[SE] GlobalDeviceMemory owns its handle

[jhen]

Summary

Final step in getting GlobalDeviceMemory to own its handle.

• Make GlobalDeviceMemory movable, but no longer copyable.
• Make Device::freeDeviceMemory function private and make GlobalDeviceMemoryBase a friend of Device so GlobalDeviceMemoryBase can free its memory in its destructor.
• Make GlobalDeviceMemory constructor private and make Device a friend so it can construct GlobalDeviceMemory.
• Remove SharedDeviceMemoryBase class because it is never used.
• Remove explicit memory freeing from example code.

This change just consumes any errors generated during device memory freeing.
The real error handling will be added in a future patch.

[jlebar]

streamexecutor/include/streamexecutor/DeviceMemory.h

126

Is this last paragraph still fully correct?

136

Do the handle and byte count still need to be optional? I guess you can change it later.

167

These don't need to be private if you don't want.

235

Why are we getting rid of SharedDeviceMemoryBase?

I don't have a strong opinion one way or another, but it's not clear to me why this is an improvement.

The main advantage of SharedDeviceMemoryBase, I guess, is that you can accept one as a const ref parameter to a function without a template.

[jhen]

• Respond to jlebar's comments

streamexecutor/include/streamexecutor/DeviceMemory.h

126

No, it was not. I just removed it completely because I think it's not needed anyway.

136

I changed it now. They don't need to be optional.

167

I'm never sure where is the best place to put these because the visibility doesn't seem to matter. I moved them into the public section to prevent creating a new section for them. Does that seem cleaner?

235

I think its an improvement because we ain't gonna need it. If it becomes useful later on, we can add it back in, but it doesn't seem like it will.

[jlebar]

streamexecutor/include/streamexecutor/DeviceMemory.h

167

Right, I don't think visibility matters because visibility doesn't factor into C++ overloading rules.

I think it's sort of cleaner to put them up top like you did here because the fact that it's not copyable is in reality part of the "public interface".

[jlebar]

I think its an improvement because we ain't gonna need it. If it becomes useful later on, we can add it back in, but it doesn't seem like it will.

That sounds great to me. :)

streamexecutor/examples/Example.cpp

int main() {
   // Wait for the stream to complete.
   se::dieIfError(Stream.blockHostUntilDone());
 
   // Process output data in HostX.
   std::vector<float> ExpectedX = {4, 47, 90, 133};
   for (size_t I = 0; I < ArraySize; ++I) {
     assert(HostX[I] == ExpectedX[I]);
   }
-
-  // Free device memory.
-  se::dieIfError(Device->freeDeviceMemory(X));
-  se::dieIfError(Device->freeDeviceMemory(Y));
   /// [Example saxpy host main]
 }

streamexecutor/include/streamexecutor/Device.h

public:
 
   /// Allocates an array of ElementCount entries of type T in device memory.
   template <typename T>
   Expected<GlobalDeviceMemory<T>> allocateDeviceMemory(size_t ElementCount) {
     Expected<void *> MaybeMemory =
         PDevice->allocateDeviceMemory(ElementCount * sizeof(T));
     if (!MaybeMemory)
       return MaybeMemory.takeError();
-    return GlobalDeviceMemory<T>::makeFromElementCount(*MaybeMemory,
-                                                       ElementCount);
-  }
-
-  /// Frees memory previously allocated with allocateDeviceMemory.
-  template <typename T> Error freeDeviceMemory(GlobalDeviceMemory<T> Memory) {
-    return PDevice->freeDeviceMemory(Memory.getHandle());
+    return GlobalDeviceMemory<T>(this, *MaybeMemory, ElementCount);
   }
 
