void PhysXRigidbody::updateMassDistribution()
{
if (((UINT32)mFlags & (UINT32)Flag::AutoTensors) == 0)
return;
if (((UINT32)mFlags & (UINT32)Flag::AutoMass) == 0)
{
PxRigidBodyExt::setMassAndUpdateInertia(*mInternal, mInternal->getMass());
}
else
{
UINT32 numShapes = mInternal->getNbShapes();
if (numShapes == 0)
{
PxRigidBodyExt::setMassAndUpdateInertia(*mInternal, mInternal->getMass());
return;
}
PxShape** shapes = (PxShape**)bs_stack_alloc(sizeof(PxShape*) * numShapes);
mInternal->getShapes(shapes, numShapes);
float* masses = (float*)bs_stack_alloc(sizeof(float) * numShapes);
for (UINT32 i = 0; i < numShapes; i++)
masses[i] = ((Collider*)shapes[i]->userData)->getMass();
PxRigidBodyExt::setMassAndUpdateInertia(*mInternal, masses, numShapes);
bs_stack_free(masses);
bs_stack_free(shapes);
}
}
void GUIPanel::_updateLayoutInternal(const GUILayoutData& data)
{
GUILayoutData childData = data;
_updateDepthRange(childData);
UINT32 numElements = (UINT32)mChildren.size();
Rect2I* elementAreas = nullptr;
if (numElements > 0)
elementAreas = bs_stack_new<Rect2I>(numElements);
_getElementAreas(data.area, elementAreas, numElements, mChildSizeRanges, mSizeRange);
UINT32 childIdx = 0;
for (auto& child : mChildren)
{
if (child->_isActive())
{
childData.area = elementAreas[childIdx];
_updateChildLayout(child, childData);
}
childIdx++;
}
if (elementAreas != nullptr)
bs_stack_free(elementAreas);
}
void GUILayoutX::_updateLayoutInternal(const GUILayoutData& data)
{
UINT32 numElements = (UINT32)mChildren.size();
Rect2I* elementAreas = nullptr;
if (numElements > 0)
elementAreas = bs_stack_new<Rect2I>(numElements);
_getElementAreas(data.area, elementAreas, numElements, mChildSizeRanges, mSizeRange);
// Now that we have all the areas, actually assign them
UINT32 childIdx = 0;
GUILayoutData childData = data;
for(auto& child : mChildren)
{
if (child->_isActive())
{
childData.area = elementAreas[childIdx];
childData.clipRect = childData.area;
childData.clipRect.clip(data.clipRect);
child->_setLayoutData(childData);
child->_updateLayoutInternal(childData);
}
childIdx++;
}
if(elementAreas != nullptr)
bs_stack_free(elementAreas);
}
VkPipelineLayout VulkanDescriptorManager::getPipelineLayout(VulkanDescriptorLayout** layouts, UINT32 numLayouts)
{
VulkanPipelineLayoutKey key(layouts, numLayouts);
auto iterFind = mPipelineLayouts.find(key);
if (iterFind != mPipelineLayouts.end())
return iterFind->second;
// Create new
VkDescriptorSetLayout* setLayouts = (VkDescriptorSetLayout*)bs_stack_alloc(sizeof(VkDescriptorSetLayout) * numLayouts);
for(UINT32 i = 0; i < numLayouts; i++)
setLayouts[i] = layouts[i]->getHandle();
VkPipelineLayoutCreateInfo layoutCI;
layoutCI.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO;
layoutCI.pNext = nullptr;
layoutCI.flags = 0;
layoutCI.pushConstantRangeCount = 0;
layoutCI.pPushConstantRanges = nullptr;
layoutCI.setLayoutCount = numLayouts;
layoutCI.pSetLayouts = setLayouts;
VkPipelineLayout pipelineLayout;
VkResult result = vkCreatePipelineLayout(mDevice.getLogical(), &layoutCI, gVulkanAllocator, &pipelineLayout);
assert(result == VK_SUCCESS);
bs_stack_free(setLayouts);
key.layouts = (VulkanDescriptorLayout**)bs_alloc(sizeof(VulkanDescriptorLayout*) * numLayouts);
memcpy(key.layouts, layouts, sizeof(VulkanDescriptorLayout*) * numLayouts);
mPipelineLayouts.insert(std::make_pair(key, pipelineLayout));
return pipelineLayout;
}
void AudioUtility::convertBitDepth(const UINT8* input, UINT32 inBitDepth, UINT8* output, UINT32 outBitDepth, UINT32 numSamples)
{
INT32* srcBuffer = nullptr;
bool needTempBuffer = inBitDepth != 32;
if (needTempBuffer)
srcBuffer = (INT32*)bs_stack_alloc(numSamples * sizeof(INT32));
else
srcBuffer = (INT32*)input;
// Note: I convert to a temporary 32-bit buffer and then use that to convert to actual requested bit depth.
// It would be more efficient to convert directly from source to requested depth without a temporary buffer,
// at the cost of additional complexity. If this method ever becomes a performance issue consider that.
switch (inBitDepth)
{
case 8:
convert8To32Bits(input, srcBuffer, numSamples);
break;
case 16:
convert16To32Bits((INT16*)input, srcBuffer, numSamples);
break;
case 24:
BansheeEngine::convert24To32Bits(input, srcBuffer, numSamples);
break;
case 32:
// Do nothing
break;
default:
assert(false);
break;
}
switch (outBitDepth)
{
case 8:
convert32To8Bits(srcBuffer, output, numSamples);
break;
case 16:
convert32To16Bits(srcBuffer, (INT16*)output, numSamples);
break;
case 24:
convert32To24Bits(srcBuffer, output, numSamples);
break;
case 32:
memcpy(output, srcBuffer, numSamples * sizeof(INT32));
break;
default:
assert(false);
break;
}
if (needTempBuffer)
{
bs_stack_free(srcBuffer);
srcBuffer = nullptr;
}
}
void PhysXRigidbody::removeColliders()
{
UINT32 numShapes = mInternal->getNbShapes();
PxShape** shapes = (PxShape**)bs_stack_alloc(sizeof(PxShape*) * numShapes);
mInternal->getShapes(shapes, sizeof(PxShape*) * numShapes);
for (UINT32 i = 0; i < numShapes; i++)
{
Collider* collider = (Collider*)shapes[i]->userData;
collider->_getInternal()->_setCCD(false);
mInternal->detachShape(*shapes[i]);
}
bs_stack_free(shapes);
}
void TriangleClipperBase::getOrderedVertices(const ClipFace& face, UINT32* sortedVerts)
{
UINT32 numEdges = (UINT32)face.edges.size();
UINT32* sortedEdges = (UINT32*)bs_stack_alloc(sizeof(UINT32) * numEdges);
for (UINT32 i = 0; i < numEdges; i++)
sortedEdges[i] = face.edges[i];
// Bubble sort to arrange edges in contiguous order
for (UINT32 i0 = 0, i1 = 1, choice = 1; i1 < numEdges - 1; i0 = i1, i1++)
{
const ClipEdge& edge0 = mesh.edges[sortedEdges[i0]];
UINT32 current = edge0.verts[choice];
for (UINT32 j = i1; j < numEdges; j++)
{
const ClipEdge& edge1 = mesh.edges[sortedEdges[j]];
if (edge1.verts[0] == current || edge1.verts[1] == current)
{
std::swap(sortedEdges[i1], sortedEdges[j]);
choice = 1;
break;
}
}
}
// Add the first two vertices
sortedVerts[0] = mesh.edges[sortedEdges[0]].verts[0];
sortedVerts[1] = mesh.edges[sortedEdges[0]].verts[1];
// Add the remaining vertices
for (UINT32 i = 1; i < numEdges; i++)
{
const ClipEdge& edge = mesh.edges[sortedEdges[i]];
if (edge.verts[0] == sortedVerts[i])
sortedVerts[i + 1] = edge.verts[1];
else
sortedVerts[i + 1] = edge.verts[0];
}
bs_stack_free(sortedEdges);
}
void TriangleClipperBase::getOrderedFaces(FrameVector<FrameVector<UINT32>>& sortedFaces)
{
for (UINT32 i = 0; i < (UINT32)mesh.faces.size(); i++)
{
const ClipFace& face = mesh.faces[i];
if (face.visible)
{
// Get the ordered vertices of the face. The first and last
// element of the array are the same since the polyline is
// closed.
UINT32 numSortedVerts = (UINT32)face.edges.size() + 1;
UINT32* sortedVerts = (UINT32*)bs_stack_alloc(sizeof(UINT32) * numSortedVerts);
getOrderedVertices(face, sortedVerts);
FrameVector<UINT32> faceVerts;
// The convention is that the vertices should be counterclockwise
// ordered when viewed from the negative side of the plane of the
// face. If you need the opposite convention, switch the
// inequality in the if-else statement.
