Files
QuaternionEngine/src/render/vk_renderpass_geometry.cpp
2025-11-03 14:37:52 +09:00

298 lines
11 KiB
C++

#include "vk_renderpass_geometry.h"
#include <chrono>
#include <unordered_set>
#include "frame_resources.h"
#include "texture_cache.h"
#include "vk_descriptor_manager.h"
#include "vk_device.h"
#include "core/engine_context.h"
#include "core/vk_initializers.h"
#include "core/vk_resource.h"
#include "vk_mem_alloc.h"
#include "vk_scene.h"
#include "vk_swapchain.h"
#include "render/rg_graph.h"
// Basic conservative frustum test against RenderObject AABB.
// Clip space uses Vulkan Z0 (0..w). Returns true if any part of the box is inside.
bool is_visible(const RenderObject &obj, const glm::mat4 &viewproj)
{
const std::array<glm::vec3, 8> corners{
glm::vec3{+1, +1, +1}, glm::vec3{+1, +1, -1}, glm::vec3{+1, -1, +1}, glm::vec3{+1, -1, -1},
glm::vec3{-1, +1, +1}, glm::vec3{-1, +1, -1}, glm::vec3{-1, -1, +1}, glm::vec3{-1, -1, -1},
};
const glm::vec3 o = obj.bounds.origin;
const glm::vec3 e = obj.bounds.extents;
const glm::mat4 m = viewproj * obj.transform; // world -> clip
glm::vec4 clip[8];
for (int i = 0; i < 8; ++i)
{
const glm::vec3 p = o + corners[i] * e;
clip[i] = m * glm::vec4(p, 1.f);
}
auto all_out = [&](auto pred) {
for (int i = 0; i < 8; ++i)
{
if (!pred(clip[i])) return false;
}
return true;
};
// Clip volume in Vulkan (ZO): -w<=x<=w, -w<=y<=w, 0<=z<=w
if (all_out([](const glm::vec4 &v) { return v.x < -v.w; })) return false; // left
if (all_out([](const glm::vec4 &v) { return v.x > v.w; })) return false; // right
if (all_out([](const glm::vec4 &v) { return v.y < -v.w; })) return false; // bottom
if (all_out([](const glm::vec4 &v) { return v.y > v.w; })) return false; // top
if (all_out([](const glm::vec4 &v) { return v.z < 0.0f; })) return false; // near (ZO)
if (all_out([](const glm::vec4 &v) { return v.z > v.w; })) return false; // far
return true; // intersects or is fully inside
}
void GeometryPass::init(EngineContext *context)
{
_context = context;
}
void GeometryPass::execute(VkCommandBuffer)
{
// Geometry is executed via the render graph now.
}
void GeometryPass::register_graph(RenderGraph *graph,
RGImageHandle gbufferPosition,
RGImageHandle gbufferNormal,
RGImageHandle gbufferAlbedo,
RGImageHandle depthHandle)
{
if (!graph || !gbufferPosition.valid() || !gbufferNormal.valid() || !gbufferAlbedo.valid() || !depthHandle.valid())
{
return;
}
graph->add_pass(
"Geometry",
RGPassType::Graphics,
[gbufferPosition, gbufferNormal, gbufferAlbedo, depthHandle](RGPassBuilder &builder, EngineContext *ctx)
{
VkClearValue clear{};
clear.color = {{0.f, 0.f, 0.f, 0.f}};
builder.write_color(gbufferPosition, true, clear);
builder.write_color(gbufferNormal, true, clear);
builder.write_color(gbufferAlbedo, true, clear);
// Reverse-Z: clear depth to 0.0
VkClearValue depthClear{};
depthClear.depthStencil = {0.f, 0};
builder.write_depth(depthHandle, true, depthClear);
