free the GPU memory.
+ *
+ * \param ptr Pointer to the memory to be freed.
+ */
+void RawDelete(void* ptr);
+#endif  // DGL_USE_CUDA
+
 }
 
 };  // namespace tensoradapter

tensoradapter/pytorch/CMakeLists.txt

@@ -17,6 +17,10 @@ list(GET TORCH_PREFIX_VER 0 TORCH_PREFIX)
 list(GET TORCH_PREFIX_VER 1 TORCH_VER)
 message(STATUS "Configuring for PyTorch ${TORCH_VER}")
 
+if(USE_CUDA)
+  add_definitions(-DDGL_USE_CUDA)
+endif()
+
 set(Torch_DIR "${TORCH_PREFIX}/Torch")
 message(STATUS "Setting directory to ${Torch_DIR}")
 find_package(Torch REQUIRED)

tensoradapter/pytorch/build.bat

@@ -11,7 +11,7 @@ IF x%1x == xx GOTO single
 
 FOR %%X IN (%*) DO (
 	DEL /S /Q *
-	"%CMAKE_COMMAND%" -DCMAKE_CONFIGURATION_TYPES=Release -DCUDA_TOOLKIT_ROOT_DIR="%CUDA_TOOLKIT_ROOT_DIR%" -DTORCH_CUDA_ARCH_LIST=%TORCH_CUDA_ARCH_LIST% -DPYTHON_INTERP=%%X .. -G "Visual Studio 16 2019" || EXIT /B 1
+	"%CMAKE_COMMAND%" -DCMAKE_CONFIGURATION_TYPES=Release -DCUDA_TOOLKIT_ROOT_DIR="%CUDA_TOOLKIT_ROOT_DIR%" -DTORCH_CUDA_ARCH_LIST=%TORCH_CUDA_ARCH_LIST% -DUSE_CUDA=%USE_CUDA% -DPYTHON_INTERP=%%X .. -G "Visual Studio 16 2019" || EXIT /B 1
 	msbuild tensoradapter_pytorch.sln /m /nr:false || EXIT /B 1
 	COPY /Y Release\*.dll "%BINDIR%\tensoradapter\pytorch" || EXIT /B 1
 )
@@ -21,7 +21,7 @@ GOTO end
 :single
 
 DEL /S /Q *
-"%CMAKE_COMMAND%" -DCMAKE_CONFIGURATION_TYPES=Release -DCUDA_TOOLKIT_ROOT_DIR="%CUDA_TOOLKIT_ROOT_DIR%" -DTORCH_CUDA_ARCH_LIST=%TORCH_CUDA_ARCH_LIST% .. -G "Visual Studio 16 2019" || EXIT /B 1
+"%CMAKE_COMMAND%" -DCMAKE_CONFIGURATION_TYPES=Release -DCUDA_TOOLKIT_ROOT_DIR="%CUDA_TOOLKIT_ROOT_DIR%" -DTORCH_CUDA_ARCH_LIST=%TORCH_CUDA_ARCH_LIST% -DUSE_CUDA=%USE_CUDA% .. -G "Visual Studio 16 2019" || EXIT /B 1
 msbuild tensoradapter_pytorch.sln /m /nr:false || EXIT /B 1
 COPY /Y Release\*.dll "%BINDIR%\tensoradapter\pytorch" || EXIT /B 1
 

tensoradapter/pytorch/build.sh

@@ -13,7 +13,7 @@ else
 	CPSOURCE=*.so
 fi
 
-CMAKE_FLAGS="-DCUDA_TOOLKIT_ROOT_DIR=$CUDA_TOOLKIT_ROOT_DIR -DTORCH_CUDA_ARCH_LIST=$TORCH_CUDA_ARCH_LIST"
+CMAKE_FLAGS="-DCUDA_TOOLKIT_ROOT_DIR=$CUDA_TOOLKIT_ROOT_DIR -DTORCH_CUDA_ARCH_LIST=$TORCH_CUDA_ARCH_LIST -DUSE_CUDA=$USE_CUDA"
 
 if [ $# -eq 0 ]; then
 	$CMAKE_COMMAND $CMAKE_FLAGS ..

tensoradapter/pytorch/torch.cpp

@@ -7,6 +7,9 @@
 #include <tensoradapter_exports.h>
 #include <torch/torch.h>
 #include <ATen/DLConvertor.h>
+#ifdef DGL_USE_CUDA
+#include <c10/cuda/CUDACachingAllocator.h>
+#endif  // DGL_USE_CUDA
 #include <vector>
 #include <iostream>
 
@@ -47,6 +50,16 @@ TA_EXPORTS DLManagedTensor* TAempty(
   return at::toDLPack(tensor);
 }
 
+#ifdef DGL_USE_CUDA
+TA_EXPORTS void* RawAlloc(size_t nbytes) {
+  return c10::cuda::CUDACachingAllocator::raw_alloc(nbytes);
+}
+
+TA_EXPORTS void RawDelete(void* ptr) {
+  c10::cuda::CUDACachingAllocator::raw_delete(ptr);
+}
+#endif  // DGL_USE_CUDA
+
 };
 
 };  // namespace tensoradapter