Vector3 normal = getNormal(sortedVerts, numSortedVerts);
if (Vector3::dot(mesh.faces[i].normal, normal) < 0)
{
// Clockwise, need to swap
for (INT32 j = (INT32)numSortedVerts - 2; j >= 0; j--)
faceVerts.push_back(sortedVerts[j]);
}
else
{
// Counterclockwise
for (int j = 0; j <= (INT32)numSortedVerts - 2; j++)
faceVerts.push_back(sortedVerts[j]);
}
sortedFaces.push_back(faceVerts);
bs_stack_free(sortedVerts);
}
}
}
void VulkanTexture::copyImage(VulkanTransferBuffer* cb, VulkanImage* srcImage, VulkanImage* dstImage,
VkImageLayout srcFinalLayout, VkImageLayout dstFinalLayout)
{
UINT32 numFaces = mProperties.getNumFaces();
UINT32 numMipmaps = mProperties.getNumMipmaps() + 1;
UINT32 mipWidth = mProperties.getWidth();
UINT32 mipHeight = mProperties.getHeight();
UINT32 mipDepth = mProperties.getDepth();
VkImageCopy* imageRegions = bs_stack_alloc<VkImageCopy>(numMipmaps);
for(UINT32 i = 0; i < numMipmaps; i++)
{
VkImageCopy& imageRegion = imageRegions[i];
imageRegion.srcOffset = { 0, 0, 0 };
imageRegion.dstOffset = { 0, 0, 0 };
imageRegion.extent = { mipWidth, mipHeight, mipDepth };
imageRegion.srcSubresource.baseArrayLayer = 0;
imageRegion.srcSubresource.layerCount = numFaces;
imageRegion.srcSubresource.mipLevel = i;
imageRegion.srcSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
imageRegion.dstSubresource.baseArrayLayer = 0;
imageRegion.dstSubresource.layerCount = numFaces;
imageRegion.dstSubresource.mipLevel = i;
imageRegion.dstSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
if (mipWidth != 1) mipWidth /= 2;
if (mipHeight != 1) mipHeight /= 2;
if (mipDepth != 1) mipDepth /= 2;
}
VkImageSubresourceRange range;
range.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
range.baseArrayLayer = 0;
range.layerCount = numFaces;
range.baseMipLevel = 0;
range.levelCount = numMipmaps;
VkImageLayout transferSrcLayout, transferDstLayout;
if (mDirectlyMappable)
{
transferSrcLayout = VK_IMAGE_LAYOUT_GENERAL;
transferDstLayout = VK_IMAGE_LAYOUT_GENERAL;
}
else
{
transferSrcLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL;
transferDstLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL;
}
// Transfer textures to a valid layout
cb->setLayout(srcImage, range, VK_ACCESS_TRANSFER_READ_BIT, transferSrcLayout);
cb->setLayout(dstImage, range, VK_ACCESS_TRANSFER_WRITE_BIT, transferDstLayout);
vkCmdCopyImage(cb->getCB()->getHandle(), srcImage->getHandle(), VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
dstImage->getHandle(), VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, numMipmaps, imageRegions);
// Transfer back to final layouts
VkAccessFlags srcAccessMask = srcImage->getAccessFlags(srcFinalLayout);
cb->setLayout(srcImage->getHandle(), VK_ACCESS_TRANSFER_READ_BIT, srcAccessMask,
transferSrcLayout, srcFinalLayout, range);
VkAccessFlags dstAccessMask = dstImage->getAccessFlags(dstFinalLayout);
cb->setLayout(dstImage->getHandle(), VK_ACCESS_TRANSFER_WRITE_BIT, dstAccessMask,
transferDstLayout, dstFinalLayout, range);
cb->getCB()->registerResource(srcImage, range, VulkanUseFlag::Read, ResourceUsage::Transfer);
cb->getCB()->registerResource(dstImage, range, VulkanUseFlag::Write, ResourceUsage::Transfer);
bs_stack_free(imageRegions);
}
Vector2I GUILayoutUtility::calcActualSizeInternal(UINT32 width, UINT32 height, GUILayout* layout)
{
UINT32 numElements = (UINT32)layout->_getNumChildren();
Rect2I* elementAreas = nullptr;
if (numElements > 0)
elementAreas = bs_stack_new<Rect2I>(numElements);
Rect2I parentArea;
parentArea.width = width;
parentArea.height = height;
layout->_getElementAreas(parentArea, elementAreas, numElements, layout->_getCachedChildSizeRanges(), layout->_getCachedSizeRange());
Rect2I* actualAreas = elementAreas; // We re-use the same array
for (UINT32 i = 0; i < numElements; i++)
{
GUIElementBase* child = layout->_getChild(i);
Rect2I childArea = elementAreas[i];
if (child->_getType() == GUIElementBase::Type::Layout || child->_getType() == GUIElementBase::Type::Panel)
{
Vector2I childActualSize = calcActualSizeInternal(childArea.width, childArea.height, static_cast<GUILayout*>(child));
actualAreas[i].width = (UINT32)childActualSize.x;
actualAreas[i].height = (UINT32)childActualSize.y;
}
else if (child->_getType() == GUIElementBase::Type::Element)
{
RectOffset padding = child->_getPadding();
actualAreas[i].width = elementAreas[i].width + padding.left + padding.right;
actualAreas[i].height = elementAreas[i].height + padding.top + padding.bottom;
}
else
{
actualAreas[i].width = elementAreas[i].width;
actualAreas[i].height = elementAreas[i].height;
}
}
Vector2I min;
Vector2I max;
if (numElements > 0)
{
Rect2I childArea = actualAreas[0];
min = Vector2I(childArea.x, childArea.y);
max = Vector2I(childArea.x + childArea.width, childArea.y + childArea.height);
}
for (UINT32 i = 1; i < numElements; i++)
{
Rect2I childArea = actualAreas[i];
min.x = std::min(min.x, childArea.x);
min.y = std::min(min.y, childArea.y);
max.x = std::max(max.x, childArea.x + (INT32)childArea.width);
max.y = std::max(max.y, childArea.y + (INT32)childArea.height);
}
Vector2I actualSize = max - min;
if (elementAreas != nullptr)
bs_stack_free(elementAreas);
return actualSize;
}
void VulkanGraphicsPipelineState::initialize()
{
Lock(mMutex);
GraphicsPipelineState::initialize();
std::pair<VkShaderStageFlagBits, GpuProgram*> stages[] =
{
{ VK_SHADER_STAGE_VERTEX_BIT, mData.vertexProgram.get() },
{ VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT, mData.hullProgram.get() },
{ VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT, mData.domainProgram.get() },
{ VK_SHADER_STAGE_GEOMETRY_BIT, mData.geometryProgram.get() },
{ VK_SHADER_STAGE_FRAGMENT_BIT, mData.fragmentProgram.get() }
};
UINT32 stageOutputIdx = 0;
UINT32 numStages = sizeof(stages) / sizeof(stages[0]);
for(UINT32 i = 0; i < numStages; i++)
{
VulkanGpuProgram* program = static_cast<VulkanGpuProgram*>(stages[i].second);
if (program == nullptr)
continue;
VkPipelineShaderStageCreateInfo& stageCI = mShaderStageInfos[stageOutputIdx];
stageCI.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
stageCI.pNext = nullptr;
stageCI.flags = 0;
stageCI.stage = stages[i].first;
stageCI.module = VK_NULL_HANDLE;
stageCI.pName = program->getProperties().getEntryPoint().c_str();
stageCI.pSpecializationInfo = nullptr;
stageOutputIdx++;
}
UINT32 numUsedStages = stageOutputIdx;
bool tesselationEnabled = mData.hullProgram != nullptr && mData.domainProgram != nullptr;
mInputAssemblyInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO;
mInputAssemblyInfo.pNext = nullptr;
mInputAssemblyInfo.flags = 0;
mInputAssemblyInfo.topology = VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST; // Assigned at runtime
mInputAssemblyInfo.primitiveRestartEnable = false;
mTesselationInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_TESSELLATION_STATE_CREATE_INFO;
mTesselationInfo.pNext = nullptr;
mTesselationInfo.flags = 0;
mTesselationInfo.patchControlPoints = 3; // Assigned at runtime
mViewportInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO;
mViewportInfo.pNext = nullptr;
mViewportInfo.flags = 0;
mViewportInfo.viewportCount = 1; // Spec says this need to be at least 1...
