path: root/src/render/renderers/opengl/graphicshelpers/submissioncontext.cpp

/****************************************************************************
**
** Copyright (C) 2017 Klaralvdalens Datakonsult AB (KDAB).
** Contact: https://www.qt.io/licensing/
**
** This file is part of the Qt3D module of the Qt Toolkit.
**
** $QT_BEGIN_LICENSE:LGPL$
** Commercial License Usage
** Licensees holding valid commercial Qt licenses may use this file in
** accordance with the commercial license agreement provided with the
** Software or, alternatively, in accordance with the terms contained in
** a written agreement between you and The Qt Company. For licensing terms
** and conditions see https://www.qt.io/terms-conditions. For further
** information use the contact form at https://www.qt.io/contact-us.
**
** GNU Lesser General Public License Usage
** Alternatively, this file may be used under the terms of the GNU Lesser
** General Public License version 3 as published by the Free Software
** Foundation and appearing in the file LICENSE.LGPL3 included in the
** packaging of this file. Please review the following information to
** ensure the GNU Lesser General Public License version 3 requirements
** will be met: https://www.gnu.org/licenses/lgpl-3.0.html.
**
** GNU General Public License Usage
** Alternatively, this file may be used under the terms of the GNU
** General Public License version 2.0 or (at your option) the GNU General
** Public license version 3 or any later version approved by the KDE Free
** Qt Foundation. The licenses are as published by the Free Software
** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3
** included in the packaging of this file. Please review the following
** information to ensure the GNU General Public License requirements will
** be met: https://www.gnu.org/licenses/gpl-2.0.html and
** https://www.gnu.org/licenses/gpl-3.0.html.
**
** $QT_END_LICENSE$
**
****************************************************************************/

#include "submissioncontext_p.h"

#include <Qt3DRender/qgraphicsapifilter.h>
#include <Qt3DRender/qparameter.h>
#include <Qt3DRender/qcullface.h>
#include <Qt3DRender/private/renderlogging_p.h>
#include <Qt3DRender/private/shader_p.h>
#include <Qt3DRender/private/material_p.h>
#include <Qt3DRender/private/gltexture_p.h>
#include <Qt3DRender/private/buffer_p.h>
#include <Qt3DRender/private/attribute_p.h>
#include <Qt3DRender/private/rendercommand_p.h>
#include <Qt3DRender/private/renderstateset_p.h>
#include <Qt3DRender/private/rendertarget_p.h>
#include <Qt3DRender/private/graphicshelperinterface_p.h>
#include <Qt3DRender/private/renderer_p.h>
#include <Qt3DRender/private/nodemanagers_p.h>
#include <Qt3DRender/private/buffermanager_p.h>
#include <Qt3DRender/private/managers_p.h>
#include <Qt3DRender/private/gltexturemanager_p.h>
#include <Qt3DRender/private/attachmentpack_p.h>
#include <Qt3DRender/private/qbuffer_p.h>
#include <Qt3DRender/private/renderbuffer_p.h>
#include <Qt3DRender/private/stringtoint_p.h>
#include <QOpenGLShaderProgram>

#if !defined(QT_OPENGL_ES_2)
#include <QOpenGLFunctions_2_0>
#include <QOpenGLFunctions_3_2_Core>
#include <QOpenGLFunctions_3_3_Core>
#include <QOpenGLFunctions_4_3_Core>
#include <Qt3DRender/private/graphicshelpergl2_p.h>
#include <Qt3DRender/private/graphicshelpergl3_2_p.h>
#include <Qt3DRender/private/graphicshelpergl3_3_p.h>
#include <Qt3DRender/private/graphicshelpergl4_p.h>
#endif
#include <Qt3DRender/private/graphicshelperes2_p.h>
#include <Qt3DRender/private/graphicshelperes3_p.h>

#include <QSurface>
#include <QWindow>
#include <QOpenGLTexture>
#include <QOpenGLDebugLogger>

QT_BEGIN_NAMESPACE

#ifndef GL_READ_FRAMEBUFFER
#define GL_READ_FRAMEBUFFER 0x8CA8
#endif

#ifndef GL_DRAW_FRAMEBUFFER
#define GL_DRAW_FRAMEBUFFER 0x8CA9
#endif

namespace Qt3DRender {
namespace Render {


class TextureExtRendererLocker
{
public:
    static void lock(GLTexture *tex)
    {
        if (!tex->isExternalRenderingEnabled())
            return;
        if (s_lockHash.keys().contains(tex)) {
            ++s_lockHash[tex];
        } else {
            tex->externalRenderingLock()->lock();
            s_lockHash[tex] = 1;
        }
    }
    static void unlock(GLTexture *tex)
    {
        if (!tex->isExternalRenderingEnabled())
            return;
        if (!s_lockHash.keys().contains(tex))
            return;

        --s_lockHash[tex];
        if (s_lockHash[tex] == 0) {
            s_lockHash.remove(tex);
            tex->externalRenderingLock()->unlock();
        }
    }
private:
    static QHash<GLTexture*, int> s_lockHash;
};

QHash<GLTexture*, int> TextureExtRendererLocker::s_lockHash = QHash<GLTexture*, int>();

static QHash<unsigned int, SubmissionContext*> static_contexts;

namespace {

GLBuffer::Type attributeTypeToGLBufferType(QAttribute::AttributeType type)
{
    switch (type) {
    case QAttribute::VertexAttribute:
        return GLBuffer::ArrayBuffer;
    case QAttribute::IndexAttribute:
        return GLBuffer::IndexBuffer;
    case QAttribute::DrawIndirectAttribute:
        return GLBuffer::DrawIndirectBuffer;
    default:
        Q_UNREACHABLE();
    }
}

void copyGLFramebufferDataToImage(QImage &img, const uchar *srcData, uint stride, uint width, uint height, QAbstractTexture::TextureFormat format)
{
    switch (format) {
    case QAbstractTexture::RGBA32F:
        {
            uchar *srcScanline = (uchar *)srcData + stride * (height - 1);
            for (uint i = 0; i < height; ++i) {
                uchar *dstScanline = img.scanLine(i);
                float *pSrc = (float*)srcScanline;
                for (uint j = 0; j < width; j++) {
                    *dstScanline++ = (uchar)(255.0f * qBound(0.0f, pSrc[4*j+2], 1.0f));
                    *dstScanline++ = (uchar)(255.0f * qBound(0.0f, pSrc[4*j+1], 1.0f));
                    *dstScanline++ = (uchar)(255.0f * qBound(0.0f, pSrc[4*j+0], 1.0f));
                    *dstScanline++ = (uchar)(255.0f * qBound(0.0f, pSrc[4*j+3], 1.0f));
                }
                srcScanline -= stride;
            }
        } break;
    default:
        {
            uchar* srcScanline = (uchar *)srcData + stride * (height - 1);
            for (uint i = 0; i < height; ++i) {
                memcpy(img.scanLine(i), srcScanline, stride);
                srcScanline -= stride;
            }
        } break;
    }
}