   /// Allocates an array of ElementCount entries of type T in host memory.
   ///
   /// Host memory allocated by this function can be used for asynchronous memory
   /// copies on streams. See Stream::thenCopyD2H and Stream::thenCopyH2D.
   template <typename T> Expected<T *> allocateHostMemory(size_t ElementCount) {
     Expected<void *> MaybeMemory =
public:
   Error synchronousCopyD2D(const GlobalDeviceMemory<T> &Src,
                            GlobalDeviceMemory<T> &Dst) {
     return synchronousCopyD2D(Src.asSlice(), Dst.asSlice());
   }
 
   ///@} End host-synchronous device memory copying functions
 
 private:
+  // Only a GlobalDeviceMemoryBase may free device memory.
+  friend GlobalDeviceMemoryBase;
+  Error freeDeviceMemory(const GlobalDeviceMemoryBase &Memory) {
+    return PDevice->freeDeviceMemory(Memory.getHandle());
+  }
+
   PlatformDevice *PDevice;
 };
 
 } // namespace streamexecutor
 
 #endif // STREAMEXECUTOR_DEVICE_H

streamexecutor/include/streamexecutor/DeviceMemory.h

 //===-- DeviceMemory.h - Types representing device memory -------*- C++ -*-===//
 //
 //                     The LLVM Compiler Infrastructure
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 ///
 /// \file
 /// This file defines types that represent device memory buffers. Two memory
 /// spaces are represented here: global and shared. Host code can have a handle
 /// to device global memory, and that handle can be used to copy data to and
 /// from the device. Host code cannot have a handle to device shared memory
 /// because that memory only exists during the execution of a kernel.
 ///
-/// GlobalDeviceMemoryBase is similar to a pair consisting of a void* pointer
-/// and a byte count to tell how much memory is pointed to by that void*.
-///
-/// GlobalDeviceMemory<T> is a subclass of GlobalDeviceMemoryBase which keeps
-/// track of the type of element to be stored in the device memory. It is
-/// similar to a pair of a T* pointer and an element count to tell how many
-/// elements of type T fit in the memory pointed to by that T*.
-///
-/// SharedDeviceMemoryBase is just the size in bytes of a shared memory buffer.
-///
-/// SharedDeviceMemory<T> is a subclass of SharedDeviceMemoryBase which knows
-/// how many elements of type T it can hold.
-///
-/// These classes are useful for keeping track of which memory space a buffer
-/// lives in, and the typed subclasses are useful for type-checking.
-///
-/// The typed subclass will be used by user code, and the untyped base classes
-/// will be used for type-unsafe operations inside of StreamExecutor.
+/// GlobalDeviceMemory<T> is a handle to an array of elements of type T in
+/// global device memory. It is similar to a pair of a std::unique_ptr<T> and an
+/// element count to tell how many elements of type T fit in the memory pointed
+/// to by that T*.
+///
+/// SharedDeviceMemory<T> is just the size in elements of an array of elements
+/// of type T in device shared memory. No resources are actually attached to
+/// this class, it is just like a memo to the device to allocate space in shared
+/// memory.
 ///
 //===----------------------------------------------------------------------===//
 