mViewportInfo.scissorCount = 1;
mViewportInfo.pViewports = nullptr; // Dynamic
mViewportInfo.pScissors = nullptr; // Dynamic
RasterizerState* rasterizerState = getRasterizerState().get();
if (rasterizerState == nullptr)
rasterizerState = RasterizerState::getDefault().get();
BlendState* blendState = getBlendState().get();
if (blendState == nullptr)
blendState = BlendState::getDefault().get();
DepthStencilState* depthStencilState = getDepthStencilState().get();
if (depthStencilState == nullptr)
depthStencilState = DepthStencilState::getDefault().get();
const RasterizerProperties& rstProps = rasterizerState->getProperties();
const BlendProperties& blendProps = blendState->getProperties();
const DepthStencilProperties dsProps = depthStencilState->getProperties();
mRasterizationInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO;
mRasterizationInfo.pNext = nullptr;
mRasterizationInfo.flags = 0;
mRasterizationInfo.depthClampEnable = !rstProps.getDepthClipEnable();
mRasterizationInfo.rasterizerDiscardEnable = VK_FALSE;
mRasterizationInfo.polygonMode = VulkanUtility::getPolygonMode(rstProps.getPolygonMode());
mRasterizationInfo.cullMode = VulkanUtility::getCullMode(rstProps.getCullMode());
mRasterizationInfo.frontFace = VK_FRONT_FACE_CLOCKWISE;
mRasterizationInfo.depthBiasEnable = rstProps.getDepthBias() != 0.0f;
mRasterizationInfo.depthBiasConstantFactor = rstProps.getDepthBias();
mRasterizationInfo.depthBiasSlopeFactor = rstProps.getSlopeScaledDepthBias();
mRasterizationInfo.depthBiasClamp = mRasterizationInfo.depthClampEnable ? rstProps.getDepthBiasClamp() : 0.0f;
mRasterizationInfo.lineWidth = 1.0f;
mMultiSampleInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO;
mMultiSampleInfo.pNext = nullptr;
mMultiSampleInfo.flags = 0;
mMultiSampleInfo.rasterizationSamples = VK_SAMPLE_COUNT_1_BIT; // Assigned at runtime
mMultiSampleInfo.sampleShadingEnable = VK_FALSE; // When enabled, perform shading per sample instead of per pixel (more expensive, essentially FSAA)
mMultiSampleInfo.minSampleShading = 1.0f; // Minimum percent of samples to run full shading for when sampleShadingEnable is enabled (1.0f to run for all)
mMultiSampleInfo.pSampleMask = nullptr; // Normally one bit for each sample: e.g. 0x0000000F to enable all samples in a 4-sample setup
mMultiSampleInfo.alphaToCoverageEnable = blendProps.getAlphaToCoverageEnabled();
mMultiSampleInfo.alphaToOneEnable = VK_FALSE;
VkStencilOpState stencilFrontInfo;
stencilFrontInfo.compareOp = VulkanUtility::getCompareOp(dsProps.getStencilFrontCompFunc());
stencilFrontInfo.depthFailOp = VulkanUtility::getStencilOp(dsProps.getStencilFrontZFailOp());
stencilFrontInfo.passOp = VulkanUtility::getStencilOp(dsProps.getStencilFrontPassOp());
stencilFrontInfo.failOp = VulkanUtility::getStencilOp(dsProps.getStencilFrontFailOp());
stencilFrontInfo.reference = 0; // Dynamic
stencilFrontInfo.compareMask = (UINT32)dsProps.getStencilReadMask();
stencilFrontInfo.writeMask = (UINT32)dsProps.getStencilWriteMask();
VkStencilOpState stencilBackInfo;
stencilBackInfo.compareOp = VulkanUtility::getCompareOp(dsProps.getStencilBackCompFunc());
stencilBackInfo.depthFailOp = VulkanUtility::getStencilOp(dsProps.getStencilBackZFailOp());
stencilBackInfo.passOp = VulkanUtility::getStencilOp(dsProps.getStencilBackPassOp());
stencilBackInfo.failOp = VulkanUtility::getStencilOp(dsProps.getStencilBackFailOp());
stencilBackInfo.reference = 0; // Dynamic
stencilBackInfo.compareMask = (UINT32)dsProps.getStencilReadMask();
stencilBackInfo.writeMask = (UINT32)dsProps.getStencilWriteMask();
mDepthStencilInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO;
mDepthStencilInfo.pNext = nullptr;
mDepthStencilInfo.flags = 0;
mDepthStencilInfo.depthBoundsTestEnable = false;
mDepthStencilInfo.minDepthBounds = 0.0f;
mDepthStencilInfo.maxDepthBounds = 1.0f;
mDepthStencilInfo.depthTestEnable = dsProps.getDepthReadEnable();
mDepthStencilInfo.depthWriteEnable = dsProps.getDepthWriteEnable();
mDepthStencilInfo.depthCompareOp = VulkanUtility::getCompareOp(dsProps.getDepthComparisonFunc());
mDepthStencilInfo.front = stencilFrontInfo;
mDepthStencilInfo.back = stencilBackInfo;
mDepthStencilInfo.stencilTestEnable = dsProps.getStencilEnable();
for(UINT32 i = 0; i < BS_MAX_MULTIPLE_RENDER_TARGETS; i++)
{
UINT32 rtIdx = 0;
if (blendProps.getIndependantBlendEnable())
rtIdx = i;
VkPipelineColorBlendAttachmentState& blendState = mAttachmentBlendStates[i];
blendState.blendEnable = blendProps.getBlendEnabled(rtIdx);
blendState.colorBlendOp = VulkanUtility::getBlendOp(blendProps.getBlendOperation(rtIdx));
blendState.srcColorBlendFactor = VulkanUtility::getBlendFactor(blendProps.getSrcBlend(rtIdx));
blendState.dstColorBlendFactor = VulkanUtility::getBlendFactor(blendProps.getDstBlend(rtIdx));
blendState.alphaBlendOp = VulkanUtility::getBlendOp(blendProps.getAlphaBlendOperation(rtIdx));
blendState.srcAlphaBlendFactor = VulkanUtility::getBlendFactor(blendProps.getAlphaSrcBlend(rtIdx));
blendState.dstAlphaBlendFactor = VulkanUtility::getBlendFactor(blendProps.getAlphaDstBlend(rtIdx));
blendState.colorWriteMask = blendProps.getRenderTargetWriteMask(rtIdx) & 0xF;
}
mColorBlendStateInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO;
mColorBlendStateInfo.pNext = nullptr;
mColorBlendStateInfo.flags = 0;
mColorBlendStateInfo.logicOpEnable = VK_FALSE;
mColorBlendStateInfo.logicOp = VK_LOGIC_OP_NO_OP;
mColorBlendStateInfo.attachmentCount = 0; // Assigned at runtime
mColorBlendStateInfo.pAttachments = mAttachmentBlendStates;
mColorBlendStateInfo.blendConstants[0] = 0.0f;
mColorBlendStateInfo.blendConstants[1] = 0.0f;
mColorBlendStateInfo.blendConstants[2] = 0.0f;
mColorBlendStateInfo.blendConstants[3] = 0.0f;
mDynamicStates[0] = VK_DYNAMIC_STATE_VIEWPORT;
mDynamicStates[1] = VK_DYNAMIC_STATE_SCISSOR;
mDynamicStates[2] = VK_DYNAMIC_STATE_STENCIL_REFERENCE;
UINT32 numDynamicStates = sizeof(mDynamicStates) / sizeof(mDynamicStates[0]);
assert(numDynamicStates == 3);
mDynamicStateInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO;
mDynamicStateInfo.pNext = nullptr;
mDynamicStateInfo.flags = 0;
mDynamicStateInfo.dynamicStateCount = numDynamicStates;
mDynamicStateInfo.pDynamicStates = mDynamicStates;
mPipelineInfo.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO;
mPipelineInfo.pNext = nullptr;
mPipelineInfo.flags = 0;
mPipelineInfo.stageCount = numUsedStages;
mPipelineInfo.pStages = mShaderStageInfos;
mPipelineInfo.pVertexInputState = nullptr; // Assigned at runtime
mPipelineInfo.pInputAssemblyState = &mInputAssemblyInfo;
mPipelineInfo.pTessellationState = tesselationEnabled ? &mTesselationInfo : nullptr;
mPipelineInfo.pViewportState = &mViewportInfo;
mPipelineInfo.pRasterizationState = &mRasterizationInfo;
mPipelineInfo.pMultisampleState = &mMultiSampleInfo;
mPipelineInfo.pDepthStencilState = nullptr; // Assigned at runtime
mPipelineInfo.pColorBlendState = nullptr; // Assigned at runtime
mPipelineInfo.pDynamicState = &mDynamicStateInfo;
mPipelineInfo.renderPass = VK_NULL_HANDLE; // Assigned at runtime
mPipelineInfo.layout = VK_NULL_HANDLE; // Assigned at runtime
mPipelineInfo.subpass = 0;
mPipelineInfo.basePipelineHandle = VK_NULL_HANDLE;
mPipelineInfo.basePipelineIndex = -1;
mScissorEnabled = rstProps.getScissorEnable();
if(mData.vertexProgram != nullptr)
mVertexDecl = mData.vertexProgram->getInputDeclaration();
VulkanRenderAPI& rapi = static_cast<VulkanRenderAPI&>(RenderAPI::instance());
VulkanDevice* devices[BS_MAX_DEVICES];
VulkanUtility::getDevices(rapi, mDeviceMask, devices);
for (UINT32 i = 0; i < BS_MAX_DEVICES; i++)
{
if (devices[i] == nullptr)
continue;
mPerDeviceData[i].device = devices[i];
VulkanDescriptorManager& descManager = mPerDeviceData[i].device->getDescriptorManager();
VulkanGpuPipelineParamInfo& vkParamInfo = static_cast<VulkanGpuPipelineParamInfo&>(*mParamInfo);
UINT32 numLayouts = vkParamInfo.getNumSets();
VulkanDescriptorLayout** layouts = (VulkanDescriptorLayout**)bs_stack_alloc(sizeof(VulkanDescriptorLayout*) * numLayouts);
for (UINT32 j = 0; j < numLayouts; j++)
layouts[j] = vkParamInfo.getLayout(i, j);
mPerDeviceData[i].pipelineLayout = descManager.getPipelineLayout(layouts, numLayouts);
bs_stack_free(layouts);
}
BS_INC_RENDER_STAT_CAT(ResCreated, RenderStatObject_PipelineState);
}
void VulkanComputePipelineState::initialize()
{
ComputePipelineState::initialize();
// This might happen fairly often if shaders with unsupported languages are loaded, in which case the pipeline
// will never get used, and its fine not to initialize it.
