Buffer.h

//
//  Created by Sam Gateau on 10/8/2014.
//  Split from Resource.h/Resource.cpp by Bradley Austin Davis on 2016/08/07
//  Copyright 2014 High Fidelity, Inc.
//
//  Distributed under the Apache License, Version 2.0.
//  See the accompanying file LICENSE or http://www.apache.org/licenses/LICENSE-2.0.html
//
#ifndef hifi_gpu_Buffer_h
#define hifi_gpu_Buffer_h
 
#include <atomic>
 
#if _DEBUG
#include <QtCore/QDebug>
#include "GPULogging.h"
#endif
 
#include "Forward.h"
#include "Format.h"
#include "Resource.h"
#include "Sysmem.h"
#include "PageManager.h"
#include "Metric.h"
 
namespace gpu {
class Buffer : public Resource {
    static ContextMetricCount _bufferCPUCount;
    static ContextMetricSize _bufferCPUMemSize;
 
public:
    using Flag = PageManager::Flag;
 
    // VKTODO: Make this independent of VkBufferUsageFlagBits.
    // Flags match VkBufferUsageFlagBits for convenience... do not modify
    enum Usage
    {
        // These values are unused in our API
        //TransferSrc = 0x0001,
        //TransferDst = 0x0002,
        //UniformTexelBuffer = 0x0004,
        //StorageTexelBuffer = 0x0008,
        UniformBuffer = 0x0010,
        ResourceBuffer = 0x0020,
        IndexBuffer = 0x0040,
        VertexBuffer = 0x0080,
        IndirectBuffer = 0x0100,
        AllFlags = 0x01F3
    };
 
    class Update {
    public:
        Update(const Buffer& buffer);
        Update(const Update& other);
        Update(Update&& other);
        void apply() const;
 
    private:
        const Buffer& buffer;
        size_t updateNumber;
        Size size;
        PageManager::Pages dirtyPages;
        std::vector<uint8> dirtyData;
    };
 
    static uint32_t getBufferCPUCount();
    static Size getBufferCPUMemSize();
 
    // VKTODO: experiment with making buffer usage inferred from other commands on the batch.
    Buffer(uint32_t usage, Size pageSize = PageManager::DEFAULT_PAGE_SIZE);
 
    template <typename T>
    static Buffer* createBuffer(uint32_t usage, const std::vector<T>& v) {
        return new Buffer(usage, sizeof(T) * v.size(), (const gpu::Byte*)v.data());
    }
 
    Buffer(uint32_t usage, Size size, const Byte* bytes, Size pageSize = PageManager::DEFAULT_PAGE_SIZE);
    Buffer(const Buffer& buf);             // deep copy of the sysmem buffer
    Buffer& operator=(const Buffer& buf);  // deep copy of the sysmem buffer
    ~Buffer();
 
    // The size in bytes of data stored in the buffer
    Size getSize() const override;
    template <typename T>
    Size getNumTypedElements() const {
        return getSize() / sizeof(T);
    };
 
    const Byte* getData() const { return getSysmem().readData(); }
 
    uint32_t getUsage() const { return _usage; }
 
    // Resize the buffer
    // Keep previous data [0 to min(pSize, mSize)]
    Size resize(Size pSize);
 
    // Assign data bytes and size (allocate for size, then copy bytes if exists)
    // \return the size of the buffer
    Size setData(Size size, const Byte* data);
 
    // Assign data bytes and size (allocate for size, then copy bytes if exists)
    // \return the number of bytes copied
    Size setSubData(Size offset, Size size, const Byte* data);
 
    template <typename T>
    Size setSubData(Size index, const T& t) {
        Size offset = index * sizeof(T);
        Size size = sizeof(T);
        return setSubData(offset, size, reinterpret_cast<const Byte*>(&t));
    }
 
    template <typename T>
    Size setSubData(Size index, const std::vector<T>& t) {
        if (t.empty()) {
            return 0;
        }
        Size offset = index * sizeof(T);
        Size size = t.size() * sizeof(T);
        return setSubData(offset, size, reinterpret_cast<const Byte*>(&t[0]));
    }
 