 #ifndef STREAMEXECUTOR_DEVICEMEMORY_H
 #define STREAMEXECUTOR_DEVICEMEMORY_H
 
 #include <cassert>
 #include <cstddef>
 
-namespace streamexecutor {
-
-/// Wrapper around a generic global device memory allocation.
-///
-/// This class represents a buffer of untyped bytes in the global memory space
-/// of a device. See GlobalDeviceMemory<T> for the corresponding type that
-/// includes type information for the elements in its buffer.
-///
-/// This is effectively a pair consisting of an opaque handle and a buffer size
-/// in bytes. The opaque handle is a platform-dependent handle to the actual
-/// memory that is allocated on the device.
-///
-/// In some cases, such as in the CUDA platform, the opaque handle may actually
-/// be a pointer in the virtual address space and it may be valid to perform
-/// arithmetic on it to obtain other device pointers, but this is not the case
-/// in general.
-///
-/// For example, in the OpenCL platform, the handle is a pointer to a _cl_mem
-/// handle object which really is completely opaque to the user.
-///
-/// The only fully platform-generic operations on handles are using them to
-/// create new GlobalDeviceMemoryBase objects, and comparing them to each other
-/// for equality.
-class GlobalDeviceMemoryBase {
-public:
-  /// Creates a GlobalDeviceMemoryBase from an optional handle and an optional
-  /// byte count.
-  explicit GlobalDeviceMemoryBase(const void *Handle = nullptr,
-                                  size_t ByteCount = 0)
-      : Handle(Handle), ByteCount(ByteCount) {}
-
-  /// Copyable like a pointer.
-  GlobalDeviceMemoryBase(const GlobalDeviceMemoryBase &) = default;
-
-  /// Copy-assignable like a pointer.
-  GlobalDeviceMemoryBase &operator=(const GlobalDeviceMemoryBase &) = default;
-
-  /// Returns the size, in bytes, for the backing memory.
-  size_t getByteCount() const { return ByteCount; }
+#include "streamexecutor/Utils/Error.h"
 
-  /// Gets the internal handle.
-  ///
-  /// Warning: note that the pointer returned is not necessarily directly to
-  /// device virtual address space, but is platform-dependent.
-  const void *getHandle() const { return Handle; }
+namespace streamexecutor {
 
-private:
-  const void *Handle; // Platform-dependent value representing allocated memory.
-  size_t ByteCount;   // Size in bytes of this allocation.
-};
+class Device;
 
 template <typename ElemT> class GlobalDeviceMemory;
 
 /// Reference to a slice of device memory.
 ///
 /// Contains a base memory handle, an element count offset into that base
 /// memory, and an element count for the size of the slice.
 template <typename ElemT> class GlobalDeviceMemorySlice {
 public:
   /// Intentionally implicit so GlobalDeviceMemory<T> can be passed to functions
   /// expecting GlobalDeviceMemorySlice<T> arguments.
   GlobalDeviceMemorySlice(const GlobalDeviceMemory<ElemT> &Memory)
       : BaseMemory(Memory), ElementOffset(0),
         ElementCount(Memory.getElementCount()) {}
 
   GlobalDeviceMemorySlice(const GlobalDeviceMemory<ElemT> &BaseMemory,
                           size_t ElementOffset, size_t ElementCount)
       : BaseMemory(BaseMemory), ElementOffset(ElementOffset),
         ElementCount(ElementCount) {
     assert(ElementOffset + ElementCount <= BaseMemory.getElementCount() &&
            "slicing past the end of a GlobalDeviceMemory buffer");
   }
 
   /// Gets the GlobalDeviceMemory backing this slice.
-  GlobalDeviceMemory<ElemT> getBaseMemory() const { return BaseMemory; }
+  const GlobalDeviceMemory<ElemT> &getBaseMemory() const { return BaseMemory; }
 
   /// Gets the offset of this slice from the base memory.
   ///
   /// The offset is measured in elements, not bytes.
   size_t getElementOffset() const { return ElementOffset; }
 
   /// Gets the number of elements in this slice.
   size_t getElementCount() const { return ElementCount; }
GlobalDeviceMemorySlice<ElemT> slice(size_t DropCount,
                                        size_t TakeCount) const {
     assert(DropCount + TakeCount <= ElementCount &&
            "sub-slice operation overruns slice bounds");
     return GlobalDeviceMemorySlice<ElemT>(BaseMemory, ElementOffset + DropCount,
                                           TakeCount);
   }
 
 private:
-  GlobalDeviceMemory<ElemT> BaseMemory;
+  const GlobalDeviceMemory<ElemT> &BaseMemory;
   size_t ElementOffset;
   size_t ElementCount;
 };
 
