Main: add lightweight DefaultBuffer class for shadow buffering

OgreMain/include/OgreDefaultHardwareBufferManager.h

@@ -43,6 +43,22 @@ namespace Ogre {
     *  @{
     */
 
+    /// Specialisation of Buffer using malloc e.g. for use as shadow buffer
+    class _OgreExport DefaultBuffer : public Buffer
+    {
+    protected:
+        unsigned char* mData;
+        void* lockImpl(size_t offset, size_t length, LockOptions options);
+        void unlockImpl(void);
+    public:
+        explicit DefaultBuffer(size_t sizeInBytes);
+        ~DefaultBuffer();
+        void readData(size_t offset, size_t length, void* pDest);
+        void writeData(size_t offset, size_t length, const void* pSource, bool discardWholeBuffer = false);
+        void* lock(size_t offset, size_t length, LockOptions options);
+        void unlock(void);
+    };
+
     /// Specialisation of HardwareVertexBuffer for emulation
     class _OgreExport DefaultHardwareVertexBuffer : public HardwareVertexBuffer 
     {

OgreMain/include/OgreHardwareBuffer.h

@@ -122,7 +122,7 @@ namespace Ogre {
             HBL_WRITE_ONLY
 
         };
-        Buffer(size_t sizeInBytes, int usage) : mSizeInBytes(sizeInBytes), mUsage(usage) {}
+        Buffer(size_t sizeInBytes, int usage) : mSizeInBytes(sizeInBytes), mUsage(usage), mIsLocked(false) {}
 
         virtual ~Buffer() {}
         /** Reads data from the buffer and places it in the memory pointed to by pDest.
@@ -131,7 +131,7 @@ namespace Ogre {
         @param pDest The area of memory in which to place the data, must be large enough to
             accommodate the data!
         */
-        virtual void readData(size_t offset, size_t length, void* pDest) = 0;
+        virtual void readData(size_t offset, size_t length, void* pDest) /* const */ = 0;
         /** Writes data to the buffer from an area of system memory; note that you must
             ensure that your buffer is big enough.
         @param offset The byte offset from the start of the buffer to start writing
@@ -152,8 +152,26 @@ namespace Ogre {
         @param length Length of the data to copy, in bytes.
         @param discardWholeBuffer If true, will discard the entire contents of this buffer before copying
         */
-        virtual void copyData(HardwareBuffer& srcBuffer, size_t srcOffset, size_t dstOffset, size_t length,
-                              bool discardWholeBuffer = false) = 0;
+        virtual void copyData(HardwareBuffer& _srcBuffer, size_t srcOffset, size_t dstOffset, size_t length,
+                              bool discardWholeBuffer = false)
+        {
+            auto& srcBuffer = (Buffer&)_srcBuffer; // backward compat
+            const void* srcData = srcBuffer.lock(srcOffset, length, HBL_READ_ONLY);
+            this->writeData(dstOffset, length, srcData, discardWholeBuffer);
+            srcBuffer.unlock();
+        }
+
+        /** Copy all data from another buffer into this one.
+        @remarks
+            Normally these buffers should be of identical size, but if they're
+            not, the routine will use the smallest of the two sizes.
+        */
+        void copyData(HardwareBuffer& _srcBuffer)
+        {
+            auto& srcBuffer = (Buffer&)_srcBuffer; // backward compat
+            size_t sz = std::min(getSizeInBytes(), srcBuffer.getSizeInBytes());
+            copyData(_srcBuffer, 0, 0, sz, true);
+        }
 
         /** Lock the buffer for (potentially) reading / writing.
         @param offset The byte offset from the start of the buffer to lock
@@ -181,15 +199,15 @@ namespace Ogre {
         virtual void unlock() = 0;
 
         /// Returns whether or not this buffer is currently locked.
-        virtual bool isLocked() const  = 0;
-
+        virtual bool isLocked() const { return mIsLocked; }
         /// Returns the size of this buffer in bytes
         size_t getSizeInBytes(void) const { return mSizeInBytes; }
         /// Returns the Usage flags with which this buffer was created
         int getUsage(void) const { return mUsage; }
     protected:
         size_t mSizeInBytes;
         int mUsage;
+        bool mIsLocked;
     };
 
