Fixing the maximum 2048 meshlet limit

This commit is contained in:
Dynamitos
2025-03-09 22:42:05 +01:00
parent 913b8391f8
commit 6f0e2fe7e7
16 changed files with 151 additions and 173 deletions
+24 -24
View File
@@ -135,10 +135,10 @@ void BasePass::beginFrame(const Component::Camera& cam) {
}
void BasePass::render() {
/*opaqueCulling->updateBuffer(0, 0, oLightIndexList);
opaqueCulling->updateTexture(1, 0, oLightGrid);
transparentCulling->updateBuffer(0, 0, tLightIndexList);
transparentCulling->updateTexture(1, 0, tLightGrid);
opaqueCulling->updateBuffer(LIGHTINDEX_NAME, 0, oLightIndexList);
opaqueCulling->updateTexture(LIGHTGRID_NAME, 0, oLightGrid);
transparentCulling->updateBuffer(LIGHTINDEX_NAME, 0, tLightIndexList);
transparentCulling->updateTexture(LIGHTGRID_NAME, 0, tLightGrid);
opaqueCulling->writeChanges();
transparentCulling->writeChanges();
@@ -153,7 +153,7 @@ void BasePass::render() {
// Base Rendering
for (VertexData* vertexData : VertexData::getList()) {
transparentData.addAll(vertexData->getTransparentData());
vertexData->getInstanceDataSet()->updateBuffer(6, 0, cullingBuffer);
vertexData->getInstanceDataSet()->updateBuffer(VertexData::CULLINGDATA_NAME, 0, cullingBuffer);
vertexData->getInstanceDataSet()->writeChanges();
permutation.setVertexData(vertexData->getTypeName());
const auto& materials = vertexData->getMaterialData();
@@ -231,7 +231,7 @@ void BasePass::render() {
Gfx::SE_SHADER_STAGE_FRAGMENT_BIT,
0, sizeof(VertexData::DrawCallOffsets), &drawCall.offsets);
if (graphics->supportMeshShading()) {
command->drawMesh(drawCall.instanceMeshData.size(), 1, 1);
//command->drawMesh(drawCall.instanceMeshData.size(), 1, 1);
} else {
command->bindIndexBuffer(vertexData->getIndexBuffer());
for (const auto& meshData : drawCall.instanceMeshData) {
@@ -247,9 +247,8 @@ void BasePass::render() {
//commands.add(waterRenderer->render(viewParamsSet));
//commands.add(terrainRenderer->render(viewParamsSet));
*/
// Skybox
graphics->beginRenderPass(renderPass);
{
Gfx::ORenderCommand skyboxCommand = graphics->createRenderCommand("SkyboxRender");
skyboxCommand->setViewport(viewport);
@@ -258,8 +257,6 @@ void BasePass::render() {
skyboxCommand->draw(36, 1, 0, 0);
graphics->executeCommands(std::move(skyboxCommand));
}
graphics->endRenderPass();
/*
// Transparent rendering
{
permutation.setDepthCulling(false); // ignore visibility infos for transparency
@@ -348,7 +345,7 @@ void BasePass::render() {
Gfx::SE_SHADER_STAGE_FRAGMENT_BIT,
0, sizeof(VertexData::DrawCallOffsets), &t.offsets);
if (graphics->supportMeshShading()) {
transparentCommand->drawMesh(1, 1, 1);
//transparentCommand->drawMesh(1, 1, 1);
} else {
// command->bindIndexBuffer(t.vertexData->getIndexBuffer());
// for (const auto& meshData : drawCall.instanceMeshData) {
@@ -376,37 +373,40 @@ void BasePass::render() {
query->endQuery();
timestamps->write(Gfx::SE_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, "BaseEnd");
gDebugVertices.clear();
*/
}
void BasePass::endFrame() {}
void BasePass::publishOutputs() {
TextureCreateInfo depthBufferInfo = {
basePassDepth = graphics->createTexture2D(TextureCreateInfo{
.format = Gfx::SE_FORMAT_D32_SFLOAT,
.width = viewport->getOwner()->getFramebufferWidth(),
.height = viewport->getOwner()->getFramebufferHeight(),
.usage = Gfx::SE_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT,
};
basePassDepth = graphics->createTexture2D(depthBufferInfo);
});
TextureCreateInfo msDepthInfo = {
basePassColor = graphics->createTexture2D(TextureCreateInfo{
.format = Gfx::SE_FORMAT_R32G32B32A32_SFLOAT,