// Render States Helpers
template<typename GenericState>
void applyStateHelper(const GenericState *state, SubmissionContext *gc)
{
    Q_UNUSED(state);
    Q_UNUSED(gc);
}

template<>
void applyStateHelper<AlphaFunc>(const AlphaFunc *state, SubmissionContext *gc)
{
    const auto values = state->values();
    gc->alphaTest(std::get<0>(values), std::get<1>(values));
}

template<>
void applyStateHelper<BlendEquationArguments>(const BlendEquationArguments *state, SubmissionContext *gc)
{
    const auto values = state->values();
    // Un-indexed BlendEquationArguments -> Use normal GL1.0 functions
    if (std::get<5>(values) < 0) {
        if (std::get<4>(values)) {
            gc->openGLContext()->functions()->glEnable(GL_BLEND);
            gc->openGLContext()->functions()->glBlendFuncSeparate(std::get<0>(values), std::get<1>(values), std::get<2>(values), std::get<3>(values));
        } else {
            gc->openGLContext()->functions()->glDisable(GL_BLEND);
        }
    }
    // BlendEquationArguments for a particular Draw Buffer. Different behaviours for
    //  (1) 3.0-3.3: only enablei/disablei supported.
    //  (2) 4.0+: all operations supported.
    // We just ignore blend func parameter for (1), so no warnings get
    // printed.
    else {
        if (std::get<4>(values)) {
            gc->enablei(GL_BLEND, std::get<5>(values));
            if (gc->supportsDrawBuffersBlend()) {
                gc->blendFuncSeparatei(std::get<5>(values), std::get<0>(values), std::get<1>(values), std::get<2>(values), std::get<3>(values));
            }
        } else {
            gc->disablei(GL_BLEND, std::get<5>(values));
        }
    }
}

template<>
void applyStateHelper<BlendEquation>(const BlendEquation *state, SubmissionContext *gc)
{
    gc->blendEquation(std::get<0>(state->values()));
}

template<>
void applyStateHelper<MSAAEnabled>(const MSAAEnabled *state, SubmissionContext *gc)
{
    gc->setMSAAEnabled(std::get<0>(state->values()));
}


template<>
void applyStateHelper<DepthTest>(const DepthTest *state, SubmissionContext *gc)
{
    gc->depthTest(std::get<0>(state->values()));
}


template<>
void applyStateHelper<NoDepthMask>(const NoDepthMask *state, SubmissionContext *gc)
{
    gc->depthMask(std::get<0>(state->values()));
}


template<>
void applyStateHelper<CullFace>(const CullFace *state, SubmissionContext *gc)
{
    const auto values = state->values();
    if (std::get<0>(values) == QCullFace::NoCulling) {
        gc->openGLContext()->functions()->glDisable(GL_CULL_FACE);
    } else {
        gc->openGLContext()->functions()->glEnable(GL_CULL_FACE);
        gc->openGLContext()->functions()->glCullFace(std::get<0>(values));
    }
}

template<>
void applyStateHelper<FrontFace>(const FrontFace *state, SubmissionContext *gc)
{
    gc->frontFace(std::get<0>(state->values()));
}

template<>
void applyStateHelper<ScissorTest>(const ScissorTest *state, SubmissionContext *gc)
{
    const auto values = state->values();
    gc->openGLContext()->functions()->glEnable(GL_SCISSOR_TEST);
    gc->openGLContext()->functions()->glScissor(std::get<0>(values), std::get<1>(values), std::get<2>(values), std::get<3>(values));
}

template<>
void applyStateHelper<StencilTest>(const StencilTest *state, SubmissionContext *gc)
{
    const auto values = state->values();
    gc->openGLContext()->functions()->glEnable(GL_STENCIL_TEST);
    gc->openGLContext()->functions()->glStencilFuncSeparate(GL_FRONT, std::get<0>(values), std::get<1>(values), std::get<2>(values));
    gc->openGLContext()->functions()->glStencilFuncSeparate(GL_BACK, std::get<3>(values), std::get<4>(values), std::get<5>(values));
}

template<>
void applyStateHelper<AlphaCoverage>(const AlphaCoverage *, SubmissionContext *gc)
{
    gc->setAlphaCoverageEnabled(true);
}

template<>
void applyStateHelper<PointSize>(const PointSize *state, SubmissionContext *gc)
{
    const auto values = state->values();
    gc->pointSize(std::get<0>(values), std::get<1>(values));
}


template<>
void applyStateHelper<PolygonOffset>(const PolygonOffset *state, SubmissionContext *gc)
{
    const auto values = state->values();
    gc->openGLContext()->functions()->glEnable(GL_POLYGON_OFFSET_FILL);
    gc->openGLContext()->functions()->glPolygonOffset(std::get<0>(values), std::get<1>(values));
}

template<>
void applyStateHelper<ColorMask>(const ColorMask *state, SubmissionContext *gc)
{
    const auto values = state->values();
    gc->openGLContext()->functions()->glColorMask(std::get<0>(values), std::get<1>(values), std::get<2>(values), std::get<3>(values));
}

template<>
void applyStateHelper<ClipPlane>(const ClipPlane *state, SubmissionContext *gc)
{
    const auto values = state->values();
    gc->enableClipPlane(std::get<0>(values));
    gc->setClipPlane(std::get<0>(values), std::get<1>(values), std::get<2>(values));
}

template<>
void applyStateHelper<SeamlessCubemap>(const SeamlessCubemap *, SubmissionContext *gc)
{
    gc->setSeamlessCubemap(true);
}

template<>
void applyStateHelper<StencilOp>(const StencilOp *state, SubmissionContext *gc)
{
    const auto values = state->values();
    gc->openGLContext()->functions()->glStencilOpSeparate(GL_FRONT, std::get<0>(values), std::get<1>(values), std::get<2>(values));
    gc->openGLContext()->functions()->glStencilOpSeparate(GL_BACK, std::get<3>(values), std::get<4>(values), std::get<5>(values));
}

template<>
void applyStateHelper<StencilMask>(const StencilMask *state, SubmissionContext *gc)
{
    const auto values = state->values();
    gc->openGLContext()->functions()->glStencilMaskSeparate(GL_FRONT, std::get<0>(values));
    gc->openGLContext()->functions()->glStencilMaskSeparate(GL_BACK, std::get<1>(values));
}

template<>
void applyStateHelper<Dithering>(const Dithering *, SubmissionContext *gc)
{
    gc->openGLContext()->functions()->glEnable(GL_DITHER);
}

#ifndef GL_LINE_SMOOTH
#define GL_LINE_SMOOTH 0x0B20
#endif

template<>
void applyStateHelper<LineWidth>(const LineWidth *state, SubmissionContext *gc)
{
    const auto values = state->values();
    if (std::get<1>(values))
        gc->openGLContext()->functions()->glEnable(GL_LINE_SMOOTH);
    else
        gc->openGLContext()->functions()->glDisable(GL_LINE_SMOOTH);

    gc->openGLContext()->functions()->glLineWidth(std::get<0>(values));
}

} // anonymous

unsigned int nextFreeContextId()
{
    for (unsigned int i=0; i < 0xffff; ++i) {
        if (!static_contexts.contains(i))
            return i;
    }

    qFatal("Couldn't find free context ID");
    return 0;
}

SubmissionContext::SubmissionContext()
    : GraphicsContext()
    , m_ownCurrent(true)
    , m_id(nextFreeContextId())
    , m_surface(nullptr)
    , m_activeShader(nullptr)
    , m_activeShaderDNA(0)
    , m_renderTargetFormat(QAbstractTexture::NoFormat)
    , m_currClearStencilValue(0)
    , m_currClearDepthValue(1.f)
    , m_currClearColorValue(0,0,0,0)
    , m_material(nullptr)
    , m_activeFBO(0)
    , m_boundArrayBuffer(nullptr)
    , m_stateSet(nullptr)
    , m_renderer(nullptr)
    , m_uboTempArray(QByteArray(1024, 0))
    , m_currentVAO(nullptr)
{
    static_contexts[m_id] = this;
}

SubmissionContext::~SubmissionContext()
{
    releaseOpenGL();

    Q_ASSERT(static_contexts[m_id] == this);
    static_contexts.remove(m_id);
}

void SubmissionContext::initialize()
{
    GraphicsContext::initialize();
    m_activeTextures.resize(maxTextureUnitsCount());
}

void SubmissionContext::resolveRenderTargetFormat()
{
    const QSurfaceFormat format = m_gl->format();
    const uint a = (format.alphaBufferSize() == -1) ? 0 : format.alphaBufferSize();
    const uint r = format.redBufferSize();
    const uint g = format.greenBufferSize();
    const uint b = format.blueBufferSize();

#define RGBA_BITS(r,g,b,a) (r | (g << 6) | (b << 12) | (a << 18))

    const uint bits = RGBA_BITS(r,g,b,a);
    switch (bits) {
    case RGBA_BITS(8,8,8,8):
        m_renderTargetFormat = QAbstractTexture::RGBA8_UNorm;
        break;
    case RGBA_BITS(8,8,8,0):
        m_renderTargetFormat = QAbstractTexture::RGB8_UNorm;
        break;
    case RGBA_BITS(5,6,5,0):
        m_renderTargetFormat = QAbstractTexture::R5G6B5;
        break;
    }
#undef RGBA_BITS
}

bool SubmissionContext::beginDrawing(QSurface *surface)
{
    Q_ASSERT(surface);
    Q_ASSERT(m_gl);

    m_surface = surface;