// Register read buffers used by all draw calls (index + vertex SSBO)
if (ctx)
{
const DrawContext &dc = ctx->getMainDrawContext();
// Collect unique buffers to avoid duplicates
std::unordered_set<VkBuffer> indexSet;
std::unordered_set<VkBuffer> vertexSet;
indexSet.reserve(dc.OpaqueSurfaces.size() + dc.TransparentSurfaces.size());
vertexSet.reserve(dc.OpaqueSurfaces.size() + dc.TransparentSurfaces.size());
auto collect = [&](const std::vector<RenderObject>& v){
for (const auto &r : v)
{
if (r.indexBuffer) indexSet.insert(r.indexBuffer);
if (r.vertexBuffer) vertexSet.insert(r.vertexBuffer);
}
};
collect(dc.OpaqueSurfaces);
collect(dc.TransparentSurfaces);
for (VkBuffer b : indexSet)
builder.read_buffer(b, RGBufferUsage::IndexRead, 0, "geom.index");
for (VkBuffer b : vertexSet)
builder.read_buffer(b, RGBufferUsage::StorageRead, 0, "geom.vertex");
}
},
[this, gbufferPosition, gbufferNormal, gbufferAlbedo, depthHandle](VkCommandBuffer cmd,
const RGPassResources &res,
EngineContext *ctx)
{
draw_geometry(cmd, ctx, res, gbufferPosition, gbufferNormal, gbufferAlbedo, depthHandle);
});
}
void GeometryPass::draw_geometry(VkCommandBuffer cmd,
EngineContext *context,
const RGPassResources &resources,
RGImageHandle gbufferPosition,
RGImageHandle gbufferNormal,
RGImageHandle gbufferAlbedo,
RGImageHandle depthHandle) const
{
EngineContext *ctxLocal = context ? context : _context;
if (!ctxLocal || !ctxLocal->currentFrame) return;
ResourceManager *resourceManager = ctxLocal->getResources();
DeviceManager *deviceManager = ctxLocal->getDevice();
DescriptorManager *descriptorLayouts = ctxLocal->getDescriptorLayouts();
if (!resourceManager || !deviceManager || !descriptorLayouts) return;
VkImageView positionView = resources.image_view(gbufferPosition);
VkImageView normalView = resources.image_view(gbufferNormal);
VkImageView albedoView = resources.image_view(gbufferAlbedo);
VkImageView depthView = resources.image_view(depthHandle);
if (positionView == VK_NULL_HANDLE || normalView == VK_NULL_HANDLE ||
albedoView == VK_NULL_HANDLE || depthView == VK_NULL_HANDLE)
{
return;
}
const auto& mainDrawContext = ctxLocal->getMainDrawContext();
const auto& sceneData = ctxLocal->getSceneData();
VkExtent2D drawExtent = ctxLocal->getDrawExtent();
auto start = std::chrono::system_clock::now();
std::vector<uint32_t> opaque_draws;
opaque_draws.reserve(mainDrawContext.OpaqueSurfaces.size());
for (int i = 0; i < mainDrawContext.OpaqueSurfaces.size(); i++)
{
if (is_visible(mainDrawContext.OpaqueSurfaces[i], sceneData.viewproj))
{
opaque_draws.push_back(i);
}
}
std::sort(opaque_draws.begin(), opaque_draws.end(), [&](const auto &iA, const auto &iB)
{
const RenderObject &A = mainDrawContext.OpaqueSurfaces[iA];
const RenderObject &B = mainDrawContext.OpaqueSurfaces[iB];
if (A.material == B.material)
{
return A.indexBuffer < B.indexBuffer;
}
return A.material < B.material;
});
// Dynamic rendering is now begun by the RenderGraph using the declared attachments.