if (!mProgram->isCompiled())
return;
VulkanGpuProgram* vkProgram = static_cast<VulkanGpuProgram*>(mProgram.get());
VkPipelineShaderStageCreateInfo stageCI;
stageCI.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
stageCI.pNext = nullptr;
stageCI.flags = 0;
stageCI.stage = VK_SHADER_STAGE_COMPUTE_BIT;
stageCI.module = VK_NULL_HANDLE;
stageCI.pName = vkProgram->getProperties().getEntryPoint().c_str();
stageCI.pSpecializationInfo = nullptr;
VkComputePipelineCreateInfo pipelineCI;
pipelineCI.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO;
pipelineCI.pNext = nullptr;
pipelineCI.flags = 0;
pipelineCI.stage = stageCI;
pipelineCI.basePipelineHandle = VK_NULL_HANDLE;
pipelineCI.basePipelineIndex = -1;
VulkanRenderAPI& rapi = static_cast<VulkanRenderAPI&>(RenderAPI::instance());
VulkanDevice* devices[BS_MAX_DEVICES];
VulkanUtility::getDevices(rapi, mDeviceMask, devices);
for (UINT32 i = 0; i < BS_MAX_DEVICES; i++)
{
if (devices[i] == nullptr)
continue;
mPerDeviceData[i].device = devices[i];
VulkanDescriptorManager& descManager = devices[i]->getDescriptorManager();
VulkanResourceManager& rescManager = devices[i]->getResourceManager();
VulkanGpuPipelineParamInfo& vkParamInfo = static_cast<VulkanGpuPipelineParamInfo&>(*mParamInfo);
UINT32 numLayouts = vkParamInfo.getNumSets();
VulkanDescriptorLayout** layouts = (VulkanDescriptorLayout**)bs_stack_alloc(sizeof(VulkanDescriptorLayout*) * numLayouts);
for (UINT32 j = 0; j < numLayouts; j++)
layouts[j] = vkParamInfo.getLayout(i, j);
VulkanShaderModule* module = vkProgram->getShaderModule(i);
if (module != nullptr)
pipelineCI.stage.module = module->getHandle();
else
pipelineCI.stage.module = VK_NULL_HANDLE;
pipelineCI.layout = descManager.getPipelineLayout(layouts, numLayouts);
VkPipeline pipeline;
VkResult result = vkCreateComputePipelines(devices[i]->getLogical(), VK_NULL_HANDLE, 1, &pipelineCI,
gVulkanAllocator, &pipeline);
assert(result == VK_SUCCESS);
mPerDeviceData[i].pipeline = rescManager.create<VulkanPipeline>(pipeline);
mPerDeviceData[i].pipelineLayout = pipelineCI.layout;
bs_stack_free(layouts);
}
BS_INC_RENDER_STAT_CAT(ResCreated, RenderStatObject_PipelineState);
}
void IconUtility::updateIconExe(const Path& path, const Map<UINT32, SPtr<PixelData>>& pixelsPerSize)
{
// A PE file is structured as such:
// - MSDOS Header
// - PE Signature
// - COFF Header
// - PE Optional Header
// - One or multiple sections
// - .code
// - .data
// - ...
// - .rsrc
// - icon/cursor/etc data
std::fstream stream;
stream.open(path.toPlatformString().c_str(), std::ios::in | std::ios::out | std::ios::binary);
// First check magic number to ensure file is even an executable
UINT16 magicNum;
stream.read((char*)&magicNum, sizeof(magicNum));
if (magicNum != MSDOS_SIGNATURE)
BS_EXCEPT(InvalidStateException, "Provided file is not a valid executable.");
// Read the MSDOS header and skip over it
stream.seekg(0);
MSDOSHeader msdosHeader;
stream.read((char*)&msdosHeader, sizeof(MSDOSHeader));
// Read PE signature
stream.seekg(msdosHeader.lfanew);
UINT32 peSignature;
stream.read((char*)&peSignature, sizeof(peSignature));
if (peSignature != PE_SIGNATURE)
BS_EXCEPT(InvalidStateException, "Provided file is not in PE format.");
// Read COFF header
COFFHeader coffHeader;
stream.read((char*)&coffHeader, sizeof(COFFHeader));
if (coffHeader.sizeOptHeader == 0) // .exe files always have an optional header
BS_EXCEPT(InvalidStateException, "Provided file is not a valid executable.");
UINT32 numSectionHeaders = coffHeader.numSections;
// Read optional header
auto optionalHeaderPos = stream.tellg();
UINT16 optionalHeaderSignature;
stream.read((char*)&optionalHeaderSignature, sizeof(optionalHeaderSignature));
PEDataDirectory* dataDirectory = nullptr;
stream.seekg(optionalHeaderPos);
if (optionalHeaderSignature == PE_32BIT_SIGNATURE)
{
PEOptionalHeader32 optionalHeader;
stream.read((char*)&optionalHeader, sizeof(optionalHeader));
dataDirectory = optionalHeader.dataDirectory + PE_IMAGE_DIRECTORY_ENTRY_RESOURCE;
}
else if (optionalHeaderSignature == PE_64BIT_SIGNATURE)
{
PEOptionalHeader64 optionalHeader;
stream.read((char*)&optionalHeader, sizeof(optionalHeader));
dataDirectory = optionalHeader.dataDirectory + PE_IMAGE_DIRECTORY_ENTRY_RESOURCE;
}
else
BS_EXCEPT(InvalidStateException, "Unrecognized PE format.");
// Read section headers
auto sectionHeaderPos = optionalHeaderPos + (std::ifstream::pos_type)coffHeader.sizeOptHeader;
stream.seekg(sectionHeaderPos);
PESectionHeader* sectionHeaders = bs_stack_alloc<PESectionHeader>(numSectionHeaders);
stream.read((char*)sectionHeaders, sizeof(PESectionHeader) * numSectionHeaders);
// Look for .rsrc section header
std::function<void(PEImageResourceDirectory*, PEImageResourceDirectory*, UINT8*, UINT32)> setIconData =
[&](PEImageResourceDirectory* base, PEImageResourceDirectory* current, UINT8* imageData, UINT32 sectionAddress)
{
UINT32 numEntries = current->numIdEntries; // Not supporting name entries
PEImageResourceEntry* entries = (PEImageResourceEntry*)(current + 1);
for (UINT32 i = 0; i < numEntries; i++)
{
// Only at root does the type identify resource type
if (base == current && entries[i].type != PE_IMAGE_RT_ICON)
continue;
if (entries[i].isDirectory)
{
PEImageResourceDirectory* child = (PEImageResourceDirectory*)(((UINT8*)base) + entries[i].offsetDirectory);
setIconData(base, child, imageData, sectionAddress);
}
else
{
PEImageResourceEntryData* data = (PEImageResourceEntryData*)(((UINT8*)base) + entries[i].offsetDirectory);
UINT8* iconData = imageData + (data->offsetData - sectionAddress);
updateIconData(iconData, pixelsPerSize);
}
}
};
for (UINT32 i = 0; i < numSectionHeaders; i++)
{
if (sectionHeaders[i].flags & PE_SECTION_UNINITIALIZED_DATA)
continue;
if (strcmp(sectionHeaders[i].name, ".rsrc") == 0)
{
UINT32 imageSize = sectionHeaders[i].physicalSize;
UINT8* imageData = (UINT8*)bs_stack_alloc(imageSize);
stream.seekg(sectionHeaders[i].physicalAddress);
stream.read((char*)imageData, imageSize);
UINT32 resourceDirOffset = dataDirectory->virtualAddress - sectionHeaders[i].relativeVirtualAddress;
PEImageResourceDirectory* resourceDirectory = (PEImageResourceDirectory*)&imageData[resourceDirOffset];
setIconData(resourceDirectory, resourceDirectory, imageData, sectionHeaders[i].relativeVirtualAddress);
stream.seekp(sectionHeaders[i].physicalAddress);
stream.write((char*)imageData, imageSize);
bs_stack_free(imageData);
}
}
bs_stack_free(sectionHeaders);
stream.close();
}
void VulkanSwapChain::rebuild(const SPtr<VulkanDevice>& device, VkSurfaceKHR surface, UINT32 width, UINT32 height,
bool vsync, VkFormat colorFormat, VkColorSpaceKHR colorSpace, bool createDepth, VkFormat depthFormat)
{
mDevice = device;
VkResult result;
VkPhysicalDevice physicalDevice = device->getPhysical();
// Determine swap chain dimensions
VkSurfaceCapabilitiesKHR surfaceCaps;
result = vkGetPhysicalDeviceSurfaceCapabilitiesKHR(physicalDevice, surface, &surfaceCaps);
assert(result == VK_SUCCESS);
VkExtent2D swapchainExtent;
// If width/height is 0xFFFFFFFF, we can manually specify width, height
if (surfaceCaps.currentExtent.width == (uint32_t)-1 || surfaceCaps.currentExtent.height == (uint32_t)-1)
{
swapchainExtent.width = width;
swapchainExtent.height = height;
}
else // Otherwise we must use the size we're given
swapchainExtent = surfaceCaps.currentExtent;
mWidth = swapchainExtent.width;
mHeight = swapchainExtent.height;
// Find present mode
uint32_t numPresentModes;
result = vkGetPhysicalDeviceSurfacePresentModesKHR(physicalDevice, surface, &numPresentModes, nullptr);
assert(result == VK_SUCCESS);
assert(numPresentModes > 0);
VkPresentModeKHR* presentModes = bs_stack_alloc<VkPresentModeKHR>(numPresentModes);
result = vkGetPhysicalDeviceSurfacePresentModesKHR(physicalDevice, surface, &numPresentModes, presentModes);
assert(result == VK_SUCCESS);
VkPresentModeKHR presentMode = VK_PRESENT_MODE_FIFO_KHR;
if(!vsync)
{
for (UINT32 i = 0; i < numPresentModes; i++)
{
if (presentModes[i] == VK_PRESENT_MODE_IMMEDIATE_KHR)
{
presentMode = VK_PRESENT_MODE_IMMEDIATE_KHR;
break;
}
if (presentModes[i] == VK_PRESENT_MODE_FIFO_RELAXED_KHR)
presentMode = VK_PRESENT_MODE_FIFO_RELAXED_KHR;
}
}
else
{
// Mailbox comes with lower input latency than FIFO, but can waste GPU power by rendering frames that are never
// displayed, especially if the app runs much faster than the refresh rate. This is a concern for mobiles.