 
@@ -248,10 +266,9 @@ namespace Ogre {
                 HBU_DYNAMIC_WRITE_ONLY_DISCARDABLE = HBU_CPU_TO_GPU,
             };
         protected:
-            bool mIsLocked;
             size_t mLockStart;
             size_t mLockSize;
-            std::unique_ptr<HardwareBuffer> mShadowBuffer;
+            std::unique_ptr<Buffer> mShadowBuffer;
             bool mSystemMemory;
             bool mUseShadowBuffer;
             bool mShadowUpdated;
@@ -265,7 +282,7 @@ namespace Ogre {
         public:
             /// Constructor, to be called by HardwareBufferManager only
             HardwareBuffer(Usage usage, bool systemMemory, bool useShadowBuffer) 
-                : Buffer(0, usage), mIsLocked(false), mLockStart(0), mLockSize(0), mSystemMemory(systemMemory),
+                : Buffer(0, usage), mLockStart(0), mLockSize(0), mSystemMemory(systemMemory),
                 mUseShadowBuffer(useShadowBuffer), mShadowUpdated(false),
                 mSuppressHardwareUpdate(false) 
             {
@@ -330,57 +347,18 @@ namespace Ogre {
 
             }
 
-            /** Copy data from another buffer into this one.
-            @remarks
-                Note that the source buffer must not be created with the
-                usage HBU_WRITE_ONLY otherwise this will fail. 
-            @param srcBuffer The buffer from which to read the copied data
-            @param srcOffset Offset in the source buffer at which to start reading
-            @param dstOffset Offset in the destination buffer to start writing
-            @param length Length of the data to copy, in bytes.
-            @param discardWholeBuffer If true, will discard the entire contents of this buffer before copying
-            */
-            void copyData(HardwareBuffer& srcBuffer, size_t srcOffset, size_t dstOffset, size_t length,
-                          bool discardWholeBuffer = false) override
-            {
-                const void *srcData = srcBuffer.lock(
-                    srcOffset, length, HBL_READ_ONLY);
-                this->writeData(dstOffset, length, srcData, discardWholeBuffer);
-                srcBuffer.unlock();
-            }
-
-            /** Copy all data from another buffer into this one. 
-            @remarks
-                Normally these buffers should be of identical size, but if they're
-                not, the routine will use the smallest of the two sizes.
-            */
-            virtual void copyData(HardwareBuffer& srcBuffer)
-            {
-                size_t sz = std::min(getSizeInBytes(), srcBuffer.getSizeInBytes()); 
-                copyData(srcBuffer, 0, 0, sz, true);
-            }
-
             /// Updates the real buffer from the shadow buffer, if required
             virtual void _updateFromShadow(void)
             {
                 if (mUseShadowBuffer && mShadowUpdated && !mSuppressHardwareUpdate)
                 {
-                    // Do this manually to avoid locking problems
-                    const void *srcData = mShadowBuffer->lockImpl(
-                        mLockStart, mLockSize, HBL_READ_ONLY);
                     // Lock with discard if the whole buffer was locked, otherwise w/o
-                    LockOptions lockOpt;
-                    if (mLockStart == 0 && mLockSize == mSizeInBytes)
-                        lockOpt = HBL_DISCARD;
-                    else
-                        lockOpt = HBL_WRITE_ONLY;
-
-                    void *destData = this->lockImpl(
-                        mLockStart, mLockSize, lockOpt);
+                    LockOptions lockOpt = mLockSize == mSizeInBytes ? HBL_DISCARD : HBL_WRITE_ONLY;
+                    // Do this manually to avoid locking problems
+                    void* destData = this->lockImpl(mLockStart, mLockSize, lockOpt);
                     // Copy shadow to real
-                    memcpy(destData, srcData, mLockSize);
+                    mShadowBuffer->readData(mLockStart, mLockSize, destData);
                     this->unlockImpl();
-                    mShadowBuffer->unlockImpl();
                     mShadowUpdated = false;
                 }
             }