    // TO DO: Find a way to make to pause work if the window is not exposed
    //    if (m_surface && m_surface->surfaceClass() == QSurface::Window) {
    //        qDebug() << Q_FUNC_INFO << 1;
    //        if (!static_cast<QWindow *>(m_surface)->isExposed())
    //            return false;
    //        qDebug() << Q_FUNC_INFO << 2;
    //    }

    // Makes the surface current on the OpenGLContext
    // and sets the right glHelper
    m_ownCurrent = !(m_gl->surface() == m_surface);
    if (m_ownCurrent && !makeCurrent(m_surface))
        return false;

    // TODO: cache surface format somewhere rather than doing this every time render surface changes
    resolveRenderTargetFormat();

    // Sets or Create the correct m_glHelper
    // for the current surface
    activateGLHelper();

#if defined(QT3D_RENDER_ASPECT_OPENGL_DEBUG)
    GLint err = m_gl->functions()->glGetError();
    if (err != 0) {
        qCWarning(Backend) << Q_FUNC_INFO << "glGetError:" << err;
    }
#endif

    if (!isInitialized())
        initialize();

    // need to reset these values every frame, may get overwritten elsewhere
    m_gl->functions()->glClearColor(m_currClearColorValue.redF(), m_currClearColorValue.greenF(), m_currClearColorValue.blueF(), m_currClearColorValue.alphaF());
    m_gl->functions()->glClearDepthf(m_currClearDepthValue);
    m_gl->functions()->glClearStencil(m_currClearStencilValue);


    if (m_activeShader) {
        m_activeShader = nullptr;
        m_activeShaderDNA = 0;
    }

    // reset active textures
    for (int u = 0; u < m_activeTextures.size(); ++u)
        m_activeTextures[u].texture = nullptr;

    m_boundArrayBuffer = nullptr;

    static int callCount = 0;
    ++callCount;
    const int shaderPurgePeriod = 600;
    if (callCount % shaderPurgePeriod == 0)
        m_shaderCache->purge();

    return true;
}

void SubmissionContext::endDrawing(bool swapBuffers)
{
    if (swapBuffers)
        m_gl->swapBuffers(m_surface);
    if (m_ownCurrent)
        m_gl->doneCurrent();
    decayTextureScores();
    for (int i = 0; i < m_activeTextures.size(); ++i)
        if (m_activeTextures[i].texture)
            TextureExtRendererLocker::unlock(m_activeTextures[i].texture);
}

void SubmissionContext::activateRenderTarget(Qt3DCore::QNodeId renderTargetNodeId, const AttachmentPack &attachments, GLuint defaultFboId)
{
    GLuint fboId = defaultFboId; // Default FBO
    if (renderTargetNodeId) {
        // New RenderTarget
        if (!m_renderTargets.contains(renderTargetNodeId)) {
            if (m_defaultFBO && fboId == m_defaultFBO) {
                // this is the default fbo that some platforms create (iOS), we just register it
                // Insert FBO into hash
                m_renderTargets.insert(renderTargetNodeId, fboId);
            } else {
                fboId = createRenderTarget(renderTargetNodeId, attachments);
            }
        } else {
            fboId = updateRenderTarget(renderTargetNodeId, attachments, true);
        }
    }
    m_activeFBO = fboId;
    m_glHelper->bindFrameBufferObject(m_activeFBO, GraphicsHelperInterface::FBODraw);
    // Set active drawBuffers
    activateDrawBuffers(attachments);
}

GLuint SubmissionContext::createRenderTarget(Qt3DCore::QNodeId renderTargetNodeId, const AttachmentPack &attachments)
{
    const GLuint fboId = m_glHelper->createFrameBufferObject();
    if (fboId) {
        // The FBO is created and its attachments are set once
        // Insert FBO into hash
        m_renderTargets.insert(renderTargetNodeId, fboId);
        // Bind FBO
        m_glHelper->bindFrameBufferObject(fboId, GraphicsHelperInterface::FBODraw);
        bindFrameBufferAttachmentHelper(fboId, attachments);
    } else {
        qCritical("Failed to create FBO");
    }
    return fboId;
}

GLuint SubmissionContext::updateRenderTarget(Qt3DCore::QNodeId renderTargetNodeId, const AttachmentPack &attachments, bool isActiveRenderTarget)
{
    const GLuint fboId = m_renderTargets.value(renderTargetNodeId);

    // We need to check if  one of the attachment was resized
    bool needsResize = !m_renderTargetsSize.contains(fboId);    // not even initialized yet?
    if (!needsResize) {
        // render target exists, has attachment been resized?
        GLTextureManager *glTextureManager = m_renderer->nodeManagers()->glTextureManager();
        const QSize s = m_renderTargetsSize[fboId];
        const auto attachments_ = attachments.attachments();
        for (const Attachment &attachment : attachments_) {
            GLTexture *rTex = glTextureManager->lookupResource(attachment.m_textureUuid);
            // ### TODO QTBUG-64757 this check is insufficient since the
            // texture may have changed to another one with the same size. That
            // case is not handled atm.
            needsResize |= (rTex != nullptr && rTex->size() != s);
            if (isActiveRenderTarget) {
                if (attachment.m_point == QRenderTargetOutput::Color0)
                    m_renderTargetFormat = rTex->properties().format;
            }
        }
    }

    if (needsResize) {
        m_glHelper->bindFrameBufferObject(fboId, GraphicsHelperInterface::FBODraw);
        bindFrameBufferAttachmentHelper(fboId, attachments);
    }

    return fboId;
}

QSize SubmissionContext::renderTargetSize(const QSize &surfaceSize) const
{
    QSize renderTargetSize;
    if (m_activeFBO != m_defaultFBO) {
        // For external FBOs we may not have a m_renderTargets entry.
        if (m_renderTargetsSize.contains(m_activeFBO)) {
            renderTargetSize = m_renderTargetsSize[m_activeFBO];
        } else if (surfaceSize.isValid()) {
            renderTargetSize = surfaceSize;
        } else {
            // External FBO (when used with QtQuick2 Scene3D)

            // Query FBO color attachment 0 size
            GLint attachmentObjectType = GL_NONE;
            GLint attachment0Name = 0;
            m_gl->functions()->glGetFramebufferAttachmentParameteriv(GL_FRAMEBUFFER,
                                                                     GL_COLOR_ATTACHMENT0,
                                                                     GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE,
                                                                     &attachmentObjectType);
            m_gl->functions()->glGetFramebufferAttachmentParameteriv(GL_FRAMEBUFFER,
                                                                     GL_COLOR_ATTACHMENT0,
                                                                     GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME,
                                                                     &attachment0Name);

            if (attachmentObjectType == GL_RENDERBUFFER && m_glHelper->supportsFeature(GraphicsHelperInterface::RenderBufferDimensionRetrieval))
                renderTargetSize = m_glHelper->getRenderBufferDimensions(attachment0Name);
            else if (attachmentObjectType == GL_TEXTURE && m_glHelper->supportsFeature(GraphicsHelperInterface::TextureDimensionRetrieval))
                // Assumes texture level 0 and GL_TEXTURE_2D target
                renderTargetSize = m_glHelper->getTextureDimensions(attachment0Name, GL_TEXTURE_2D);
            else
                return renderTargetSize;
        }
    } else {
        renderTargetSize = m_surface->size();
        if (m_surface->surfaceClass() == QSurface::Window) {
            int dpr = static_cast<QWindow *>(m_surface)->devicePixelRatio();
            renderTargetSize *= dpr;
        }
    }
    return renderTargetSize;
}

QImage SubmissionContext::readFramebuffer(const QRect &rect)
{
    QImage img;
    const unsigned int area = rect.width() * rect.height();
    unsigned int bytes;
    GLenum format, type;
    QImage::Format imageFormat;
    uint stride;