#if BS_PLATFORM != BS_PLATFORM_ANDROID && BS_PLATFORM != BS_PLATFORM_IOS
for (UINT32 i = 0; i < numPresentModes; i++)
{
if (presentModes[i] == VK_PRESENT_MODE_MAILBOX_KHR)
{
presentMode = VK_PRESENT_MODE_MAILBOX_KHR;
break;
}
}
#endif
}
bs_stack_free(presentModes);
uint32_t numImages = surfaceCaps.minImageCount;
VkSurfaceTransformFlagsKHR transform;
if (surfaceCaps.supportedTransforms & VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR)
transform = VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR;
else
transform = surfaceCaps.currentTransform;
VkSwapchainCreateInfoKHR swapChainCI;
swapChainCI.sType = VK_STRUCTURE_TYPE_SWAPCHAIN_CREATE_INFO_KHR;
swapChainCI.pNext = nullptr;
swapChainCI.flags = 0;
swapChainCI.surface = surface;
swapChainCI.minImageCount = numImages;
swapChainCI.imageFormat = colorFormat;
swapChainCI.imageColorSpace = colorSpace;
swapChainCI.imageExtent = { swapchainExtent.width, swapchainExtent.height };
swapChainCI.imageUsage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
swapChainCI.preTransform = (VkSurfaceTransformFlagBitsKHR)transform;
swapChainCI.imageArrayLayers = 1;
swapChainCI.imageSharingMode = VK_SHARING_MODE_EXCLUSIVE;
swapChainCI.queueFamilyIndexCount = 0;
swapChainCI.pQueueFamilyIndices = NULL;
swapChainCI.presentMode = presentMode;
swapChainCI.oldSwapchain = mSwapChain;
swapChainCI.clipped = VK_TRUE;
swapChainCI.compositeAlpha = VK_COMPOSITE_ALPHA_OPAQUE_BIT_KHR;
VkSwapchainKHR oldSwapChain = mSwapChain;
VkDevice logicalDevice = device->getLogical();
result = vkCreateSwapchainKHR(logicalDevice, &swapChainCI, gVulkanAllocator, &mSwapChain);
assert(result == VK_SUCCESS);
clear(oldSwapChain);
result = vkGetSwapchainImagesKHR(logicalDevice, mSwapChain, &numImages, nullptr);
assert(result == VK_SUCCESS);
// Get the swap chain images
VkImage* images = bs_stack_alloc<VkImage>(numImages);
result = vkGetSwapchainImagesKHR(logicalDevice, mSwapChain, &numImages, images);
assert(result == VK_SUCCESS);
VulkanResourceManager& resManager = device->getResourceManager();
VULKAN_IMAGE_DESC imageDesc;
imageDesc.format = colorFormat;
imageDesc.type = TEX_TYPE_2D;
imageDesc.usage = TU_RENDERTARGET;
imageDesc.layout = VK_IMAGE_LAYOUT_UNDEFINED;
imageDesc.numFaces = 1;
imageDesc.numMipLevels = 1;
imageDesc.memory = VK_NULL_HANDLE;
mSurfaces.resize(numImages);
for (UINT32 i = 0; i < numImages; i++)
{
imageDesc.image = images[i];
mSurfaces[i].acquired = false;
mSurfaces[i].needsWait = false;
mSurfaces[i].image = resManager.create<VulkanImage>(imageDesc, false);
mSurfaces[i].sync = resManager.create<VulkanSemaphore>();
}
bs_stack_free(images);
// Create depth stencil image
if (createDepth)
{
VkImageCreateInfo depthStencilImageCI;
depthStencilImageCI.sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO;
depthStencilImageCI.pNext = nullptr;
depthStencilImageCI.flags = 0;
depthStencilImageCI.imageType = VK_IMAGE_TYPE_2D;
depthStencilImageCI.format = depthFormat;
depthStencilImageCI.extent = { width, height, 1 };
depthStencilImageCI.mipLevels = 1;
depthStencilImageCI.arrayLayers = 1;
depthStencilImageCI.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED;
depthStencilImageCI.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
depthStencilImageCI.samples = VK_SAMPLE_COUNT_1_BIT;
depthStencilImageCI.tiling = VK_IMAGE_TILING_OPTIMAL;
depthStencilImageCI.usage = VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
depthStencilImageCI.pQueueFamilyIndices = nullptr;
depthStencilImageCI.queueFamilyIndexCount = 0;
VkImage depthStencilImage;
result = vkCreateImage(logicalDevice, &depthStencilImageCI, gVulkanAllocator, &depthStencilImage);
assert(result == VK_SUCCESS);
imageDesc.image = depthStencilImage;
imageDesc.usage = TU_DEPTHSTENCIL;
imageDesc.format = depthFormat;
imageDesc.memory = mDevice->allocateMemory(depthStencilImage, VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT);
mDepthStencilImage = resManager.create<VulkanImage>(imageDesc, true);
}
else
mDepthStencilImage = nullptr;
// Create a framebuffer for each swap chain buffer
UINT32 numFramebuffers = (UINT32)mSurfaces.size();
for (UINT32 i = 0; i < numFramebuffers; i++)
{
VULKAN_FRAMEBUFFER_DESC& desc = mSurfaces[i].framebufferDesc;
desc.width = getWidth();
desc.height = getHeight();
desc.layers = 1;
desc.numSamples = 1;
desc.offscreen = false;
desc.color[0].format = colorFormat;
desc.color[0].image = mSurfaces[i].image;
desc.color[0].surface = TextureSurface::COMPLETE;
desc.color[0].baseLayer = 0;
desc.depth.format = depthFormat;
desc.depth.image = mDepthStencilImage;
desc.depth.surface = TextureSurface::COMPLETE;
desc.depth.baseLayer = 0;
mSurfaces[i].framebuffer = resManager.create<VulkanFramebuffer>(desc);
}
}
void FMODAudioClip::initialize()
{
AudioFileInfo info;
info.bitDepth = mDesc.bitDepth;
info.numChannels = mDesc.numChannels;
info.numSamples = mNumSamples;
info.sampleRate = mDesc.frequency;
// If we need to keep source data, read everything into memory and keep a copy
if (mKeepSourceData)
{
mStreamData->seek(mStreamOffset);
UINT8* sampleBuffer = (UINT8*)bs_alloc(mStreamSize);
mStreamData->read(sampleBuffer, mStreamSize);
mSourceStreamData = bs_shared_ptr_new<MemoryDataStream>(sampleBuffer, mStreamSize);
mSourceStreamSize = mStreamSize;
}
FMOD::System* fmod = gFMODAudio()._getFMOD();
FMOD_MODE flags = FMOD_OPENMEMORY;
if (is3D())
flags |= FMOD_3D;
else
flags |= FMOD_2D;
// Load data into a sound buffer
// TODO - Vorbis cannot be decompressed from memory by FMOD. Instead we force AudioReadMode::Stream for it.