    // Append new data at the end of the current buffer
    // do a resize( size + getSize) and copy the new data
    // \return the number of bytes copied
    Size append(Size size, const Byte* data);
 
    template <typename T>
    Size append(const T& t) {
        return append(sizeof(t), reinterpret_cast<const Byte*>(&t));
    }
 
    template <typename T>
    Size append(const std::vector<T>& t) {
        if (t.empty()) {
            return _end;
        }
        return append(sizeof(T) * t.size(), reinterpret_cast<const Byte*>(&t[0]));
    }
 
    const GPUObjectPointer gpuObject{};
 
    // Access the sysmem object, limited to ourselves and GPUObject derived classes
    const Sysmem& getSysmem() const { return _sysmem; }
 
    bool isDirty() const { return _pages(PageManager::DIRTY); }
 
    void applyUpdate(const Update& update);
 
    // Main thread operation to say that the buffer is ready to be used as a frame
    Update getUpdate() const;
 
protected:
    // For use by the render thread to avoid the intermediate step of getUpdate/applyUpdate
    void flush() const;
 
    // FIXME don't maintain a second buffer continuously.  We should be able to apply updates
    // directly to the GL object and discard _renderSysmem and _renderPages
    mutable PageManager _renderPages;
    mutable Sysmem _renderSysmem;
 
    mutable std::atomic<size_t> _getUpdateCount{ 0 };
    mutable std::atomic<size_t> _applyUpdateCount{ 0 };
 
    void markDirty(Size offset, Size bytes);
 
    template <typename T>
    void markDirty(Size index, Size count = 1) {
        markDirty(sizeof(T) * index, sizeof(T) * count);
    }
 
    Sysmem& editSysmem() { return _sysmem; }
    Byte* editData() { return editSysmem().editData(); }
 
    mutable PageManager _pages;
    Size _end{ 0 };
    Sysmem _sysmem;
    const uint32_t _usage{ 0 };
 
    friend class Serializer;
    friend class Deserializer;
    friend class BufferView;
    friend class Frame;
    friend class Batch;
 
    // FIXME find a more generic way to do this.
    friend class ::gpu::vk::VKBuffer;
    friend class ::gpu::vk::VKBackend;
    friend class gl::GLBackend;
    friend class gl::GLBuffer;
    friend class gl41::GL41Buffer;
    friend class gl45::GL45Buffer;
    friend class gles::GLESBuffer;
    friend class vk::VKBuffer;
};
 
using BufferUpdates = std::vector<Buffer::Update>;
 
typedef std::shared_ptr<Buffer> BufferPointer;
typedef std::vector<BufferPointer> Buffers;
 
class BufferView {
protected:
    static const Resource::Size DEFAULT_OFFSET{ 0 };
    static const Element DEFAULT_ELEMENT;
 
public:
    using Size = Resource::Size;
    using Index = int32_t;
 
    BufferPointer _buffer;
    Size _offset{ 0 };
    Size _size{ 0 };
    Element _element{ DEFAULT_ELEMENT };
    uint16 _stride{ 0 };
 
    BufferView(const BufferView& view) = default;
    BufferView& operator=(const BufferView& view) = default;
 
    BufferView();
    BufferView(const Element& element);
    BufferView(Buffer* newBuffer, const Element& element = DEFAULT_ELEMENT);
    BufferView(const BufferPointer& buffer, const Element& element = DEFAULT_ELEMENT);
    BufferView(const BufferPointer& buffer, Size offset, Size size, const Element& element = DEFAULT_ELEMENT);
    BufferView(const BufferPointer& buffer, Size offset, Size size, uint16 stride, const Element& element = DEFAULT_ELEMENT);
 
    Size getNumElements() const { return _size / _stride; }
 
    //Template iterator with random access on the buffer sysmem
    template<typename T>
    class Iterator {
    public:
        using iterator_category = std::random_access_iterator_tag;
        using value_type = T;
        using difference_type = Index;
        using pointer = const value_type*;
        using reference = const value_type&;
 