    /* format value should match GL internalFormat */
    GLenum internalFormat = m_renderTargetFormat;

    switch (m_renderTargetFormat) {
    case QAbstractTexture::RGBAFormat:
    case QAbstractTexture::RGBA8_SNorm:
    case QAbstractTexture::RGBA8_UNorm:
    case QAbstractTexture::RGBA8U:
    case QAbstractTexture::SRGB8_Alpha8:
#ifdef QT_OPENGL_ES_2
        format = GL_RGBA;
        imageFormat = QImage::Format_RGBA8888_Premultiplied;
#else
        format = GL_BGRA;
        imageFormat = QImage::Format_ARGB32_Premultiplied;
        internalFormat = GL_RGBA8;
#endif
        type = GL_UNSIGNED_BYTE;
        bytes = area * 4;
        stride = rect.width() * 4;
        break;
    case QAbstractTexture::SRGB8:
    case QAbstractTexture::RGBFormat:
    case QAbstractTexture::RGB8U:
    case QAbstractTexture::RGB8_UNorm:
#ifdef QT_OPENGL_ES_2
        format = GL_RGBA;
        imageFormat = QImage::Format_RGBX8888;
#else
        format = GL_BGRA;
        imageFormat = QImage::Format_RGB32;
        internalFormat = GL_RGB8;
#endif
        type = GL_UNSIGNED_BYTE;
        bytes = area * 4;
        stride = rect.width() * 4;
        break;
#ifndef QT_OPENGL_ES_2
    case QAbstractTexture::RG11B10F:
        bytes = area * 4;
        format = GL_RGB;
        type = GL_UNSIGNED_INT_10F_11F_11F_REV;
        imageFormat = QImage::Format_RGB30;
        stride = rect.width() * 4;
        break;
    case QAbstractTexture::RGB10A2:
        bytes = area * 4;
        format = GL_RGBA;
        type = GL_UNSIGNED_INT_2_10_10_10_REV;
        imageFormat = QImage::Format_A2BGR30_Premultiplied;
        stride = rect.width() * 4;
        break;
    case QAbstractTexture::R5G6B5:
        bytes = area * 2;
        format = GL_RGB;
        type = GL_UNSIGNED_SHORT;
        internalFormat = GL_UNSIGNED_SHORT_5_6_5_REV;
        imageFormat = QImage::Format_RGB16;
        stride = rect.width() * 2;
        break;
    case QAbstractTexture::RGBA16F:
    case QAbstractTexture::RGBA16U:
    case QAbstractTexture::RGBA32F:
    case QAbstractTexture::RGBA32U:
        bytes = area * 16;
        format = GL_RGBA;
        type = GL_FLOAT;
        imageFormat = QImage::Format_ARGB32_Premultiplied;
        stride = rect.width() * 16;
        break;
#endif
    default:
        // unsupported format
        Q_UNREACHABLE();
        return img;
    }

    GLint samples = 0;
    m_gl->functions()->glGetIntegerv(GL_SAMPLES, &samples);
    if (samples > 0 && !m_glHelper->supportsFeature(GraphicsHelperInterface::BlitFramebuffer)) {
        qWarning () << Q_FUNC_INFO << "Unable to capture multisampled framebuffer; "
                       "Required feature BlitFramebuffer is missing.";
        return img;
    }

    img = QImage(rect.width(), rect.height(), imageFormat);

    QScopedArrayPointer<uchar> data(new uchar [bytes]);

    if (samples > 0) {
        // resolve multisample-framebuffer to renderbuffer and read pixels from it
        GLuint fbo, rb;
        QOpenGLFunctions *gl = m_gl->functions();
        gl->glGenFramebuffers(1, &fbo);
        gl->glBindFramebuffer(GL_DRAW_FRAMEBUFFER, fbo);
        gl->glGenRenderbuffers(1, &rb);
        gl->glBindRenderbuffer(GL_RENDERBUFFER, rb);
        gl->glRenderbufferStorage(GL_RENDERBUFFER, internalFormat, rect.width(), rect.height());
        gl->glFramebufferRenderbuffer(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER, rb);

        const GLenum status = gl->glCheckFramebufferStatus(GL_DRAW_FRAMEBUFFER);
        if (status != GL_FRAMEBUFFER_COMPLETE) {
            gl->glDeleteRenderbuffers(1, &rb);
            gl->glDeleteFramebuffers(1, &fbo);
            qWarning () << Q_FUNC_INFO << "Copy-framebuffer not complete: " << status;
            return img;
        }

        m_glHelper->blitFramebuffer(rect.x(), rect.y(), rect.x() + rect.width(), rect.y() + rect.height(),
                                    0, 0, rect.width(), rect.height(),
                                    GL_COLOR_BUFFER_BIT, GL_NEAREST);
        gl->glBindFramebuffer(GL_READ_FRAMEBUFFER, fbo);
        gl->glReadPixels(0,0,rect.width(), rect.height(), format, type, data.data());

        copyGLFramebufferDataToImage(img, data.data(), stride, rect.width(), rect.height(), m_renderTargetFormat);

        gl->glBindRenderbuffer(GL_RENDERBUFFER, rb);
        gl->glDeleteRenderbuffers(1, &rb);
        gl->glBindFramebuffer(GL_FRAMEBUFFER, m_activeFBO);
        gl->glDeleteFramebuffers(1, &fbo);
    } else {
        // read pixels directly from framebuffer
        m_gl->functions()->glReadPixels(rect.x(), rect.y(), rect.width(), rect.height(), format, type, data.data());
        copyGLFramebufferDataToImage(img, data.data(), stride, rect.width(), rect.height(), m_renderTargetFormat);
    }

    return img;
}

void SubmissionContext::setViewport(const QRectF &viewport, const QSize &surfaceSize)
{
    //    // save for later use; this has nothing to do with the viewport but it is
    //    // here that we get to know the surfaceSize from the RenderView.
        m_surfaceSize = surfaceSize;

    m_viewport = viewport;
    QSize size = renderTargetSize(surfaceSize);

    // Check that the returned size is before calling glViewport
    if (size.isEmpty())
        return;

    // Qt3D 0------------------> 1  OpenGL  1^
    //      |                                |
    //      |                                |
    //      |                                |
    //      V                                |
    //      1                                0---------------------> 1
    // The Viewport is defined between 0 and 1 which allows us to automatically
    // scale to the size of the provided window surface
    m_gl->functions()->glViewport(m_viewport.x() * size.width(),
                                  (1.0 - m_viewport.y() - m_viewport.height()) * size.height(),
                                  m_viewport.width() * size.width(),
                                  m_viewport.height() * size.height());
}

void SubmissionContext::releaseOpenGL()
{
    m_shaderCache->clear();
    m_renderBufferHash.clear();

    // Stop and destroy the OpenGL logger
    if (m_debugLogger) {
        m_debugLogger->stopLogging();
        m_debugLogger.reset(nullptr);
    }
}

// The OpenGLContext is not current on any surface at this point
void SubmissionContext::setOpenGLContext(QOpenGLContext* ctx)
{
    Q_ASSERT(ctx && m_shaderCache);

    releaseOpenGL();
    m_gl = ctx;
}

void SubmissionContext::activateGLHelper()
{
    // Sets the correct GL Helper depending on the surface
    // If no helper exists, create one
    m_glHelper = m_glHelpers.value(m_surface);
    if (!m_glHelper) {
        m_glHelper = resolveHighestOpenGLFunctions();
        m_glHelpers.insert(m_surface, m_glHelper);
        // Note: OpenGLContext is current at this point
        m_gl->functions()->glDisable(GL_DITHER);
    }
}


// Called only from RenderThread
bool SubmissionContext::activateShader(ProgramDNA shaderDNA)
{
    if (shaderDNA != m_activeShaderDNA) {
        // Ensure material uniforms are re-applied
        m_material = nullptr;

        m_activeShader = m_shaderCache->getShaderProgramForDNA(shaderDNA);
        if (Q_LIKELY(m_activeShader != nullptr)) {
            m_activeShader->bind();
            m_activeShaderDNA = shaderDNA;
        } else {
            m_glHelper->useProgram(0);
            qWarning() << "No shader program found for DNA";
            m_activeShaderDNA = 0;
            return false;
        }
    }
    return true;
}

void SubmissionContext::bindFrameBufferAttachmentHelper(GLuint fboId, const AttachmentPack &attachments)
{
    // Set FBO attachments. These are normally textures, except that on Open GL
    // ES <= 3.1 we must use a renderbuffer if a combined depth+stencil is
    // desired since this cannot be achieved neither with a single texture (not
    // before GLES 3.2) nor with separate textures (no suitable format for
    // stencil before 3.1 with the appropriate extension).