if(mDesc.readMode == AudioReadMode::LoadDecompressed ||
(mDesc.readMode == AudioReadMode::LoadCompressed && mDesc.format != AudioFormat::VORBIS))
{
// Read all data into memory
SPtr<DataStream> stream;
UINT32 offset = 0;
if (mSourceStreamData != nullptr) // If it's already loaded in memory, use it directly
stream = mSourceStreamData;
else
{
stream = mStreamData;
offset = mStreamOffset;
}
UINT32 bufferSize = info.numSamples * (info.bitDepth / 8);
UINT8* sampleBuffer = (UINT8*)bs_stack_alloc(bufferSize);
stream->seek(offset);
stream->read(sampleBuffer, bufferSize);
FMOD_CREATESOUNDEXINFO exInfo;
memset(&exInfo, 0, sizeof(exInfo));
exInfo.cbsize = sizeof(exInfo);
exInfo.length = bufferSize;
bool loadCompressed = mDesc.readMode == AudioReadMode::LoadCompressed && mDesc.format != AudioFormat::PCM;
if (loadCompressed)
flags |= FMOD_CREATECOMPRESSEDSAMPLE;
else
flags |= FMOD_CREATESAMPLE;
if(fmod->createSound((const char*)sampleBuffer, flags, &exInfo, &mSound) != FMOD_OK)
{
LOGERR("Failed playing sound.");
}
else
{
mSound->setMode(FMOD_LOOP_OFF);
}
mStreamData = nullptr;
mStreamOffset = 0;
mStreamSize = 0;
bs_stack_free(sampleBuffer);
}
else // Streaming
{
// Do nothing, we rely on AudioSource from creating sounds as only one streaming sound can ever be playing
}
AudioClip::initialize();
}
void GLSLGpuProgram::initialize()
{
#if BS_OPENGL_4_5
static const char* VERSION_CHARS = "450";
#elif BS_OPENGL_4_4
static const char* VERSION_CHARS = "440";
#elif BS_OPENGL_4_3
static const char* VERSION_CHARS = "430";
#elif BS_OPENGL_4_2
static const char* VERSION_CHARS = "420";
#else
static const char* VERSION_CHARS = "410";
#endif
if (!isSupported())
{
mIsCompiled = false;
mCompileMessages = "Specified GPU program type is not supported by the current render system.";
GpuProgram::initialize();
return;
}
GLenum shaderType = 0x0000;
switch (mType)
{
case GPT_VERTEX_PROGRAM:
shaderType = GL_VERTEX_SHADER;
mProgramID = ++sVertexShaderCount;
break;
case GPT_FRAGMENT_PROGRAM:
shaderType = GL_FRAGMENT_SHADER;
mProgramID = ++sFragmentShaderCount;
break;
#if BS_OPENGL_4_1 || BS_OPENGLES_3_2
case GPT_GEOMETRY_PROGRAM:
shaderType = GL_GEOMETRY_SHADER;
mProgramID = ++sGeometryShaderCount;
break;
case GPT_HULL_PROGRAM:
shaderType = GL_TESS_CONTROL_SHADER;
mProgramID = ++sDomainShaderCount;
break;
case GPT_DOMAIN_PROGRAM:
shaderType = GL_TESS_EVALUATION_SHADER;
mProgramID = ++sHullShaderCount;
break;
#endif
#if BS_OPENGL_4_3 || BS_OPENGLES_3_1
case GPT_COMPUTE_PROGRAM:
shaderType = GL_COMPUTE_SHADER;
mProgramID = ++sComputeShaderCount;
break;
#endif
default:
break;
}
// Add preprocessor extras and main source
const String& source = mSource;
if (!source.empty())
{
Vector<GLchar*> lines;
const char* versionStr = "#version ";
UINT32 versionStrLen = (UINT32)strlen(versionStr);
UINT32 lineLength = 0;
INT32 versionLineNum = -1;
for (UINT32 i = 0; i < source.size(); i++)
{
if (source[i] == '\n' || source[i] == '\r')
{
assert(sizeof(source[i]) == sizeof(GLchar));
GLchar* lineData = (GLchar*)bs_stack_alloc(sizeof(GLchar) * (lineLength + 2));
memcpy(lineData, &source[i - lineLength], sizeof(GLchar) * lineLength);
lineData[lineLength] = '\n';
lineData[lineLength + 1] = '\0';
if(versionLineNum == -1 && lineLength >= versionStrLen)
{
bool isEqual = true;
for (UINT32 j = 0; j < versionStrLen; ++j)
{
if(lineData[j] != versionStr[j])
{
isEqual = false;
break;
}
}
if (isEqual)
versionLineNum = (INT32)lines.size();
}
lines.push_back(lineData);
lineLength = 0;
}
else
{
lineLength++;
}
}
if (lineLength > 0)
{
UINT32 end = (UINT32)source.size() - 1;
assert(sizeof(source[end]) == sizeof(GLchar));
GLchar* lineData = (GLchar*)bs_stack_alloc(sizeof(GLchar) * (lineLength + 1));
memcpy(lineData, &source[source.size() - lineLength], sizeof(GLchar) * lineLength);
lineData[lineLength] = '\0';
lines.push_back(lineData);
lineLength = 0;
}
int numInsertedLines = 0;
if(versionLineNum == -1)
{
char versionLine[50];
strcpy(versionLine, "#version ");
strcat(versionLine, VERSION_CHARS);
strcat(versionLine, "\n");
UINT32 length = (UINT32)strlen(versionLine) + 1;
GLchar* extraLineData = (GLchar*)bs_stack_alloc(length);
memcpy(extraLineData, versionLine, length);
lines.insert(lines.begin(), extraLineData);
numInsertedLines++;
}
char versionDefine[50];
strcpy(versionDefine, "#define OPENGL");
strcat(versionDefine, VERSION_CHARS);
strcat(versionDefine, "\n");
const char* EXTRA_LINES[] =
{
"#define OPENGL\n",
versionDefine
};
UINT32 numExtraLines = sizeof(EXTRA_LINES) / sizeof(EXTRA_LINES[0]);
UINT32 extraLineOffset = versionLineNum != -1 ? versionLineNum + 1 : 0;
for (UINT32 i = 0; i < numExtraLines; i++)
{
UINT32 length = (UINT32)strlen(EXTRA_LINES[i]) + 1;
GLchar* extraLineData = (GLchar*)bs_stack_alloc(length);
memcpy(extraLineData, EXTRA_LINES[i], length);
lines.insert(lines.begin() + extraLineOffset + numInsertedLines, extraLineData);
numInsertedLines++;
}
StringStream codeStream;
for(auto& entry : lines)
codeStream << entry;
for (INT32 i = numInsertedLines - 1; i >= 0; i--)
bs_stack_free(lines[extraLineOffset + i]);
if (numInsertedLines > 0)
lines.erase(lines.begin() + extraLineOffset, lines.begin() + extraLineOffset + numInsertedLines);
for (auto iter = lines.rbegin(); iter != lines.rend(); ++iter)
bs_stack_free(*iter);
String code = codeStream.str();
const char* codeRaw = code.c_str();
mGLHandle = glCreateShaderProgramv(shaderType, 1, (const GLchar**)&codeRaw);
BS_CHECK_GL_ERROR();
mCompileMessages = "";
mIsCompiled = !checkForGLSLError(mGLHandle, mCompileMessages);
}
if (mIsCompiled)
{
GLSLParamParser paramParser;
paramParser.buildUniformDescriptions(mGLHandle, mType, *mParametersDesc);
if (mType == GPT_VERTEX_PROGRAM)
{
Vector<VertexElement> elementList = paramParser.buildVertexDeclaration(mGLHandle);
mInputDeclaration = HardwareBufferManager::instance().createVertexDeclaration(elementList);
}
}
BS_INC_RENDER_STAT_CAT(ResCreated, RenderStatObject_GpuProgram);
GpuProgram::initialize();
}
void MonoAssembly::load(MonoDomain* domain)
{
if (mIsLoaded)
unload();
// Load assembly from memory because mono_domain_assembly_open keeps a lock on the file
SPtr<DataStream> assemblyStream = FileSystem::openFile(mPath, true);
if (assemblyStream == nullptr)
{
LOGERR("Cannot load assembly at path \"" + toString(mPath) + "\" because the file doesn't exist");
return;
}
UINT32 assemblySize = (UINT32)assemblyStream->size();
char* assemblyData = (char*)bs_stack_alloc(assemblySize);
assemblyStream->read(assemblyData, assemblySize);
String imageName = Path(mPath).getFilename();
MonoImageOpenStatus status = MONO_IMAGE_OK;
MonoImage* image = mono_image_open_from_data_with_name(assemblyData, assemblySize, true, &status, false, imageName.c_str());
bs_stack_free(assemblyData);
if (status != MONO_IMAGE_OK || image == nullptr)
{
LOGERR("Failed loading image data for assembly \"" + toString(mPath) + "\"");
return;
}
// Load MDB file
#if BS_DEBUG_MODE
Path mdbPath = mPath + L".mdb";
if (FileSystem::exists(mdbPath))
{
SPtr<DataStream> mdbStream = FileSystem::openFile(mdbPath, true);
if (mdbStream != nullptr)
{
UINT32 mdbSize = (UINT32)mdbStream->size();
mDebugData = (UINT8*)bs_alloc(mdbSize);
mdbStream->read(mDebugData, mdbSize);
mono_debug_open_image_from_memory(image, mDebugData, mdbSize);
}
}
#endif