AllocatedBuffer gpuSceneDataBuffer = resourceManager->create_buffer(sizeof(GPUSceneData),
VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT,
VMA_MEMORY_USAGE_CPU_TO_GPU);
ctxLocal->currentFrame->_deletionQueue.push_function([resourceManager, gpuSceneDataBuffer]()
{
resourceManager->destroy_buffer(gpuSceneDataBuffer);
});
VmaAllocationInfo allocInfo{};
vmaGetAllocationInfo(deviceManager->allocator(), gpuSceneDataBuffer.allocation, &allocInfo);
auto *sceneUniformData = static_cast<GPUSceneData *>(allocInfo.pMappedData);
*sceneUniformData = sceneData;
vmaFlushAllocation(deviceManager->allocator(), gpuSceneDataBuffer.allocation, 0, sizeof(GPUSceneData));
VkDescriptorSet globalDescriptor = ctxLocal->currentFrame->_frameDescriptors.allocate(
deviceManager->device(), descriptorLayouts->gpuSceneDataLayout());
DescriptorWriter writer;
writer.write_buffer(0, gpuSceneDataBuffer.buffer, sizeof(GPUSceneData), 0, VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER);
writer.update_set(deviceManager->device(), globalDescriptor);
MaterialPipeline *lastPipeline = nullptr;
MaterialInstance *lastMaterial = nullptr;
VkBuffer lastIndexBuffer = VK_NULL_HANDLE;
auto draw = [&](const RenderObject &r)
{
if (r.material != lastMaterial)
{
lastMaterial = r.material;
if (r.material->pipeline != lastPipeline)
{
lastPipeline = r.material->pipeline;
vkCmdBindPipeline(cmd, VK_PIPELINE_BIND_POINT_GRAPHICS, r.material->pipeline->pipeline);
vkCmdBindDescriptorSets(cmd, VK_PIPELINE_BIND_POINT_GRAPHICS, r.material->pipeline->layout, 0, 1,
&globalDescriptor, 0, nullptr);
VkViewport viewport{};
viewport.x = 0;
viewport.y = 0;
viewport.width = static_cast<float>(drawExtent.width);
viewport.height = static_cast<float>(drawExtent.height);
viewport.minDepth = 0.f;
viewport.maxDepth = 1.f;
vkCmdSetViewport(cmd, 0, 1, &viewport);
VkRect2D scissor{};
scissor.offset.x = 0;
scissor.offset.y = 0;
scissor.extent.width = drawExtent.width;
scissor.extent.height = drawExtent.height;
vkCmdSetScissor(cmd, 0, 1, &scissor);
}
vkCmdBindDescriptorSets(cmd, VK_PIPELINE_BIND_POINT_GRAPHICS, r.material->pipeline->layout, 1, 1,
&r.material->materialSet, 0, nullptr);
if (ctxLocal->textures)
{
ctxLocal->textures->markSetUsed(r.material->materialSet, ctxLocal->frameIndex);
}
}
if (r.indexBuffer != lastIndexBuffer)
{
lastIndexBuffer = r.indexBuffer;
vkCmdBindIndexBuffer(cmd, r.indexBuffer, 0, VK_INDEX_TYPE_UINT32);
}
GPUDrawPushConstants push_constants{};
push_constants.worldMatrix = r.transform;
push_constants.vertexBuffer = r.vertexBufferAddress;
vkCmdPushConstants(cmd, r.material->pipeline->layout, VK_SHADER_STAGE_VERTEX_BIT, 0,
sizeof(GPUDrawPushConstants), &push_constants);
vkCmdDrawIndexed(cmd, r.indexCount, 1, r.firstIndex, 0, 0);
if (ctxLocal->stats)
{
ctxLocal->stats->drawcall_count++;
ctxLocal->stats->triangle_count += r.indexCount / 3;
}
};
if (ctxLocal->stats)
{
ctxLocal->stats->drawcall_count = 0;
ctxLocal->stats->triangle_count = 0;
}
for (auto &r: opaque_draws)
{
draw(mainDrawContext.OpaqueSurfaces[r]);
}
// Transparent surfaces are rendered in a separate Transparent pass after lighting.
// RenderGraph will end dynamic rendering for this pass.
auto end = std::chrono::system_clock::now();
auto elapsed = std::chrono::duration_cast<std::chrono::microseconds>(end - start);
if (ctxLocal->stats)
{
ctxLocal->stats->mesh_draw_time = elapsed.count() / 1000.f;
}
}
void GeometryPass::cleanup()
{
fmt::print("GeometryPass::cleanup()\n");
}