    QSize fboSize;
    GLTextureManager *glTextureManager = m_renderer->nodeManagers()->glTextureManager();
    const auto attachments_ = attachments.attachments();
    for (const Attachment &attachment : attachments_) {
        GLTexture *rTex = glTextureManager->lookupResource(attachment.m_textureUuid);
        if (!m_glHelper->frameBufferNeedsRenderBuffer(attachment)) {
            QOpenGLTexture *glTex = rTex ? rTex->getGLTexture() : nullptr;
            if (glTex != nullptr) {
                // The texture can not be rendered simultaniously by another renderer
                Q_ASSERT(!rTex->isExternalRenderingEnabled());
                if (fboSize.isEmpty())
                    fboSize = QSize(glTex->width(), glTex->height());
                else
                    fboSize = QSize(qMin(fboSize.width(), glTex->width()), qMin(fboSize.height(), glTex->height()));
                m_glHelper->bindFrameBufferAttachment(glTex, attachment);
            }
        } else {
            RenderBuffer *renderBuffer = rTex ? rTex->getOrCreateRenderBuffer() : nullptr;
            if (renderBuffer) {
                if (fboSize.isEmpty())
                    fboSize = QSize(renderBuffer->width(), renderBuffer->height());
                else
                    fboSize = QSize(qMin(fboSize.width(), renderBuffer->width()), qMin(fboSize.height(), renderBuffer->height()));
                m_glHelper->bindFrameBufferAttachment(renderBuffer, attachment);
            }
        }
    }
    m_renderTargetsSize.insert(fboId, fboSize);
}

void SubmissionContext::activateDrawBuffers(const AttachmentPack &attachments)
{
    const QVector<int> activeDrawBuffers = attachments.getGlDrawBuffers();

    if (m_glHelper->checkFrameBufferComplete()) {
        if (activeDrawBuffers.size() > 1) {// We need MRT
            if (m_glHelper->supportsFeature(GraphicsHelperInterface::MRT)) {
                // Set up MRT, glDrawBuffers...
                m_glHelper->drawBuffers(activeDrawBuffers.size(), activeDrawBuffers.data());
            }
        }
    } else {
        qWarning() << "FBO incomplete";
    }
}


void SubmissionContext::setActiveMaterial(Material *rmat)
{
    if (m_material == rmat)
        return;

    deactivateTexturesWithScope(TextureScopeMaterial);
    m_material = rmat;
}

int SubmissionContext::activateTexture(TextureScope scope, GLTexture *tex, int onUnit)
{
    // Returns the texture unit to use for the texture
    // This always return a valid unit, unless there are more textures than
    // texture unit available for the current material
    onUnit = assignUnitForTexture(tex);

    // check we didn't overflow the available units
    if (onUnit == -1)
        return -1;

    // actually re-bind if required, the tex->dna on the unit not being the same
    // Note: tex->dna() could be 0 if the texture has not been created yet
    if (m_activeTextures[onUnit].texture != tex) {
        // Texture must have been created and updated at this point
        QOpenGLTexture *glTex = tex->getGLTexture();
        if (glTex == nullptr)
            return -1;
        glTex->bind(onUnit);
        if (m_activeTextures[onUnit].texture)
            TextureExtRendererLocker::unlock(m_activeTextures[onUnit].texture);
        m_activeTextures[onUnit].texture = tex;
        TextureExtRendererLocker::lock(tex);
    }

#if defined(QT3D_RENDER_ASPECT_OPENGL_DEBUG)
    int err = m_gl->functions()->glGetError();
    if (err)
        qCWarning(Backend) << "GL error after activating texture" << QString::number(err, 16)
                           << tex->getGLTexture()->textureId() << "on unit" << onUnit;
#endif

    m_activeTextures[onUnit].score = 200;
    m_activeTextures[onUnit].pinned = true;
    m_activeTextures[onUnit].scope = scope;

    return onUnit;
}

void SubmissionContext::deactivateTexturesWithScope(TextureScope ts)
{
    for (int u=0; u<m_activeTextures.size(); ++u) {
        if (!m_activeTextures[u].pinned)
            continue; // inactive, ignore

        if (m_activeTextures[u].scope == ts) {
            m_activeTextures[u].pinned = false;
            m_activeTextures[u].score = qMax(m_activeTextures[u].score, 1) - 1;
        }
    } // of units iteration
}

void SubmissionContext::deactivateTexture(GLTexture* tex)
{
    for (int u=0; u<m_activeTextures.size(); ++u) {
        if (m_activeTextures[u].texture == tex) {
            Q_ASSERT(m_activeTextures[u].pinned);
            m_activeTextures[u].pinned = false;
            return;
        }
    } // of units iteration

    qCWarning(Backend) << Q_FUNC_INFO << "texture not active:" << tex;
}

/*!
    \internal
    Returns a texture unit for a texture, -1 if all texture units are assigned.
    Tries to use the texture unit with the texture that hasn't been used for the longest time
    if the texture happens not to be already pinned on a texture unit.
 */
GLint SubmissionContext::assignUnitForTexture(GLTexture *tex)
{
    int lowestScoredUnit = -1;
    int lowestScore = 0xfffffff;

    for (int u=0; u<m_activeTextures.size(); ++u) {
        if (m_activeTextures[u].texture == tex)
            return u;

        // No texture is currently active on the texture unit
        // we save the texture unit with the texture that has been on there
        // the longest time while not being used
        if (!m_activeTextures[u].pinned) {
            int score = m_activeTextures[u].score;
            if (score < lowestScore) {
                lowestScore = score;
                lowestScoredUnit = u;
            }
        }
    } // of units iteration

    if (lowestScoredUnit == -1)
        qCWarning(Backend) << Q_FUNC_INFO << "No free texture units!";

    return lowestScoredUnit;
}

void SubmissionContext::decayTextureScores()
{
    for (int u = 0; u < m_activeTextures.size(); u++)
        m_activeTextures[u].score = qMax(m_activeTextures[u].score - 1, 0);
}

void SubmissionContext::setCurrentStateSet(RenderStateSet *ss)
{
    if (ss == m_stateSet)
        return;
    if (ss)
        applyStateSet(ss);
    m_stateSet = ss;
}

RenderStateSet *SubmissionContext::currentStateSet() const
{
    return m_stateSet;
}

void SubmissionContext::applyState(const StateVariant &stateVariant)
{
    switch (stateVariant.type) {

    case AlphaCoverageStateMask: {
        applyStateHelper<AlphaCoverage>(static_cast<const AlphaCoverage *>(stateVariant.constState()), this);
        break;
    }
    case AlphaTestMask: {
        applyStateHelper<AlphaFunc>(static_cast<const AlphaFunc *>(stateVariant.constState()), this);
        break;
    }
    case BlendStateMask: {
        applyStateHelper<BlendEquation>(static_cast<const BlendEquation *>(stateVariant.constState()), this);
        break;
    }
    case BlendEquationArgumentsMask: {
        applyStateHelper<BlendEquationArguments>(static_cast<const BlendEquationArguments *>(stateVariant.constState()), this);
        break;
    }
    case MSAAEnabledStateMask: {
        applyStateHelper<MSAAEnabled>(static_cast<const MSAAEnabled *>(stateVariant.constState()), this);
        break;
    }

    case CullFaceStateMask: {
        applyStateHelper<CullFace>(static_cast<const CullFace *>(stateVariant.constState()), this);
        break;
    }

    case DepthWriteStateMask: {
        applyStateHelper<NoDepthMask>(static_cast<const NoDepthMask *>(stateVariant.constState()), this);
        break;
    }

    case DepthTestStateMask: {
        applyStateHelper<DepthTest>(static_cast<const DepthTest *>(stateVariant.constState()), this);
        break;
    }