mMonoAssembly = mono_assembly_load_from_full(image, imageName.c_str(), &status, false);
if (status != MONO_IMAGE_OK || mMonoAssembly == nullptr)
{
LOGERR("Failed loading assembly \"" + toString(mPath) + "\"");
return;
}
mMonoImage = image;
if(mMonoImage == nullptr)
{
BS_EXCEPT(InvalidParametersException, "Cannot get script assembly image.");
}
mIsLoaded = true;
mIsDependency = false;
}
void OAAudioClip::initialize()
{
{
Lock lock(mMutex); // Needs to be called even if stream data is null, to ensure memory fence is added so the
// other thread sees properly initialized AudioClip members
AudioDataInfo info;
info.bitDepth = mDesc.bitDepth;
info.numChannels = mDesc.numChannels;
info.numSamples = mNumSamples;
info.sampleRate = mDesc.frequency;
// If we need to keep source data, read everything into memory and keep a copy
if (mKeepSourceData)
{
mStreamData->seek(mStreamOffset);
UINT8* sampleBuffer = (UINT8*)bs_alloc(mStreamSize);
mStreamData->read(sampleBuffer, mStreamSize);
mSourceStreamData = bs_shared_ptr_new<MemoryDataStream>(sampleBuffer, mStreamSize);
mSourceStreamSize = mStreamSize;
}
// Load decompressed data into a sound buffer
bool loadDecompressed =
mDesc.readMode == AudioReadMode::LoadDecompressed ||
(mDesc.readMode == AudioReadMode::LoadCompressed && mDesc.format == AudioFormat::PCM);
if(loadDecompressed)
{
// Read all data into memory
SPtr<DataStream> stream;
UINT32 offset = 0;
if (mSourceStreamData != nullptr) // If it's already loaded in memory, use it directly
stream = mSourceStreamData;
else
{
stream = mStreamData;
offset = mStreamOffset;
}
UINT32 bufferSize = info.numSamples * (info.bitDepth / 8);
UINT8* sampleBuffer = (UINT8*)bs_stack_alloc(bufferSize);
// Decompress from Ogg
if (mDesc.format == AudioFormat::VORBIS)
{
OggVorbisDecoder reader;
if (reader.open(stream, info, offset))
reader.read(sampleBuffer, info.numSamples);
else
LOGERR("Failed decompressing AudioClip stream.");
}
// Load directly
else
{
stream->seek(offset);
stream->read(sampleBuffer, bufferSize);
}
alGenBuffers(1, &mBufferId);
gOAAudio()._writeToOpenALBuffer(mBufferId, sampleBuffer, info);
mStreamData = nullptr;
mStreamOffset = 0;
mStreamSize = 0;
bs_stack_free(sampleBuffer);
}
// Load compressed data for streaming from memory
else if(mDesc.readMode == AudioReadMode::LoadCompressed)
{
// If reading from file, make a copy of data in memory, otherwise just take ownership of the existing buffer
if (mStreamData->isFile())
{
if (mSourceStreamData != nullptr) // If it's already loaded in memory, use it directly
mStreamData = mSourceStreamData;
else
{
UINT8* data = (UINT8*)bs_alloc(mStreamSize);
mStreamData->seek(mStreamOffset);
mStreamData->read(data, mStreamSize);
mStreamData = bs_shared_ptr_new<MemoryDataStream>(data, mStreamSize);
}
mStreamOffset = 0;
}
}
// Keep original stream for streaming from file
else
{
// Do nothing
}
if (mDesc.format == AudioFormat::VORBIS && mDesc.readMode != AudioReadMode::LoadDecompressed)
{
mNeedsDecompression = true;
if (mStreamData != nullptr)
{
if (!mVorbisReader.open(mStreamData, info, mStreamOffset))
LOGERR("Failed decompressing AudioClip stream.");
}
}
}
AudioClip::initialize();
}
void GUIScrollArea::_updateLayoutInternal(const GUILayoutData& data)
{
UINT32 numElements = (UINT32)mChildren.size();
Rect2I* elementAreas = nullptr;
if (numElements > 0)
elementAreas = bs_stack_new<Rect2I>(numElements);
UINT32 layoutIdx = 0;
UINT32 horzScrollIdx = 0;
UINT32 vertScrollIdx = 0;
for (UINT32 i = 0; i < numElements; i++)
{
GUIElementBase* child = _getChild(i);
if (child == mContentLayout)
layoutIdx = i;
if (child == mHorzScroll)
horzScrollIdx = i;
if (child == mVertScroll)
vertScrollIdx = i;
}
_getElementAreas(data.area, elementAreas, numElements, mChildSizeRanges, mVisibleSize, mContentSize);
Rect2I& layoutBounds = elementAreas[layoutIdx];
Rect2I& horzScrollBounds = elementAreas[horzScrollIdx];
Rect2I& vertScrollBounds = elementAreas[vertScrollIdx];
// Recalculate offsets in case scroll percent got updated externally (this needs to be delayed to this point because
// at the time of the update content and visible sizes might be out of date).
UINT32 scrollableHeight = (UINT32)std::max(0, INT32(mContentSize.y) - INT32(mVisibleSize.y));
if (mRecalculateVertOffset)
{
mVertOffset = scrollableHeight * Math::clamp01(mVertScroll->getScrollPos());
mRecalculateVertOffset = false;
}
UINT32 scrollableWidth = (UINT32)std::max(0, INT32(mContentSize.x) - INT32(mVisibleSize.x));
if (mRecalculateHorzOffset)
{
mHorzOffset = scrollableWidth * Math::clamp01(mHorzScroll->getScrollPos());
mRecalculateHorzOffset = false;
}
// Reset offset in case layout size changed so everything fits
mVertOffset = Math::clamp(mVertOffset, 0.0f, (float)scrollableHeight);
mHorzOffset = Math::clamp(mHorzOffset, 0.0f, (float)scrollableWidth);
// Layout
if (mContentLayout->_isActive())
{
layoutBounds.x -= Math::floorToInt(mHorzOffset);
layoutBounds.y -= Math::floorToInt(mVertOffset);
Rect2I layoutClipRect = data.clipRect;
layoutClipRect.width = (UINT32)mVisibleSize.x;
layoutClipRect.height = (UINT32)mVisibleSize.y;
layoutClipRect.clip(data.clipRect);
GUILayoutData layoutData = data;
layoutData.area = layoutBounds;
layoutData.clipRect = layoutClipRect;
mContentLayout->_setLayoutData(layoutData);
mContentLayout->_updateLayoutInternal(layoutData);
}
// Vertical scrollbar
{
GUILayoutData vertScrollData = data;
vertScrollData.area = vertScrollBounds;
vertScrollData.clipRect = vertScrollBounds;
vertScrollData.clipRect.clip(data.clipRect);
mVertScroll->_setLayoutData(vertScrollData);
mVertScroll->_updateLayoutInternal(vertScrollData);
// Set new handle size and update position to match the new size
UINT32 scrollableHeight = (UINT32)std::max(0, INT32(mContentSize.y) - INT32(vertScrollBounds.height));
float newScrollPct = 0.0f;
if (scrollableHeight > 0)
newScrollPct = mVertOffset / scrollableHeight;
mVertScroll->_setHandleSize(vertScrollBounds.height / (float)mContentSize.y);
mVertScroll->_setScrollPos(newScrollPct);
}
// Horizontal scrollbar
{
GUILayoutData horzScrollData = data;
horzScrollData.area = horzScrollBounds;
horzScrollData.clipRect = horzScrollBounds;
horzScrollData.clipRect.clip(data.clipRect);
mHorzScroll->_setLayoutData(horzScrollData);
mHorzScroll->_updateLayoutInternal(horzScrollData);
// Set new handle size and update position to match the new size
UINT32 scrollableWidth = (UINT32)std::max(0, INT32(mContentSize.x) - INT32(horzScrollBounds.width));
float newScrollPct = 0.0f;
if (scrollableWidth > 0)
newScrollPct = mHorzOffset / scrollableWidth;
mHorzScroll->_setHandleSize(horzScrollBounds.width / (float)mContentSize.x);
mHorzScroll->_setScrollPos(newScrollPct);
}
if (elementAreas != nullptr)
bs_stack_free(elementAreas);
}
void GUILayoutX::_getElementAreas(const Rect2I& layoutArea, Rect2I* elementAreas, UINT32 numElements,
const Vector<LayoutSizeRange>& sizeRanges, const LayoutSizeRange& mySizeRange) const
{
assert(mChildren.size() == numElements);
UINT32 totalOptimalSize = mySizeRange.optimal.x;
UINT32 totalNonClampedSize = 0;
UINT32 numNonClampedElements = 0;
UINT32 numFlexibleSpaces = 0;
bool* processedElements = nullptr;
float* elementScaleWeights = nullptr;
if (mChildren.size() > 0)
{
processedElements = bs_stack_alloc<bool>((UINT32)mChildren.size());