        Iterator(T* ptr = NULL, int stride = sizeof(T)): _ptr(ptr), _stride(stride) { }
        Iterator(const Iterator<T>& iterator) = default;
        ~Iterator() {}
 
        Iterator<T>& operator=(const Iterator<T>& iterator) = default;
        Iterator<T>& operator=(T* ptr) {
            _ptr = ptr;
            // stride is left unchanged
            return (*this);
        }
 
        operator bool() const {
            if (_ptr)
                return true;
            else
                return false;
        }
 
        bool operator==(const Iterator<T>& iterator) const { return (_ptr == iterator.getConstPtr()); }
        bool operator!=(const Iterator<T>& iterator) const { return (_ptr != iterator.getConstPtr()); }
        bool operator<(const Iterator<T>& iterator) const { return (_ptr < iterator.getConstPtr()); }
        bool operator>(const Iterator<T>& iterator) const { return (_ptr > iterator.getConstPtr()); }
 
        void movePtr(const Index& movement) {
            auto byteptr = ((Byte*)_ptr);
            byteptr += _stride * movement;
            _ptr = (T*)byteptr;
        }
 
        Iterator<T>& operator+=(const Index& movement) {
            movePtr(movement);
            return (*this);
        }
        Iterator<T>& operator-=(const Index& movement) {
            movePtr(-movement);
            return (*this);
        }
        Iterator<T>& operator++() {
            movePtr(1);
            return (*this);
        }
        Iterator<T>& operator--() {
            movePtr(-1);
            return (*this);
        }
        Iterator<T> operator++(Index) {
            auto temp(*this);
            movePtr(1);
            return temp;
        }
        Iterator<T> operator--(Index) {
            auto temp(*this);
            movePtr(-1);
            return temp;
        }
        Iterator<T> operator+(const Index& movement) {
            auto oldPtr = _ptr;
            movePtr(movement);
            auto temp(*this);
            _ptr = oldPtr;
            return temp;
        }
        Iterator<T> operator-(const Index& movement) {
            auto oldPtr = _ptr;
            movePtr(-movement);
            auto temp(*this);
            _ptr = oldPtr;
            return temp;
        }
 
        Index operator-(const Iterator<T>& iterator) { return (iterator.getPtr() - this->getPtr()) / sizeof(T); }
 
        T& operator*() { return *_ptr; }
        const T& operator*() const { return *_ptr; }
        T* operator->() { return _ptr; }
 
        T* getPtr() const { return _ptr; }
        const T* getConstPtr() const { return _ptr; }
 
    protected:
        T* _ptr;
        int _stride;
    };
 
#if 0
    // Direct memory access to the buffer contents is incompatible with the paging memory scheme
    template <typename T> Iterator<T> begin() { return Iterator<T>(&edit<T>(0), _stride); }
    template <typename T> Iterator<T> end() { return Iterator<T>(&edit<T>(getNum<T>()), _stride); }
#else
    template <typename T>
    Iterator<const T> begin() const {
        return Iterator<const T>(&get<T>(), _stride);
    }
    template <typename T>
    Iterator<const T> end() const {
        // reimplement get<T> without bounds checking
        Resource::Size elementOffset = getNum<T>() * _stride + _offset;
        return Iterator<const T>((reinterpret_cast<const T*>(_buffer->getData() + elementOffset)), _stride);
    }
#endif
    template <typename T>
    Iterator<const T> cbegin() const {
        return Iterator<const T>(&get<T>(), _stride);
    }
    template <typename T>
    Iterator<const T> cend() const {
        // reimplement get<T> without bounds checking
        Resource::Size elementOffset = getNum<T>() * _stride + _offset;
        return Iterator<const T>((reinterpret_cast<const T*>(_buffer->getData() + elementOffset)), _stride);
    }
 
