/**************************************************************************** Copyright (c) 2020-2022 Xiamen Yaji Software Co., Ltd. http://www.cocos.com Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated engine source code (the "Software"), a limited, worldwide, royalty-free, non-assignable, revocable and non-exclusive license to use Cocos Creator solely to develop games on your target platforms. You shall not use Cocos Creator software for developing other software or tools that's used for developing games. You are not granted to publish, distribute, sublicense, and/or sell copies of Cocos Creator. The software or tools in this License Agreement are licensed, not sold. Xiamen Yaji Software Co., Ltd. reserves all rights not expressly granted to you. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. ****************************************************************************/ #include "BatchedBuffer.h" #include "gfx-base/GFXBuffer.h" #include "gfx-base/GFXDescriptorSet.h" #include "gfx-base/GFXDevice.h" #include "gfx-base/GFXInputAssembler.h" namespace cc { namespace pipeline { map> BatchedBuffer::buffers; BatchedBuffer * BatchedBuffer::get(scene::Pass *pass) { return BatchedBuffer::get(pass, 0); } BatchedBuffer *BatchedBuffer::get(scene::Pass *pass, uint extraKey) { auto &record = BatchedBuffer::buffers[pass]; auto &buffer = record[extraKey]; if (buffer == nullptr) buffer = CC_NEW(BatchedBuffer(pass)); return buffer; } void BatchedBuffer::destroyBatchedBuffer() { for (auto &pair : BatchedBuffer::buffers) { const map &bufferItem = pair.second; for (const auto &item : bufferItem) { BatchedBuffer *batchedBuffer = item.second; if (batchedBuffer) { batchedBuffer->destroy(); } } } BatchedBuffer::buffers.clear(); } BatchedBuffer::BatchedBuffer(const scene::Pass *pass) : _pass(pass), _device(gfx::Device::getInstance()) { } BatchedBuffer::~BatchedBuffer() = default; void BatchedBuffer::destroy() { for (auto &batch : _batches) { for (auto *vb : batch.vbs) { CC_SAFE_DESTROY(vb); } for (auto *data : batch.vbDatas) { CC_FREE(data); } CC_SAFE_DESTROY(batch.indexBuffer); CC_SAFE_DESTROY(batch.ia); CC_SAFE_DESTROY(batch.ubo); CC_FREE(batch.indexData); } _batches.clear(); } void BatchedBuffer::merge(const scene::SubModel *subModel, uint passIdx, const scene::Model *model) { const auto *subMesh = subModel->getSubMesh(); const auto &flatBuffers = subMesh->flatBuffers; auto flatBuffersCount = static_cast(flatBuffers.size()); if (0 == flatBuffersCount) { return; } const auto & flatBuffer = flatBuffers[0]; uint vbSize = 0; uint indexSize = 0; const auto vbCount = flatBuffer.count; const auto *const pass = subModel->getPass(passIdx); auto *const shader = subModel->getShader(passIdx); auto *const descriptorSet = subModel->getDescriptorSet(); bool isBatchExist = false; for (auto &batch : _batches) { if (batch.vbs.size() == flatBuffersCount && batch.mergeCount < UBOLocalBatched::BATCHING_COUNT) { isBatchExist = true; for (uint j = 0; j < flatBuffersCount; ++j) { auto *const vb = batch.vbs[j]; if (vb->getStride() != flatBuffers[j].stride) { isBatchExist = false; break; } } if (isBatchExist) { for (uint j = 0; j < flatBuffersCount; ++j) { const auto &flatBuffer = flatBuffers[j]; auto * batchVB = batch.vbs[j]; auto * vbData = batch.vbDatas[j]; const uint vbBufSizeOld = batchVB->getSize(); vbSize = (vbCount + batch.vbCount) * flatBuffer.stride; if (vbSize > vbBufSizeOld) { auto *vbDataNew = static_cast(CC_MALLOC(vbSize)); memcpy(vbDataNew, vbData, vbBufSizeOld); batchVB->resize(vbSize); CC_FREE(vbData); batch.vbDatas[j] = vbDataNew; vbData = vbDataNew; } auto offset = batch.vbCount * flatBuffer.stride; memcpy(vbData + offset, flatBuffer.data, flatBuffer.size); } auto *indexData = batch.indexData; indexSize = (vbCount + batch.vbCount) * sizeof(float); if (indexSize > batch.indexBuffer->getSize()) { auto *newIndexData = static_cast(CC_MALLOC(indexSize)); memcpy(newIndexData, indexData, batch.indexBuffer->getSize()); CC_FREE(indexData); batch.indexData = newIndexData; indexData = batch.indexData; batch.indexBuffer->resize(indexSize); } const auto start = batch.vbCount; const auto end = start + vbCount; const auto mergeCount = batch.mergeCount; if (indexData[start] != mergeCount || indexData[end - 1] != mergeCount) { for (auto j = start; j < end; j++) { indexData[j] = mergeCount + 0.1F; // guard against underflow } } // update world matrix const auto offset = UBOLocalBatched::MAT_WORLDS_OFFSET + batch.mergeCount * 16; const auto &worldMatrix = model->getTransform()->getWorldMatrix(); memcpy(batch.uboData.data() + offset, worldMatrix.m, sizeof(worldMatrix)); if (!batch.mergeCount) { descriptorSet->bindBuffer(UBOLocalBatched::BINDING, batch.ubo); descriptorSet->update(); batch.pass = pass; batch.shader = shader; batch.descriptorSet = descriptorSet; } ++batch.mergeCount; batch.vbCount += vbCount; auto prevCount = batch.ia->getDrawInfo().vertexCount; batch.ia->setVertexCount(prevCount + vbCount); return; } } } // Create a new batch vector vbs(flatBuffersCount, nullptr); vector vbDatas(flatBuffersCount, nullptr); vector totalVBs(flatBuffersCount + 1, nullptr); for (uint i = 0; i < flatBuffersCount; ++i) { const auto &flatBuffer = flatBuffers[i]; auto * newVB = _device->createBuffer({ gfx::BufferUsageBit::VERTEX | gfx::BufferUsageBit::TRANSFER_DST, gfx::MemoryUsageBit::HOST | gfx::MemoryUsageBit::DEVICE, flatBuffer.count * flatBuffer.stride, flatBuffer.stride, }); auto size = 0U; newVB->update(flatBuffer.data, flatBuffer.size); vbs[i] = newVB; vbDatas[i] = static_cast(CC_MALLOC(newVB->getSize())); memset(vbDatas[i], 0, newVB->getSize()); totalVBs[i] = newVB; } const auto indexBufferSize = vbCount * sizeof(float); auto * indexBuffer = _device->createBuffer({ gfx::BufferUsageBit::VERTEX | gfx::BufferUsageBit::TRANSFER_DST, gfx::MemoryUsageBit::HOST | gfx::MemoryUsageBit::DEVICE, static_cast(indexBufferSize), sizeof(float), }); auto * indexData = static_cast(CC_MALLOC(indexBufferSize)); memset(indexData, 0, indexBufferSize); indexBuffer->update(indexData, static_cast(indexBufferSize)); totalVBs[flatBuffersCount] = indexBuffer; vector attributes = subModel->getInputAssembler()->getAttributes(); gfx::Attribute attrib = { "a_dyn_batch_id", gfx::Format::R32F, false, flatBuffersCount, }; attributes.emplace_back(std::move(attrib)); auto *ia = _device->createInputAssembler({std::move(attributes), std::move(totalVBs)}); auto *ubo = _device->createBuffer({ gfx::BufferUsageBit::UNIFORM | gfx::BufferUsageBit::TRANSFER_DST, gfx::MemoryUsageBit::HOST | gfx::MemoryUsageBit::DEVICE, UBOLocalBatched::SIZE, UBOLocalBatched::SIZE, }); descriptorSet->bindBuffer(UBOLocalBatched::BINDING, ubo); descriptorSet->update(); std::array uboData; const auto & worldMatrix = model->getTransform()->getWorldMatrix(); memcpy(uboData.data() + UBOLocalBatched::MAT_WORLDS_OFFSET, worldMatrix.m, sizeof(worldMatrix)); BatchedItem item = { std::move(vbs), //vbs std::move(vbDatas), //vbDatas indexBuffer, //indexBuffer static_cast(indexData), //indexData vbCount, //vbCount 1, //mergeCount ia, //ia ubo, //ubo uboData, //uboData descriptorSet, //descriptorSet pass, //pass shader, //shader }; _batches.emplace_back(std::move(item)); } void BatchedBuffer::clear() { for (auto &batch : _batches) { batch.vbCount = 0; batch.mergeCount = 0; batch.ia->setVertexCount(0); } } void BatchedBuffer::setDynamicOffset(uint idx, uint value) { _dynamicOffsets[idx] = value; } } // namespace pipeline } // namespace cc