memset(processedElements, 0, mChildren.size() * sizeof(bool));
elementScaleWeights = bs_stack_alloc<float>((UINT32)mChildren.size());
memset(elementScaleWeights, 0, mChildren.size() * sizeof(float));
}
// Set initial sizes, count number of children per type and mark fixed elements as already processed
UINT32 childIdx = 0;
for (auto& child : mChildren)
{
elementAreas[childIdx].width = sizeRanges[childIdx].optimal.x;
if (child->_getType() == GUIElementBase::Type::FixedSpace)
{
processedElements[childIdx] = true;
}
else if (child->_getType() == GUIElementBase::Type::FlexibleSpace)
{
if (child->_isActive())
{
numFlexibleSpaces++;
numNonClampedElements++;
}
else
processedElements[childIdx] = true;
}
else
{
const GUIDimensions& dimensions = child->_getDimensions();
if (dimensions.fixedWidth())
processedElements[childIdx] = true;
else
{
if (elementAreas[childIdx].width > 0)
{
numNonClampedElements++;
totalNonClampedSize += elementAreas[childIdx].width;
}
else
processedElements[childIdx] = true;
}
}
childIdx++;
}
// If there is some room left, calculate flexible space sizes (since they will fill up all that extra room)
if ((UINT32)layoutArea.width > totalOptimalSize)
{
UINT32 extraSize = layoutArea.width - totalOptimalSize;
UINT32 remainingSize = extraSize;
// Flexible spaces always expand to fill up all unused space
if (numFlexibleSpaces > 0)
{
float avgSize = remainingSize / (float)numFlexibleSpaces;
childIdx = 0;
for (auto& child : mChildren)
{
if (processedElements[childIdx])
{
childIdx++;
continue;
}
UINT32 extraWidth = std::min((UINT32)Math::ceilToInt(avgSize), remainingSize);
UINT32 elementWidth = elementAreas[childIdx].width + extraWidth;
// Clamp if needed
if (child->_getType() == GUIElementBase::Type::FlexibleSpace)
{
processedElements[childIdx] = true;
numNonClampedElements--;
elementAreas[childIdx].width = elementWidth;
remainingSize = (UINT32)std::max(0, (INT32)remainingSize - (INT32)extraWidth);
}
childIdx++;
}
totalOptimalSize = layoutArea.width;
}
}
// Determine weight scale for every element. When scaling elements up/down they will be scaled based on this weight.
// Weight is to ensure all elements are scaled fairly, so elements that are large will get effected more than smaller elements.
childIdx = 0;
float invOptimalSize = 1.0f / totalNonClampedSize;
UINT32 childCount = (UINT32)mChildren.size();
for (UINT32 i = 0; i < childCount; i++)
{
if (processedElements[childIdx])
{
childIdx++;
continue;
}
elementScaleWeights[childIdx] = invOptimalSize * elementAreas[childIdx].width;
childIdx++;
}
// Our optimal size is larger than maximum allowed, so we need to reduce size of some elements
if (totalOptimalSize > (UINT32)layoutArea.width)
{
UINT32 extraSize = totalOptimalSize - layoutArea.width;
UINT32 remainingSize = extraSize;
// Iterate until we reduce everything so it fits, while maintaining
// equal average sizes using the weights we calculated earlier
while (remainingSize > 0 && numNonClampedElements > 0)
{
UINT32 totalRemainingSize = remainingSize;
childIdx = 0;
for (auto& child : mChildren)
{
if (processedElements[childIdx])
{
childIdx++;
continue;
}
float avgSize = totalRemainingSize * elementScaleWeights[childIdx];
UINT32 extraWidth = std::min((UINT32)Math::ceilToInt(avgSize), remainingSize);
UINT32 elementWidth = (UINT32)std::max(0, (INT32)elementAreas[childIdx].width - (INT32)extraWidth);
// Clamp if needed
switch (child->_getType())
{
case GUIElementBase::Type::FlexibleSpace:
elementAreas[childIdx].width = 0;
processedElements[childIdx] = true;
numNonClampedElements--;
break;
case GUIElementBase::Type::Element:
case GUIElementBase::Type::Layout:
case GUIElementBase::Type::Panel:
{
const LayoutSizeRange& childSizeRange = sizeRanges[childIdx];
if (elementWidth == 0)
{
processedElements[childIdx] = true;
numNonClampedElements--;
}
else if (childSizeRange.min.x > 0 && (INT32)elementWidth < childSizeRange.min.x)
{
elementWidth = childSizeRange.min.x;
processedElements[childIdx] = true;
numNonClampedElements--;
}
extraWidth = elementAreas[childIdx].width - elementWidth;
elementAreas[childIdx].width = elementWidth;
remainingSize = (UINT32)std::max(0, (INT32)remainingSize - (INT32)extraWidth);
}
break;
case GUIElementBase::Type::FixedSpace:
break;
}
childIdx++;
}
}
}
else // We are smaller than the allowed maximum, so try to expand some elements
{
UINT32 extraSize = layoutArea.width - totalOptimalSize;
UINT32 remainingSize = extraSize;
// Iterate until we reduce everything so it fits, while maintaining
// equal average sizes using the weights we calculated earlier
while (remainingSize > 0 && numNonClampedElements > 0)
{
UINT32 totalRemainingSize = remainingSize;
childIdx = 0;
for (auto& child : mChildren)
{
if (processedElements[childIdx])
{
childIdx++;
continue;
}
float avgSize = totalRemainingSize * elementScaleWeights[childIdx];
UINT32 extraWidth = std::min((UINT32)Math::ceilToInt(avgSize), remainingSize);
UINT32 elementWidth = elementAreas[childIdx].width + extraWidth;
// Clamp if needed
switch (child->_getType())
{
case GUIElementBase::Type::FlexibleSpace:
processedElements[childIdx] = true;
numNonClampedElements--;
break;
case GUIElementBase::Type::Element:
case GUIElementBase::Type::Layout:
case GUIElementBase::Type::Panel:
{
const LayoutSizeRange& childSizeRange = sizeRanges[childIdx];
if (elementWidth == 0)
{
processedElements[childIdx] = true;
numNonClampedElements--;
}
else if (childSizeRange.max.x > 0 && (INT32)elementWidth > childSizeRange.max.x)
{
elementWidth = childSizeRange.max.x;
processedElements[childIdx] = true;
numNonClampedElements--;
}
extraWidth = elementWidth - elementAreas[childIdx].width;
elementAreas[childIdx].width = elementWidth;
remainingSize = (UINT32)std::max(0, (INT32)remainingSize - (INT32)extraWidth);
}
break;
case GUIElementBase::Type::FixedSpace:
break;
}
childIdx++;
}
}
}
// Compute offsets and height
UINT32 xOffset = 0;
childIdx = 0;
for (auto& child : mChildren)
{
UINT32 elemWidth = elementAreas[childIdx].width;
xOffset += child->_getPadding().left;
const LayoutSizeRange& sizeRange = sizeRanges[childIdx];
UINT32 elemHeight = (UINT32)sizeRange.optimal.y;
const GUIDimensions& dimensions = child->_getDimensions();
if (!dimensions.fixedHeight())
{
elemHeight = layoutArea.height;
if (sizeRange.min.y > 0 && elemHeight < (UINT32)sizeRange.min.y)
elemHeight = (UINT32)sizeRange.min.y;
if (sizeRange.max.y > 0 && elemHeight > (UINT32)sizeRange.max.y)
elemHeight = (UINT32)sizeRange.max.y;
}
elementAreas[childIdx].height = elemHeight;
if (child->_getType() == GUIElementBase::Type::Element)
{
GUIElement* element = static_cast<GUIElement*>(child);
UINT32 yPadding = element->_getPadding().top + element->_getPadding().bottom;
INT32 yOffset = Math::ceilToInt(((INT32)layoutArea.height - (INT32)(elemHeight + yPadding)) * 0.5f);
yOffset = std::max(0, yOffset);
elementAreas[childIdx].x = layoutArea.x + xOffset;
elementAreas[childIdx].y = layoutArea.y + yOffset;
}
else
{
elementAreas[childIdx].x = layoutArea.x + xOffset;
elementAreas[childIdx].y = layoutArea.y;
}
xOffset += elemWidth + child->_getPadding().right;
childIdx++;
}
if (elementScaleWeights != nullptr)
bs_stack_free(elementScaleWeights);
if (processedElements != nullptr)
bs_stack_free(processedElements);
}