    // the number of elements of the specified type fitting in the view size
    template <typename T>
    Index getNum() const {
        return Index(_size / _stride);
    }
 
    template <typename T>
    const T& get() const {
#if _DEBUG
        if (!_buffer) {
            qCDebug(gpulogging) << "Accessing null gpu::buffer!";
        }
        if (sizeof(T) > (_buffer->getSize() - _offset)) {
            qCDebug(gpulogging) << "Accessing buffer in non allocated memory, element size = " << sizeof(T)
                                << " available space in buffer at offset is = " << (_buffer->getSize() - _offset);
        }
        if (sizeof(T) > _size) {
            qCDebug(gpulogging) << "Accessing buffer outside the BufferView range, element size = " << sizeof(T)
                                << " when bufferView size = " << _size;
        }
#endif
        const T* t = (reinterpret_cast<const T*>(_buffer->getData() + _offset));
        return *(t);
    }
 
    template <typename T>
    T& edit() {
#if _DEBUG
        if (!_buffer) {
            qCDebug(gpulogging) << "Accessing null gpu::buffer!";
        }
        if (sizeof(T) > (_buffer->getSize() - _offset)) {
            qCDebug(gpulogging) << "Accessing buffer in non allocated memory, element size = " << sizeof(T)
                                << " available space in buffer at offset is = " << (_buffer->getSize() - _offset);
        }
        if (sizeof(T) > _size) {
            qCDebug(gpulogging) << "Accessing buffer outside the BufferView range, element size = " << sizeof(T)
                                << " when bufferView size = " << _size;
        }
#endif
        _buffer->markDirty(_offset, sizeof(T));
        T* t = (reinterpret_cast<T*>(_buffer->editData() + _offset));
        return *(t);
    }
 
    template <typename T>
    const T& get(const Index index) const {
        Resource::Size elementOffset = index * _stride + _offset;
#if _DEBUG
        if (!_buffer) {
            qCDebug(gpulogging) << "Accessing null gpu::buffer!";
        }
        if (sizeof(T) > (_buffer->getSize() - elementOffset)) {
            qCDebug(gpulogging) << "Accessing buffer in non allocated memory, index = " << index
                                << ", element size = " << sizeof(T)
                                << " available space in buffer at offset is = " << (_buffer->getSize() - elementOffset);
        }
        if (index > getNum<T>()) {
            qCDebug(gpulogging) << "Accessing buffer outside the BufferView range, index = " << index
                                << " number elements = " << getNum<T>();
        }
#endif
        return *(reinterpret_cast<const T*>(_buffer->getData() + elementOffset));
    }
 
    template <typename T>
    T& edit(const Index index) const {
        Resource::Size elementOffset = index * _stride + _offset;
#if _DEBUG
        if (!_buffer) {
            qCDebug(gpulogging) << "Accessing null gpu::buffer!";
        }
        if (sizeof(T) > (_buffer->getSize() - elementOffset)) {
            qCDebug(gpulogging) << "Accessing buffer in non allocated memory, index = " << index
                                << ", element size = " << sizeof(T)
                                << " available space in buffer at offset is = " << (_buffer->getSize() - elementOffset);
        }
        if (index > getNum<T>()) {
            qCDebug(gpulogging) << "Accessing buffer outside the BufferView range, index = " << index
                                << " number elements = " << getNum<T>();
        }
#endif
        _buffer->markDirty(elementOffset, sizeof(T));
        return *(reinterpret_cast<T*>(_buffer->editData() + elementOffset));
    }
};
 
template <class T>
class StructBuffer : public gpu::BufferView {
public:
    template <class U>
    static BufferPointer makeBuffer(uint32_t usage = gpu::Buffer::UniformBuffer) {
        U t;
        return std::make_shared<gpu::Buffer>(usage, sizeof(U), (const gpu::Byte*)&t, sizeof(U));
    }
    ~StructBuffer(){};
    StructBuffer() : gpu::BufferView(makeBuffer<T>()) {}
 
    T& edit() { return BufferView::edit<T>(0); }
    const T& get() const { return BufferView::get<T>(0); }
    const T* operator->() const { return &get(); }
};
};  // namespace gpu
 
#endif