.width = viewport->getOwner()->getFramebufferWidth(),
.height = viewport->getOwner()->getFramebufferHeight(),
.usage = Gfx::SE_IMAGE_USAGE_COLOR_ATTACHMENT_BIT,
});
msBasePassDepth = graphics->createTexture2D(TextureCreateInfo{
.format = Gfx::SE_FORMAT_D32_SFLOAT,
.width = viewport->getOwner()->getFramebufferWidth(),
.height = viewport->getOwner()->getFramebufferHeight(),
.samples = viewport->getSamples(),
.usage = Gfx::SE_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT | Gfx::SE_IMAGE_USAGE_TRANSIENT_ATTACHMENT_BIT,
};
msBasePassDepth = graphics->createTexture2D(msDepthInfo);
TextureCreateInfo msBaseColorInfo = {
.format = viewport->getOwner()->getSwapchainFormat(),
});
msBasePassColor = graphics->createTexture2D(TextureCreateInfo{
.format = Gfx::SE_FORMAT_R32G32B32A32_SFLOAT,
.width = viewport->getOwner()->getFramebufferWidth(),
.height = viewport->getOwner()->getFramebufferHeight(),
.samples = viewport->getSamples(),
.usage = Gfx::SE_IMAGE_USAGE_COLOR_ATTACHMENT_BIT | Gfx::SE_IMAGE_USAGE_TRANSIENT_ATTACHMENT_BIT,
};
msBasePassColor = graphics->createTexture2D(msBaseColorInfo);
});
depthAttachment =
Gfx::RenderTargetAttachment(basePassDepth, Gfx::SE_IMAGE_LAYOUT_UNDEFINED, Gfx::SE_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL,
@@ -418,7 +418,7 @@ void BasePass::publishOutputs() {
msDepthAttachment.clear.depthStencil.depth = 0.0f;
colorAttachment = Gfx::RenderTargetAttachment(viewport, Gfx::SE_IMAGE_LAYOUT_UNDEFINED, Gfx::SE_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL,
Gfx::SE_ATTACHMENT_LOAD_OP_DONT_CARE, Gfx::SE_ATTACHMENT_STORE_OP_DONT_CARE);
Gfx::SE_ATTACHMENT_LOAD_OP_DONT_CARE, Gfx::SE_ATTACHMENT_STORE_OP_STORE);
msColorAttachment =
Gfx::RenderTargetAttachment(msBasePassColor, Gfx::SE_IMAGE_LAYOUT_UNDEFINED, Gfx::SE_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL,
@@ -19,6 +19,7 @@ class BasePass : public RenderPass {
virtual void createRenderPass() override;
private:
// hdr
Gfx::RenderTargetAttachment msColorAttachment;
Gfx::RenderTargetAttachment colorAttachment;
Gfx::RenderTargetAttachment msDepthAttachment;
@@ -37,7 +38,9 @@ class BasePass : public RenderPass {
// use a different texture here so we can do multisampling
Gfx::OTexture2D msBasePassDepth;
Gfx::OTexture2D basePassDepth;
// hdr
Gfx::OTexture2D msBasePassColor;
Gfx::OTexture2D basePassColor;
// used for transparency sorting
Vector cameraPos;
@@ -43,7 +43,7 @@ CachedDepthPass::~CachedDepthPass() {}
void CachedDepthPass::beginFrame(const Component::Camera& cam) { RenderPass::beginFrame(cam); }
void CachedDepthPass::render() {
/* query->beginQuery();
query->beginQuery();
timestamps->write(Gfx::SE_PIPELINE_STAGE_TOP_OF_PIPE_BIT, "CachedBegin");
graphics->beginRenderPass(renderPass);
Array<Gfx::ORenderCommand> commands;
@@ -53,7 +53,7 @@ void CachedDepthPass::render() {
permutation.setDepthCulling(true);
for (VertexData* vertexData : VertexData::getList()) {
permutation.setVertexData(vertexData->getTypeName());
vertexData->getInstanceDataSet()->updateBuffer(6, 0, cullingBuffer);
vertexData->getInstanceDataSet()->updateBuffer(VertexData::CULLINGDATA_NAME, 0, cullingBuffer);
vertexData->getInstanceDataSet()->writeChanges();
// Create Pipeline(VertexData)
@@ -134,9 +134,8 @@ void CachedDepthPass::render() {
graphics->executeCommands(std::move(commands));
graphics->endRenderPass();
graphics->waitDeviceIdle();
timestamps->write(Gfx::SE_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, "CachedEnd");
query->endQuery();*/
query->endQuery();
}
void CachedDepthPass::endFrame() {}
@@ -68,7 +68,7 @@ DepthCullingPass::~DepthCullingPass() {}