    case FrontFaceStateMask: {
        applyStateHelper<FrontFace>(static_cast<const FrontFace *>(stateVariant.constState()), this);
        break;
    }

    case ScissorStateMask: {
        applyStateHelper<ScissorTest>(static_cast<const ScissorTest *>(stateVariant.constState()), this);
        break;
    }

    case StencilTestStateMask: {
        applyStateHelper<StencilTest>(static_cast<const StencilTest *>(stateVariant.constState()), this);
        break;
    }

    case PointSizeMask: {
        applyStateHelper<PointSize>(static_cast<const PointSize *>(stateVariant.constState()), this);
        break;
    }

    case PolygonOffsetStateMask: {
        applyStateHelper<PolygonOffset>(static_cast<const PolygonOffset *>(stateVariant.constState()), this);
        break;
    }

    case ColorStateMask: {
        applyStateHelper<ColorMask>(static_cast<const ColorMask *>(stateVariant.constState()), this);
        break;
    }

    case ClipPlaneMask: {
        applyStateHelper<ClipPlane>(static_cast<const ClipPlane *>(stateVariant.constState()), this);
        break;
    }

    case SeamlessCubemapMask: {
        applyStateHelper<SeamlessCubemap>(static_cast<const SeamlessCubemap *>(stateVariant.constState()), this);
        break;
    }

    case StencilOpMask: {
        applyStateHelper<StencilOp>(static_cast<const StencilOp *>(stateVariant.constState()), this);
        break;
    }

    case StencilWriteStateMask: {
        applyStateHelper<StencilMask>(static_cast<const StencilMask *>(stateVariant.constState()), this);
        break;
    }

    case DitheringStateMask: {
        applyStateHelper<Dithering>(static_cast<const Dithering *>(stateVariant.constState()), this);
        break;
    }

    case LineWidthMask: {
        applyStateHelper<LineWidth>(static_cast<const LineWidth *>(stateVariant.constState()), this);
        break;
    }
    default:
        Q_UNREACHABLE();
    }
}

void SubmissionContext::resetMasked(qint64 maskOfStatesToReset)
{
    // TO DO -> Call gcHelper methods instead of raw GL
    // QOpenGLFunctions shouldn't be used here directly
    QOpenGLFunctions *funcs = m_gl->functions();

    if (maskOfStatesToReset & ScissorStateMask)
        funcs->glDisable(GL_SCISSOR_TEST);

    if (maskOfStatesToReset & BlendStateMask)
        funcs->glDisable(GL_BLEND);

    if (maskOfStatesToReset & StencilWriteStateMask)
        funcs->glStencilMask(0);

    if (maskOfStatesToReset & StencilTestStateMask)
        funcs->glDisable(GL_STENCIL_TEST);

    if (maskOfStatesToReset & DepthTestStateMask)
        funcs->glDisable(GL_DEPTH_TEST);

    if (maskOfStatesToReset & DepthWriteStateMask)
        funcs->glDepthMask(GL_TRUE); // reset to default

    if (maskOfStatesToReset & FrontFaceStateMask)
        funcs->glFrontFace(GL_CCW); // reset to default

    if (maskOfStatesToReset & CullFaceStateMask)
        funcs->glDisable(GL_CULL_FACE);

    if (maskOfStatesToReset & DitheringStateMask)
        funcs->glDisable(GL_DITHER);

    if (maskOfStatesToReset & AlphaCoverageStateMask)
        setAlphaCoverageEnabled(false);

    if (maskOfStatesToReset & PointSizeMask)
        pointSize(false, 1.0f);    // reset to default

    if (maskOfStatesToReset & PolygonOffsetStateMask)
        funcs->glDisable(GL_POLYGON_OFFSET_FILL);

    if (maskOfStatesToReset & ColorStateMask)
        funcs->glColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);

    if (maskOfStatesToReset & ClipPlaneMask) {
        GLint max = maxClipPlaneCount();
        for (GLint i = 0; i < max; ++i)
            disableClipPlane(i);
    }

    if (maskOfStatesToReset & SeamlessCubemapMask)
        setSeamlessCubemap(false);

    if (maskOfStatesToReset & StencilOpMask)
        funcs->glStencilOp(GL_KEEP, GL_KEEP, GL_KEEP);

    if (maskOfStatesToReset & LineWidthMask)
        funcs->glLineWidth(1.0f);
}

void SubmissionContext::applyStateSet(RenderStateSet *ss)
{
    RenderStateSet* previousStates = currentStateSet();

    const StateMaskSet invOurState = ~ss->stateMask();
    // generate a mask for each set bit in previous, where we do not have
    // the corresponding bit set.

    StateMaskSet stateToReset = 0;
    if (previousStates) {
        stateToReset = previousStates->stateMask() & invOurState;
        qCDebug(RenderStates) << "previous states " << QString::number(previousStates->stateMask(), 2);
    }
    qCDebug(RenderStates) << " current states " << QString::number(ss->stateMask(), 2)  << "inverse " << QString::number(invOurState, 2) << " -> states to change:  " << QString::number(stateToReset, 2);

    // Reset states that aren't active in the current state set
    resetMasked(stateToReset);

    // Apply states that weren't in the previous state or that have
    // different values
    const QVector<StateVariant> statesToSet = ss->states();
    for (const StateVariant &ds : statesToSet) {
        if (previousStates && previousStates->contains(ds))
            continue;
        applyState(ds);
    }
}

void SubmissionContext::clearColor(const QColor &color)
{
    if (m_currClearColorValue != color) {
        m_currClearColorValue = color;
        m_gl->functions()->glClearColor(color.redF(), color.greenF(), color.blueF(), color.alphaF());
    }
}

void SubmissionContext::clearDepthValue(float depth)
{
    if (m_currClearDepthValue != depth) {
        m_currClearDepthValue = depth;
        m_gl->functions()->glClearDepthf(depth);
    }
}

void SubmissionContext::clearStencilValue(int stencil)
{
    if (m_currClearStencilValue != stencil) {
        m_currClearStencilValue = stencil;
        m_gl->functions()->glClearStencil(stencil);
    }
}

// It will be easier if the QGraphicContext applies the QUniformPack
// than the other way around
bool SubmissionContext::setParameters(ShaderParameterPack &parameterPack)
{
    static const int irradianceId = StringToInt::lookupId(QLatin1String("envLight.irradiance"));
    static const int specularId = StringToInt::lookupId(QLatin1String("envLight.specular"));
    // Activate textures and update TextureUniform in the pack
    // with the correct textureUnit

    // Set the pinned texture of the previous material texture
    // to pinable so that we should easily find an available texture unit
    NodeManagers *manager = m_renderer->nodeManagers();

    deactivateTexturesWithScope(TextureScopeMaterial);
    // Update the uniforms with the correct texture unit id's
    PackUniformHash &uniformValues = parameterPack.uniforms();

    for (int i = 0; i < parameterPack.textures().size(); ++i) {
        const ShaderParameterPack::NamedTexture &namedTex = parameterPack.textures().at(i);
        // Given a Texture QNodeId, we retrieve the associated shared GLTexture
        if (uniformValues.contains(namedTex.glslNameId)) {
            GLTexture *t = manager->glTextureManager()->lookupResource(namedTex.texId);
            if (t != nullptr) {
                UniformValue &texUniform = uniformValues[namedTex.glslNameId];
                Q_ASSERT(texUniform.valueType() == UniformValue::TextureValue);
                const int texUnit = activateTexture(TextureScopeMaterial, t);
                texUniform.data<int>()[namedTex.uniformArrayIndex] = texUnit;
                if (texUnit == -1) {
                    if (namedTex.glslNameId != irradianceId &&
                        namedTex.glslNameId != specularId) {
                        // Only return false if we are not dealing with env light textures
                        return false;
                    }
                }
            }
        }
    }

    QOpenGLShaderProgram *shader = activeShader();

    // TO DO: We could cache the binding points somehow and only do the binding when necessary
    // for SSBO and UBO