+/// Wrapper around a generic global device memory allocation.
+///
+/// This class represents a buffer of untyped bytes in the global memory space
+/// of a device. See GlobalDeviceMemory<T> for the corresponding type that
+/// includes type information for the elements in its buffer.
+///
+/// This is effectively a pair consisting of an opaque handle and a buffer size
+/// in bytes. The opaque handle is a platform-dependent handle to the actual
+/// memory that is allocated on the device.
+///
+/// In some cases, such as in the CUDA platform, the opaque handle may actually
+/// be a pointer in the virtual address space and it may be valid to perform
+/// arithmetic on it to obtain other device pointers, but this is not the case
+/// in general.
+///
+/// For example, in the OpenCL platform, the handle is a pointer to a _cl_mem
+/// handle object which really is completely opaque to the user.
+class GlobalDeviceMemoryBase {
+public:
+  /// Returns an opaque handle to the underlying memory.
+  const void *getHandle() const { return Handle; }
+
+  // Cannot copy because the handle must be owned by a single object.
+  GlobalDeviceMemoryBase(const GlobalDeviceMemoryBase &) = delete;
+  GlobalDeviceMemoryBase &operator=(const GlobalDeviceMemoryBase &) = delete;
+
+protected:
+  /// Creates a GlobalDeviceMemoryBase from a handle and a byte count.
+  GlobalDeviceMemoryBase(Device *D, const void *Handle, size_t ByteCount)
+      : TheDevice(D), Handle(Handle), ByteCount(ByteCount) {}
+
+  /// Transfer ownership of the underlying handle.
+  GlobalDeviceMemoryBase(GlobalDeviceMemoryBase &&Other)
+      : TheDevice(Other.TheDevice), Handle(Other.Handle),
+        ByteCount(Other.ByteCount) {
+    Other.TheDevice = nullptr;
+    Other.Handle = nullptr;
+    Other.ByteCount = 0;
+  }
+
+  GlobalDeviceMemoryBase &operator=(GlobalDeviceMemoryBase &&Other) {
+    TheDevice = Other.TheDevice;
+    Handle = Other.Handle;
+    ByteCount = Other.ByteCount;
+    Other.TheDevice = nullptr;
+    Other.Handle = nullptr;
+    Other.ByteCount = 0;
+    return *this;
+  }
+
+  ~GlobalDeviceMemoryBase();
+
+  Device *TheDevice;  // Pointer to the device on which this memory lives.
+  const void *Handle; // Platform-dependent value representing allocated memory.
+  size_t ByteCount;   // Size in bytes of this allocation.
+};
+
 /// Typed wrapper around the "void *"-like GlobalDeviceMemoryBase class.
 ///
 /// For example, GlobalDeviceMemory<int> is a simple wrapper around
 /// GlobalDeviceMemoryBase that represents a buffer of integers stored in global
 /// device memory.
 template <typename ElemT>
 class GlobalDeviceMemory : public GlobalDeviceMemoryBase {
 public:
-  /// Creates a typed area of GlobalDeviceMemory with a given opaque handle and
-  /// the given element count.
-  static GlobalDeviceMemory<ElemT> makeFromElementCount(const void *Handle,
-                                                        size_t ElementCount) {
-    return GlobalDeviceMemory<ElemT>(Handle, ElementCount);
-  }
-
-  /// Creates a typed device memory region from an untyped device memory region.
-  ///
-  /// This effectively amounts to a cast from a void* to an ElemT*, but it also
-  /// manages the difference in the size measurements when
-  /// GlobalDeviceMemoryBase is measured in bytes and GlobalDeviceMemory is
-  /// measured in elements.
-  explicit GlobalDeviceMemory(const GlobalDeviceMemoryBase &Other)
-      : GlobalDeviceMemoryBase(Other.getHandle(), Other.getByteCount()) {}
-
-  /// Copyable like a pointer.
-  GlobalDeviceMemory(const GlobalDeviceMemory &) = default;
-
-  /// Copy-assignable like a pointer.
-  GlobalDeviceMemory &operator=(const GlobalDeviceMemory &) = default;
+  GlobalDeviceMemory(GlobalDeviceMemory &&Other) = default;
+  GlobalDeviceMemory &operator=(GlobalDeviceMemory &&Other) = default;
 