void DepthCullingPass::beginFrame(const Component::Camera& cam) { RenderPass::beginFrame(cam); }
void DepthCullingPass::render() {
/*query->beginQuery();
query->beginQuery();
depthAttachment.getTexture()->changeLayout(Gfx::SE_IMAGE_LAYOUT_GENERAL, Gfx::SE_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
Gfx::SE_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT, Gfx::SE_ACCESS_SHADER_READ_BIT,
Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
@@ -125,7 +125,7 @@ void DepthCullingPass::render() {
permutation.setDepthCulling(getGlobals().useDepthCulling);
for (VertexData* vertexData : VertexData::getList()) {
permutation.setVertexData(vertexData->getTypeName());
vertexData->getInstanceDataSet()->updateBuffer(6, 0, cullingBuffer);
vertexData->getInstanceDataSet()->updateBuffer(VertexData::CULLINGDATA_NAME, 0, cullingBuffer);
vertexData->getInstanceDataSet()->writeChanges();
// Create Pipeline(VertexData)
// Descriptors:
@@ -217,7 +217,7 @@ void DepthCullingPass::render() {
// Sync visibility write with compute read
visibilityAttachment.getTexture()->pipelineBarrier(Gfx::SE_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
Gfx::SE_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, Gfx::SE_ACCESS_SHADER_READ_BIT,
Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);*/
Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
}
void DepthCullingPass::endFrame() {}
@@ -37,19 +37,19 @@ void LightCullingPass::beginFrame(const Component::Camera& cam) {
}
void LightCullingPass::render() {
/*query->beginQuery();
query->beginQuery();
timestamps->write(Gfx::SE_PIPELINE_STAGE_TOP_OF_PIPE_BIT, "LightCullBegin");
oLightGrid->pipelineBarrier(Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_FRAGMENT_SHADER_BIT, Gfx::SE_ACCESS_SHADER_WRITE_BIT,
Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
tLightGrid->pipelineBarrier(Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_FRAGMENT_SHADER_BIT, Gfx::SE_ACCESS_SHADER_WRITE_BIT,
Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
cullingDescriptorSet->updateTexture(0, 0, depthAttachment);
cullingDescriptorSet->updateBuffer(1, 0, oLightIndexCounter);
cullingDescriptorSet->updateBuffer(2, 0, tLightIndexCounter);
cullingDescriptorSet->updateBuffer(3, 0, oLightIndexList);
cullingDescriptorSet->updateBuffer(4, 0, tLightIndexList);
cullingDescriptorSet->updateTexture(5, 0, oLightGrid);
cullingDescriptorSet->updateTexture(6, 0, tLightGrid);
cullingDescriptorSet->updateTexture(DEPTHATTACHMENT_NAME, 0, depthAttachment);
cullingDescriptorSet->updateBuffer(OLIGHTINDEXCOUNTER_NAME, 0, oLightIndexCounter);
cullingDescriptorSet->updateBuffer(TLIGHTINDEXCOUNTER_NAME, 0, tLightIndexCounter);
cullingDescriptorSet->updateBuffer(OLIGHTINDEXLIST_NAME, 0, oLightIndexList);
cullingDescriptorSet->updateBuffer(TLIGHTINDEXLIST_NAME, 0, tLightIndexList);
cullingDescriptorSet->updateTexture(OLIGHTGRID_NAME, 0, oLightGrid);
cullingDescriptorSet->updateTexture(TLIGHTGRID_NAME, 0, tLightGrid);
cullingDescriptorSet->writeChanges();
Gfx::OComputeCommand computeCommand = graphics->createComputeCommand("CullingCommand");
if (getGlobals().useLightCulling) {
@@ -58,7 +58,7 @@ void LightCullingPass::render() {
computeCommand->bindPipeline(cullingPipeline);
}
computeCommand->bindDescriptor({viewParamsSet, dispatchParamsSet, cullingDescriptorSet, lightEnv->getDescriptorSet()});
computeCommand->dispatch(dispatchParams.numThreadGroups.x, dispatchParams.numThreadGroups.y, dispatchParams.numThreadGroups.z);
computeCommand->dispatch(numThreadGroups.x, numThreadGroups.y, numThreadGroups.z);