    // Bind Shader Storage block to SSBO and update SSBO
    const QVector<BlockToSSBO> blockToSSBOs = parameterPack.shaderStorageBuffers();
    int ssboIndex = 0;
    for (const BlockToSSBO b : blockToSSBOs) {
        Buffer *cpuBuffer = m_renderer->nodeManagers()->bufferManager()->lookupResource(b.m_bufferID);
        GLBuffer *ssbo = glBufferForRenderBuffer(cpuBuffer, GLBuffer::ShaderStorageBuffer);
        bindShaderStorageBlock(shader->programId(), b.m_blockIndex, ssboIndex);
        // Needed to avoid conflict where the buffer would already
        // be bound as a VertexArray
        bindGLBuffer(ssbo, GLBuffer::ShaderStorageBuffer);
        ssbo->bindBufferBase(this, ssboIndex++, GLBuffer::ShaderStorageBuffer);
        // TO DO: Make sure that there's enough binding points
    }

    // Bind UniformBlocks to UBO and update UBO from Buffer
    // TO DO: Convert ShaderData to Buffer so that we can use that generic process
    const QVector<BlockToUBO> blockToUBOs = parameterPack.uniformBuffers();
    int uboIndex = 0;
    for (const BlockToUBO &b : blockToUBOs) {
        Buffer *cpuBuffer = m_renderer->nodeManagers()->bufferManager()->lookupResource(b.m_bufferID);
        GLBuffer *ubo = glBufferForRenderBuffer(cpuBuffer, GLBuffer::UniformBuffer);
        bindUniformBlock(shader->programId(), b.m_blockIndex, uboIndex);
        // Needed to avoid conflict where the buffer would already
        // be bound as a VertexArray
        bindGLBuffer(ubo, GLBuffer::UniformBuffer);
        ubo->bindBufferBase(this, uboIndex++, GLBuffer::UniformBuffer);
        // TO DO: Make sure that there's enough binding points
    }

    // Update uniforms in the Default Uniform Block
    const PackUniformHash values = parameterPack.uniforms();
    const QVector<ShaderUniform> activeUniforms = parameterPack.submissionUniforms();

    for (const ShaderUniform &uniform : activeUniforms) {
        // We can use [] as we are sure the the uniform wouldn't
        // be un activeUniforms if there wasn't a matching value
        const UniformValue &v = values[uniform.m_nameId];

        // skip invalid textures
        if (v.valueType() == UniformValue::TextureValue && *v.constData<int>() == -1)
            continue;

        applyUniform(uniform, v);
    }
    // if not all data is valid, the next frame will be rendered immediately
    return true;
}

void SubmissionContext::enableAttribute(const VAOVertexAttribute &attr)
{
    // Bind buffer within the current VAO
    GLBuffer *buf = m_renderer->nodeManagers()->glBufferManager()->data(attr.bufferHandle);
    Q_ASSERT(buf);
    bindGLBuffer(buf, attr.attributeType);

    // Don't use QOpenGLShaderProgram::setAttributeBuffer() because of QTBUG-43199.
    // Use the introspection data and set the attribute explicitly
    m_glHelper->enableVertexAttributeArray(attr.location);
    m_glHelper->vertexAttributePointer(attr.shaderDataType,
                                       attr.location,
                                       attr.vertexSize,
                                       attr.dataType,
                                       GL_TRUE, // TODO: Support normalization property on QAttribute
                                       attr.byteStride,
                                       reinterpret_cast<const void *>(qintptr(attr.byteOffset)));


    // Done by the helper if it supports it
    if (attr.divisor != 0)
        m_glHelper->vertexAttribDivisor(attr.location, attr.divisor);
}

void SubmissionContext::disableAttribute(const SubmissionContext::VAOVertexAttribute &attr)
{
    QOpenGLShaderProgram *prog = activeShader();
    prog->disableAttributeArray(attr.location);
}

// Note: needs to be called while VAO is bound
void SubmissionContext::specifyAttribute(const Attribute *attribute,
                                       Buffer *buffer,
                                       const ShaderAttribute *attributeDescription)
{
    const int location = attributeDescription->m_location;
    if (location < 0) {
        qCWarning(Backend) << "failed to resolve location for attribute:" << attribute->name();
        return;
    }

    const GLint attributeDataType = glDataTypeFromAttributeDataType(attribute->vertexBaseType());
    const HGLBuffer glBufferHandle = m_renderer->nodeManagers()->glBufferManager()->lookupHandle(buffer->peerId());
    Q_ASSERT(!glBufferHandle.isNull());
    const GLBuffer::Type attributeType = attributeTypeToGLBufferType(attribute->attributeType());

    int typeSize = 0;
    int attrCount = 0;

    if (attribute->vertexSize() >= 1 && attribute->vertexSize() <= 4) {
        attrCount = 1;
    } else if (attribute->vertexSize() == 9) {
        typeSize = byteSizeFromType(attributeDataType);
        attrCount = 3;
    } else if (attribute->vertexSize() == 16) {
        typeSize = byteSizeFromType(attributeDataType);
        attrCount = 4;
    } else {
        Q_UNREACHABLE();
    }

    VAOVertexAttribute attr;
    attr.bufferHandle = glBufferHandle;
    attr.attributeType = attributeType;
    attr.dataType = attributeDataType;
    attr.divisor = attribute->divisor();
    attr.vertexSize = attribute->vertexSize() / attrCount;
    attr.byteStride = (attribute->byteStride() != 0) ? attribute->byteStride() : (attrCount * attrCount * typeSize);
    attr.shaderDataType = attributeDescription->m_type;

    for (int i = 0; i < attrCount; i++) {
        attr.location = location + i;
        attr.byteOffset = attribute->byteOffset() + (i * attrCount * typeSize);

        enableAttribute(attr);

        // Save this in the current emulated VAO
        if (m_currentVAO)
            m_currentVAO->saveVertexAttribute(attr);
    }
}

void SubmissionContext::specifyIndices(Buffer *buffer)
{
    GLBuffer *buf = glBufferForRenderBuffer(buffer, GLBuffer::IndexBuffer);
    if (!bindGLBuffer(buf, GLBuffer::IndexBuffer))
        qCWarning(Backend) << Q_FUNC_INFO << "binding index buffer failed";

    // bound within the current VAO
    // Save this in the current emulated VAO
    if (m_currentVAO)
        m_currentVAO->saveIndexAttribute(m_renderer->nodeManagers()->glBufferManager()->lookupHandle(buffer->peerId()));
}

void SubmissionContext::updateBuffer(Buffer *buffer)
{
    const QHash<Qt3DCore::QNodeId, HGLBuffer>::iterator it = m_renderBufferHash.find(buffer->peerId());
    if (it != m_renderBufferHash.end())
        uploadDataToGLBuffer(buffer, m_renderer->nodeManagers()->glBufferManager()->data(it.value()));
}

QByteArray SubmissionContext::downloadBufferContent(Buffer *buffer)
{
    const QHash<Qt3DCore::QNodeId, HGLBuffer>::iterator it = m_renderBufferHash.find(buffer->peerId());
    if (it != m_renderBufferHash.end())
        return downloadDataFromGLBuffer(buffer, m_renderer->nodeManagers()->glBufferManager()->data(it.value()));
    return QByteArray();
}

void SubmissionContext::releaseBuffer(Qt3DCore::QNodeId bufferId)
{
    auto it = m_renderBufferHash.find(bufferId);
    if (it != m_renderBufferHash.end()) {
        HGLBuffer glBuffHandle = it.value();
        GLBuffer *glBuff = m_renderer->nodeManagers()->glBufferManager()->data(glBuffHandle);

        Q_ASSERT(glBuff);
        // Destroy the GPU resource
        glBuff->destroy(this);
        // Destroy the GLBuffer instance
        m_renderer->nodeManagers()->glBufferManager()->releaseResource(bufferId);
        // Remove Id - HGLBuffer entry
        m_renderBufferHash.erase(it);
    }
}

bool SubmissionContext::hasGLBufferForBuffer(Buffer *buffer)
{
    const QHash<Qt3DCore::QNodeId, HGLBuffer>::iterator it = m_renderBufferHash.find(buffer->peerId());
    return (it != m_renderBufferHash.end());
}