   /// Returns the number of elements of type ElemT that constitute this
   /// allocation.
-  size_t getElementCount() const { return getByteCount() / sizeof(ElemT); }
+  size_t getElementCount() const { return ByteCount / sizeof(ElemT); }
 
   /// Converts this memory object into a slice.
   GlobalDeviceMemorySlice<ElemT> asSlice() const {
     return GlobalDeviceMemorySlice<ElemT>(*this);
   }
 
 private:
-  /// Constructs a GlobalDeviceMemory instance from an opaque handle and an
-  /// element count.
-  ///
-  /// This constructor is not public because there is a potential for confusion
-  /// between the size of the buffer in bytes and the size of the buffer in
-  /// elements.
-  ///
-  /// The static method makeFromElementCount is provided for users of this class
-  /// because its name makes the meaning of the size parameter clear.
-  GlobalDeviceMemory(const void *Handle, size_t ElementCount)
-      : GlobalDeviceMemoryBase(Handle, ElementCount * sizeof(ElemT)) {}
+  GlobalDeviceMemory(const GlobalDeviceMemory &) = delete;
+  GlobalDeviceMemory &operator=(const GlobalDeviceMemory &) = delete;
+
+  // Only a Device can create a GlobalDeviceMemory instance.
+  friend Device;
+  GlobalDeviceMemory(Device *D, const void *Handle, size_t ElementCount)
+      : GlobalDeviceMemoryBase(D, Handle, ElementCount * sizeof(ElemT)) {}
 };
 
-/// A class to represent the size of a dynamic shared memory buffer on a device.
-///
-/// This class maintains no information about the types to be stored in the
-/// buffer. For the typed version of this class see SharedDeviceMemory<ElemT>.
+/// A class to represent the size of a dynamic shared memory buffer of elements
+/// of type T on a device.
 ///
 /// Shared memory buffers exist only on the device and cannot be manipulated
 /// from the host, so instances of this class do not have an opaque handle, only
 /// a size.
 ///
 /// This type of memory is called "local" memory in OpenCL and "shared" memory
 /// in CUDA, and both platforms follow the rule that the host code only knows
 /// the size of these buffers and does not have a handle to them.
 ///
 /// The treatment of shared memory in StreamExecutor matches the way it is done
 /// in OpenCL, where a kernel takes any number of shared memory sizes as kernel
 /// function arguments.
 ///
 /// In CUDA only one shared memory size argument is allowed per kernel call.
 /// StreamExecutor handles this by allowing CUDA kernel signatures that take
 /// multiple SharedDeviceMemory arguments, and simply adding together all the
 /// shared memory sizes to get the final shared memory size that is used to
 /// launch the kernel.
-class SharedDeviceMemoryBase {
-public:
-  /// Creates an untyped shared memory array from a byte count.
-  SharedDeviceMemoryBase(size_t ByteCount) : ByteCount(ByteCount) {}
-
-  /// Copyable because it is just an array size.
-  SharedDeviceMemoryBase(const SharedDeviceMemoryBase &) = default;
-
-  /// Copy-assignable because it is just an array size.
-  SharedDeviceMemoryBase &operator=(const SharedDeviceMemoryBase &) = default;
-
-  /// Gets the byte count.
-  size_t getByteCount() const { return ByteCount; }
-
-private:
-  size_t ByteCount;
-};
-
-/// Typed wrapper around the untyped SharedDeviceMemoryBase class.
-///
-/// For example, SharedDeviceMemory<int> is a wrapper around
-/// SharedDeviceMemoryBase that represents a buffer of integers stored in shared
-/// device memory.
-template <typename ElemT>
-class SharedDeviceMemory : public SharedDeviceMemoryBase {
+template <typename ElemT> class SharedDeviceMemory {
 public:
   /// Creates a typed area of shared device memory with a given number of
   /// elements.
   static SharedDeviceMemory<ElemT> makeFromElementCount(size_t ElementCount) {
     return SharedDeviceMemory(ElementCount);
   }
 