OgreMain/src/OgreDefaultHardwareBufferManager.cpp

@@ -29,6 +29,39 @@ THE SOFTWARE.
 #include "OgreDefaultHardwareBufferManager.h"
 
 namespace Ogre {
+    DefaultBuffer::DefaultBuffer(size_t sizeInBytes)
+        : Buffer(sizeInBytes, HBU_CPU_ONLY)
+    {
+        // Allocate aligned memory for better SIMD processing friendly.
+        mData = static_cast<unsigned char*>(AlignedMemory::allocate(mSizeInBytes));
+    }
+    //-----------------------------------------------------------------------
+    DefaultBuffer::~DefaultBuffer() { AlignedMemory::deallocate(mData); }
+    //-----------------------------------------------------------------------
+    void* DefaultBuffer::lockImpl(size_t offset, size_t length, LockOptions options) { return mData + offset; }
+    //-----------------------------------------------------------------------
+    void DefaultBuffer::unlockImpl() {}
+    //-----------------------------------------------------------------------
+    void* DefaultBuffer::lock(size_t offset, size_t length, LockOptions options)
+    {
+        mIsLocked = true;
+        return mData + offset;
+    }
+    void DefaultBuffer::unlock(void) { mIsLocked = false; }
+    //-----------------------------------------------------------------------
+    void DefaultBuffer::readData(size_t offset, size_t length, void* pDest)
+    {
+        assert((offset + length) <= mSizeInBytes);
+        memcpy(pDest, mData + offset, length);
+    }
+    //-----------------------------------------------------------------------
+    void DefaultBuffer::writeData(size_t offset, size_t length, const void* pSource, bool discardWholeBuffer)
+    {
+        assert((offset + length) <= mSizeInBytes);
+        // ignore discard, memory is not guaranteed to be zeroised
+        memcpy(mData + offset, pSource, length);
+    }
+
 
     DefaultHardwareVertexBuffer::DefaultHardwareVertexBuffer(size_t vertexSize, size_t numVertices, 
                                                              HardwareBuffer::Usage usage)

OgreMain/src/OgreHardwareIndexBuffer.cpp

@@ -56,8 +56,7 @@ namespace Ogre {
         // Create a shadow buffer if required
         if (mUseShadowBuffer)
         {
-            mShadowBuffer.reset(new DefaultHardwareIndexBuffer(mIndexType,
-                mNumIndexes, HardwareBuffer::HBU_DYNAMIC));
+            mShadowBuffer.reset(new DefaultBuffer(mSizeInBytes));
         }
 
 

OgreMain/src/OgreHardwareUniformBuffer.cpp

@@ -44,7 +44,7 @@ namespace Ogre {
         // Create a shadow buffer if required
         if (mUseShadowBuffer)
         {
-            mShadowBuffer.reset(new DefaultHardwareUniformBuffer(mMgr, sizeBytes, HardwareBuffer::HBU_DYNAMIC, false));
+            mShadowBuffer.reset(new DefaultBuffer(sizeBytes));
         }
     }
 

OgreMain/src/OgreHardwareVertexBuffer.cpp

@@ -48,8 +48,7 @@ namespace Ogre {
         // Create a shadow buffer if required
         if (mUseShadowBuffer)
         {
-            mShadowBuffer.reset(new DefaultHardwareVertexBuffer(mMgr, mVertexSize,
-                    mNumVertices, HardwareBuffer::HBU_DYNAMIC));
+            mShadowBuffer.reset(new DefaultBuffer(mSizeInBytes));
         }
 
     }

RenderSystems/Direct3D11/include/OgreD3D11HardwareBuffer.h

@@ -74,7 +74,7 @@ namespace Ogre {
         /** See HardwareBuffer. We perform a hardware copy here. */
         void copyData(HardwareBuffer& srcBuffer, size_t srcOffset, 
             size_t dstOffset, size_t length, bool discardWholeBuffer = false);
-		void copyDataImpl(HardwareBuffer& srcBuffer, size_t srcOffset,
+		void copyDataImpl(Buffer& srcBuffer, size_t srcOffset,
 			size_t dstOffset, size_t length, bool discardWholeBuffer = false);
 		/// Updates the real buffer from the shadow buffer, if required
 		virtual void _updateFromShadow(void);

RenderSystems/Direct3D11/src/OgreD3D11HardwareBuffer.cpp

@@ -213,7 +213,7 @@ namespace Ogre {
 		copyDataImpl(srcBuffer, srcOffset, dstOffset, length, discardWholeBuffer);
 	}
 	//---------------------------------------------------------------------
-	void D3D11HardwareBuffer::copyDataImpl(HardwareBuffer& srcBuffer, size_t srcOffset, 
+	void D3D11HardwareBuffer::copyDataImpl(Buffer& srcBuffer, size_t srcOffset,
 		size_t dstOffset, size_t length, bool discardWholeBuffer)
 	{
         // If we're copying same-size buffers in their entirety...