Array<Gfx::OComputeCommand> commands;
commands.add(std::move(computeCommand));
// std::cout << "Execute" << std::endl;
@@ -72,7 +72,7 @@ void LightCullingPass::render() {
oLightGrid->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT, Gfx::SE_ACCESS_SHADER_READ_BIT,
Gfx::SE_PIPELINE_STAGE_FRAGMENT_SHADER_BIT);
tLightGrid->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT, Gfx::SE_ACCESS_SHADER_READ_BIT,
Gfx::SE_PIPELINE_STAGE_FRAGMENT_SHADER_BIT);*/
Gfx::SE_PIPELINE_STAGE_FRAGMENT_SHADER_BIT);
}
void LightCullingPass::endFrame() {}
@@ -81,8 +81,8 @@ void LightCullingPass::publishOutputs() {
setupFrustums();
uint32_t viewportWidth = viewport->getWidth();
uint32_t viewportHeight = viewport->getHeight();
UVector4 numThreadGroups = glm::ceil(glm::vec4(viewportWidth / (float)BLOCK_SIZE, viewportHeight / (float)BLOCK_SIZE, 1, 0));
UVector4 numThreads = numThreadGroups * glm::uvec4(BLOCK_SIZE, BLOCK_SIZE, 1, 0);
numThreadGroups = glm::ceil(glm::vec4(viewportWidth / (float)BLOCK_SIZE, viewportHeight / (float)BLOCK_SIZE, 1, 0));
numThreads = numThreadGroups * glm::uvec4(BLOCK_SIZE, BLOCK_SIZE, 1, 0);
dispatchParamsSet = dispatchParamsLayout->allocateDescriptorSet();
dispatchParamsSet->updateConstants("numThreadGroups", 0, &numThreadGroups);
dispatchParamsSet->updateConstants("numThreads", 0, &numThreads);
@@ -225,8 +225,8 @@ void LightCullingPass::setupFrustums() {
uint32_t viewportWidth = viewport->getWidth();
uint32_t viewportHeight = viewport->getHeight();
glm::uvec4 numThreads = glm::ceil(glm::vec4(viewportWidth / (float)BLOCK_SIZE, viewportHeight / (float)BLOCK_SIZE, 1, 0));
glm::uvec4 numThreadGroups = glm::ceil(glm::vec4(numThreads.x / (float)BLOCK_SIZE, numThreads.y / (float)BLOCK_SIZE, 1, 0));
numThreads = glm::ceil(glm::vec4(viewportWidth / (float)BLOCK_SIZE, viewportHeight / (float)BLOCK_SIZE, 1, 0));
numThreadGroups = glm::ceil(glm::vec4(numThreads.x / (float)BLOCK_SIZE, numThreads.y / (float)BLOCK_SIZE, 1, 0));
RenderPass::beginFrame(Component::Camera());
@@ -59,7 +59,9 @@ class LightCullingPass : public RenderPass {
Gfx::PComputePipeline cullingEnabledPipeline;
Gfx::OPipelineStatisticsQuery query;
Gfx::PTimestampQuery timestamps;
UVector4 numThreadGroups;
UVector4 numThreads;
PScene scene;
};
DEFINE_REF(LightCullingPass)
@@ -13,11 +13,12 @@ void VisibilityPass::beginFrame(const Component::Camera& cam) {
}
void VisibilityPass::render() {
/*cullingBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_MESH_SHADER_BIT_EXT,
cullingBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_MESH_SHADER_BIT_EXT,
Gfx::SE_ACCESS_TRANSFER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_TRANSFER_BIT);
cullingBuffer->clear();
cullingBuffer->pipelineBarrier(Gfx::SE_ACCESS_TRANSFER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_TRANSFER_BIT, Gfx::SE_ACCESS_SHADER_READ_BIT,
cullingBuffer->pipelineBarrier(Gfx::SE_ACCESS_TRANSFER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_TRANSFER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT | Gfx::SE_ACCESS_SHADER_WRITE_BIT,
Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
visibilityDescriptor->reset();
visibilitySet = visibilityDescriptor->allocateDescriptorSet();
@@ -38,7 +39,7 @@ void VisibilityPass::render() {
query->endQuery();
cullingBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT,
Gfx::SE_PIPELINE_STAGE_TASK_SHADER_BIT_EXT | Gfx::SE_PIPELINE_STAGE_MESH_SHADER_BIT_EXT);*/
Gfx::SE_PIPELINE_STAGE_TASK_SHADER_BIT_EXT | Gfx::SE_PIPELINE_STAGE_MESH_SHADER_BIT_EXT);
}
void VisibilityPass::endFrame() {}