   /// Copyable because it is just an array size.
   SharedDeviceMemory(const SharedDeviceMemory &) = default;
 
   /// Copy-assignable because it is just an array size.
   SharedDeviceMemory &operator=(const SharedDeviceMemory &) = default;
 
   /// Returns the number of elements of type ElemT that can fit this memory
   /// buffer.
-  size_t getElementCount() const { return getByteCount() / sizeof(ElemT); }
+  size_t getElementCount() const { return ElementCount; }
 
   /// Returns whether this is a single-element memory buffer.
   bool isScalar() const { return getElementCount() == 1; }
 
 private:
   /// Constructs a SharedDeviceMemory instance from an element count.
   ///
   /// This constructor is not public because there is a potential for confusion
   /// between the size of the buffer in bytes and the size of the buffer in
   /// elements.
   ///
   /// The static method makeFromElementCount is provided for users of this class
   /// because its name makes the meaning of the size parameter clear.
   explicit SharedDeviceMemory(size_t ElementCount)
-      : SharedDeviceMemoryBase(ElementCount * sizeof(ElemT)) {}
+      : ElementCount(ElementCount) {}
+
+  size_t ElementCount;
 };
 
 } // namespace streamexecutor
 
 #endif // STREAMEXECUTOR_DEVICEMEMORY_H

streamexecutor/lib/CMakeLists.txt

 add_library(
     utils
     OBJECT
     Utils/Error.cpp)
 
 add_library(
     streamexecutor
     $<TARGET_OBJECTS:utils>
     Device.cpp
+    DeviceMemory.cpp
     Kernel.cpp
     KernelSpec.cpp
     PackedKernelArgumentArray.cpp
     Platform.cpp
     PlatformInterfaces.cpp
     PlatformManager.cpp
     Stream.cpp)
 target_link_libraries(streamexecutor ${llvm_libs})
 
 if(STREAM_EXECUTOR_UNIT_TESTS)
     add_subdirectory(unittests)
 endif()

streamexecutor/lib/DeviceMemory.cpp

+//===-- DeviceMemory.cpp - DeviceMemory implementation --------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+///
+/// \file
+/// Implementation of DeviceMemory class internals.
+///
+//===----------------------------------------------------------------------===//
+
+#include "streamexecutor/DeviceMemory.h"
+
+#include "streamexecutor/Device.h"
+
+namespace streamexecutor {
+
+GlobalDeviceMemoryBase::~GlobalDeviceMemoryBase() {
+  if (Handle) {
+    // TODO(jhen): How to handle errors here.
+    consumeError(TheDevice->freeDeviceMemory(*this));
+  }
+}
+
+} // namespace streamexecutor

streamexecutor/lib/unittests/DeviceTest.cpp

 
 using llvm::ArrayRef;
 using llvm::MutableArrayRef;
 
 TEST_F(DeviceTest, AllocateAndFreeDeviceMemory) {
   se::Expected<se::GlobalDeviceMemory<int>> MaybeMemory =
       Device.allocateDeviceMemory<int>(10);
   EXPECT_TRUE(static_cast<bool>(MaybeMemory));
-  EXPECT_NO_ERROR(Device.freeDeviceMemory(*MaybeMemory));
 }
 
 TEST_F(DeviceTest, AllocateAndFreeHostMemory) {
   se::Expected<int *> MaybeMemory = Device.allocateHostMemory<int>(10);
   EXPECT_TRUE(static_cast<bool>(MaybeMemory));
   EXPECT_NO_ERROR(Device.freeHostMemory(*MaybeMemory));
 }