GLBuffer *SubmissionContext::glBufferForRenderBuffer(Buffer *buf, GLBuffer::Type type)
{
    if (!m_renderBufferHash.contains(buf->peerId()))
        m_renderBufferHash.insert(buf->peerId(), createGLBufferFor(buf, type));
    return m_renderer->nodeManagers()->glBufferManager()->data(m_renderBufferHash.value(buf->peerId()));
}

HGLBuffer SubmissionContext::createGLBufferFor(Buffer *buffer, GLBuffer::Type type)
{
    GLBuffer *b = m_renderer->nodeManagers()->glBufferManager()->getOrCreateResource(buffer->peerId());
    //    b.setUsagePattern(static_cast<QOpenGLBuffer::UsagePattern>(buffer->usage()));
    Q_ASSERT(b);
    if (!b->create(this))
        qCWarning(Render::Io) << Q_FUNC_INFO << "buffer creation failed";

    if (!bindGLBuffer(b, type))
        qCWarning(Render::Io) << Q_FUNC_INFO << "buffer binding failed";

    return m_renderer->nodeManagers()->glBufferManager()->lookupHandle(buffer->peerId());
}

bool SubmissionContext::bindGLBuffer(GLBuffer *buffer, GLBuffer::Type type)
{
    if (type == GLBuffer::ArrayBuffer && buffer == m_boundArrayBuffer)
        return true;

    if (buffer->bind(this, type)) {
        if (type == GLBuffer::ArrayBuffer)
            m_boundArrayBuffer = buffer;
        return true;
    }
    return false;
}

void SubmissionContext::uploadDataToGLBuffer(Buffer *buffer, GLBuffer *b, bool releaseBuffer)
{
    if (!bindGLBuffer(b, GLBuffer::ArrayBuffer)) // We're uploading, the type doesn't matter here
        qCWarning(Render::Io) << Q_FUNC_INFO << "buffer bind failed";
    // If the buffer is dirty (hence being called here)
    // there are two possible cases
    // * setData was called changing the whole data or functor (or the usage pattern)
    // * partial buffer updates where received

    // TO DO: Handle usage pattern
    QVector<Qt3DRender::QBufferUpdate> updates = std::move(buffer->pendingBufferUpdates());
    for (auto it = updates.begin(); it != updates.end(); ++it) {
        auto update = it;
        // We have a partial update
        if (update->offset >= 0) {
            //accumulate sequential updates as single one
            int bufferSize = update->data.size();
            auto it2 = it + 1;
            while ((it2 != updates.end())
                   && (it2->offset - update->offset == bufferSize)) {
                bufferSize += it2->data.size();
                ++it2;
            }
            update->data.resize(bufferSize);
            while (it + 1 != it2) {
                ++it;
                update->data.replace(it->offset - update->offset, it->data.size(), it->data);
                it->data.clear();
            }
            // TO DO: based on the number of updates .., it might make sense to
            // sometime use glMapBuffer rather than glBufferSubData
            b->update(this, update->data.constData(), update->data.size(), update->offset);
        } else {
            // We have an update that was done by calling QBuffer::setData
            // which is used to resize or entirely clear the buffer
            // Note: we use the buffer data directly in that case
            const int bufferSize = buffer->data().size();
            b->allocate(this, bufferSize, false); // orphan the buffer
            b->allocate(this, buffer->data().constData(), bufferSize, false);
        }
    }

    if (releaseBuffer) {
        b->release(this);
        m_boundArrayBuffer = nullptr;
    }
    qCDebug(Render::Io) << "uploaded buffer size=" << buffer->data().size();
}

QByteArray SubmissionContext::downloadDataFromGLBuffer(Buffer *buffer, GLBuffer *b)
{
    if (!bindGLBuffer(b, GLBuffer::ArrayBuffer)) // We're downloading, the type doesn't matter here
        qCWarning(Render::Io) << Q_FUNC_INFO << "buffer bind failed";

    QByteArray data = b->download(this, buffer->data().size());
    return data;
}

void SubmissionContext::blitFramebuffer(Qt3DCore::QNodeId inputRenderTargetId,
                                        Qt3DCore::QNodeId outputRenderTargetId,
                                        QRect inputRect, QRect outputRect,
                                        uint defaultFboId,
                                        QRenderTargetOutput::AttachmentPoint inputAttachmentPoint,
                                        QRenderTargetOutput::AttachmentPoint outputAttachmentPoint,
                                        QBlitFramebuffer::InterpolationMethod interpolationMethod)
{
    GLuint inputFboId = defaultFboId;
    bool inputBufferIsDefault = true;
    if (!inputRenderTargetId.isNull()) {
        RenderTarget *renderTarget = m_renderer->nodeManagers()->renderTargetManager()->lookupResource(inputRenderTargetId);
        if (renderTarget) {
            AttachmentPack attachments(renderTarget, m_renderer->nodeManagers()->attachmentManager());
            if (m_renderTargets.contains(inputRenderTargetId))
                inputFboId = updateRenderTarget(inputRenderTargetId, attachments, false);
            else
                inputFboId = createRenderTarget(inputRenderTargetId, attachments);
        }
        inputBufferIsDefault = false;
    }

    GLuint outputFboId = defaultFboId;
    bool outputBufferIsDefault = true;
    if (!outputRenderTargetId.isNull()) {
        RenderTarget *renderTarget = m_renderer->nodeManagers()->renderTargetManager()->lookupResource(outputRenderTargetId);
        if (renderTarget) {
            AttachmentPack attachments(renderTarget, m_renderer->nodeManagers()->attachmentManager());
            if (m_renderTargets.contains(outputRenderTargetId))
                outputFboId = updateRenderTarget(outputRenderTargetId, attachments, false);
            else
                outputFboId = createRenderTarget(outputRenderTargetId, attachments);
        }
        outputBufferIsDefault = false;
    }

    // Up until this point the input and output rects are normal Qt rectangles.
    // Convert them to GL rectangles (Y at bottom).
    const int inputFboHeight = inputFboId == defaultFboId ? m_surfaceSize.height() : m_renderTargetsSize[inputFboId].height();
    const GLint srcX0 = inputRect.left();
    const GLint srcY0 = inputFboHeight - (inputRect.top() + inputRect.height());
    const GLint srcX1 = srcX0 + inputRect.width();
    const GLint srcY1 = srcY0 + inputRect.height();

    const int outputFboHeight = outputFboId == defaultFboId ? m_surfaceSize.height() : m_renderTargetsSize[outputFboId].height();
    const GLint dstX0 = outputRect.left();
    const GLint dstY0 = outputFboHeight - (outputRect.top() + outputRect.height());
    const GLint dstX1 = dstX0 + outputRect.width();
    const GLint dstY1 = dstY0 + outputRect.height();

    //Get the last bounded framebuffers
    const GLuint lastDrawFboId = boundFrameBufferObject();

    // Activate input framebuffer for reading
    bindFramebuffer(inputFboId, GraphicsHelperInterface::FBORead);

    // Activate output framebuffer for writing
    bindFramebuffer(outputFboId, GraphicsHelperInterface::FBODraw);

    //Bind texture
    if (!inputBufferIsDefault)
        readBuffer(GL_COLOR_ATTACHMENT0 + inputAttachmentPoint);

    if (!outputBufferIsDefault) {
        // Note that we use glDrawBuffers, not glDrawBuffer. The
        // latter is not available with GLES.
        const int buf = outputAttachmentPoint;
        drawBuffers(1, &buf);
    }

    // Blit framebuffer
    const GLenum mode = interpolationMethod ? GL_NEAREST : GL_LINEAR;
    m_glHelper->blitFramebuffer(srcX0, srcY0, srcX1, srcY1,
                                dstX0, dstY0, dstX1, dstY1,
                                GL_COLOR_BUFFER_BIT, mode);

    // Reset draw buffer
    bindFramebuffer(lastDrawFboId, GraphicsHelperInterface::FBOReadAndDraw);
    if (outputAttachmentPoint != QRenderTargetOutput::Color0) {
        const int buf = QRenderTargetOutput::Color0;
        drawBuffers(1, &buf);
    }
}

} // namespace Render
} // namespace Qt3DRender of namespace

QT_END_NAMESPACE