英文:
VULKAN - Uniform Buffers (Dynamic) - Rendering freezes and app crashes
问题
我昨晚成功实现了动态统一缓冲区。
在下面的代码中，我初始化了一个动态偏移量，最初设置了投影/视图，然后在循环中将动态偏移量设置为i * bufferSize，ubo的pvm矩阵，然后绑定描述符集并绘制。
这个方法对我来说非常有效，我尝试了一个非常随机的魔术数字（21），它运行得很顺利。但是，如果我尝试22或更高的数字，程序就会冻结和崩溃。只是好奇，我的方法是否不正确？如果我无法超过21个循环，肯定是我遗漏了什么。我使用的是GTX 3060。
context中的bufferSize是uniformBufferObject结构的大小，它只包含三个mat4矩阵（64 * 3 = 192）。
尝试使用动态UBO渲染数百个球体，但当我尝试访问uniformBufferObject[索引= bufferSize（192）* 21（4032）]时，应用程序会崩溃。
创建统一缓冲区的代码如下：
创建缓冲区的函数：
findMemoryType函数：
UBO结构体，描述符集布局，描述符池，描述符写入：
英文:
I got dynamic uniform buffers working last night.
In the code below, I initialize a dynamic offset, set the projection/view initially, then in a loop, set the dynamic offset to be i * the bufferSize, ubo's pvm matricies, then bind the descriptor set and draw.
This thing worked like a charm with a very random magic number I tried (21). But if I try 22 or higher, the thing just freezes and crashes. Just curious, is my approach to this incorrect? I must be missing something if I can't go over 21 loops. I have a GTX 3060 as well.
bufferSize for context is the size of the uniformBufferObject struct, which is just three mat4s (64 * 3 = 192).
uint32_t dynamicOffset = 0;glm::mat4 proj = glm::perspective(fov, (float)screenWidth / (float)screenHeight, zNear, zFar);proj[1][1] *= -1;glm::mat4 view = glm::lookAt(glm::vec3(0.0f, 0.0f, 2.0f), glm::vec3(0.0f, 0.0f, 0.0f), glm::vec3(0.0f, 1.0f, 0.0f));for (int i = 0; i < 21; i++){dynamicOffset = bufferSize * i;// Update and draw the cubeuniformBufferObject[i].proj = proj;uniformBufferObject[i].view = view;uniformBufferObject[i].model = glm::mat4(1.0f);uniformBufferObject[i].model = glm::translate(uniformBufferObject[i].model, glm::vec3(0.0f, 0.0f, -3.0f));uniformBufferObject[i].model = glm::rotate(uniformBufferObject[i].model, 0.0f, glm::vec3(1.0f, 1.0f, 1.0f));uniformBufferObject[i].model = glm::scale(uniformBufferObject[i].model, glm::vec3(0.45f));vkCmdBindDescriptorSets(commandBuffer, VK_PIPELINE_BIND_POINT_GRAPHICS, graphicsPipelineLayout, 0, 1, &descriptorSet, 1, &dynamicOffset);vkCmdDraw(commandBuffer, static_cast<uint32_t>(vertices.size()), 1, 0, 0);}
Trying to render hundreds of spheres using a dynamic UBO but when I try to access uniformBufferObject[where index = bufferSize (192) * 21 (4032)], the application crashes.
Uniform buffer creation code is here:
void createUniformBuffer(){createBuffer(bufferSize, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT, VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT, uniformBuffer, uniformBufferMemory);vkMapMemory(device, uniformBufferMemory, 0, bufferSize, 0, &uniformBufferMapped);uniformBufferObject = static_cast<UniformBufferObject*>(uniformBufferMapped);}
Create buffer function:
void createBuffer(VkDeviceSize size, VkBufferUsageFlags usage, VkMemoryPropertyFlags properties, VkBuffer& buffer, VkDeviceMemory& bufferMemory){VkBufferCreateInfo bufferInfo{};bufferInfo.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;bufferInfo.size = size;bufferInfo.usage = usage;bufferInfo.sharingMode = VK_SHARING_MODE_EXCLUSIVE;vkCreateBuffer(device, &bufferInfo, nullptr, &buffer);VkMemoryRequirements memRequirements;vkGetBufferMemoryRequirements(device, buffer, &memRequirements);VkMemoryAllocateInfo allocInfo{};allocInfo.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO;allocInfo.allocationSize = memRequirements.size;allocInfo.memoryTypeIndex = findMemoryType(memRequirements.memoryTypeBits, properties);vkAllocateMemory(device, &allocInfo, nullptr, &bufferMemory);vkBindBufferMemory(device, buffer, bufferMemory, 0);}
findMemoryType function:
uint32_t findMemoryType(uint32_t typeFilter, VkMemoryPropertyFlags properties) {VkPhysicalDeviceMemoryProperties memProperties;vkGetPhysicalDeviceMemoryProperties(physicalDevice, &memProperties);for (uint32_t i = 0; i < memProperties.memoryTypeCount; i++) {if ((typeFilter & (1 << i)) && (memProperties.memoryTypes[i].propertyFlags & properties) == properties) {return i;}}}
ubo struct, descriptorsetlayout, descriptorpool, descriptorwrite:
struct UniformBufferObject{glm::mat4 model;glm::mat4 view;glm::mat4 proj;};VkDescriptorSetLayoutBinding uboLayoutBinding{};uboLayoutBinding.binding = 0;uboLayoutBinding.descriptorCount = 1;uboLayoutBinding.descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC;uboLayoutBinding.stageFlags = VK_SHADER_STAGE_VERTEX_BIT;std::array<VkDescriptorPoolSize, 2> poolSizes{};poolSizes[0].type = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC;poolSizes[0].descriptorCount = static_cast<uint32_t>(1);poolSizes[1].type = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;poolSizes[1].descriptorCount = static_cast<uint32_t>(1);descriptorWrites[0].sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET;descriptorWrites[0].dstSet = descriptorSet;descriptorWrites[0].dstBinding = 0;descriptorWrites[0].dstArrayElement = 0;descriptorWrites[0].descriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC;descriptorWrites[0].descriptorCount = 1;descriptorWrites[0].pBufferInfo = &bufferInfo;
答案1
得分: 1
createBuffer(bufferSize, ...
假设这个bufferSize与填充结构的循环中使用的bufferSize相同，那么它需要变成
createBuffer(bufferSize * <你想要激活的绘制数量>, ...)
你正在分配一个单一的192字节缓冲区，恰好位于页面的开头，所以你可以在缓冲区的末尾之前写入，直到触及页面的末尾，然后出现错误。
英文:
> createBuffer(bufferSize, ...
Assuming this is the same bufferSize as your loop that populates the structure, this needs to be
createBuffer(bufferSize * <whatever number of draws you want active>, ...
You're allocating a single 192 byte buffer, which happens to be at the start of a page, and so you you are able write over the end of the buffer until you hit the end of the page which then faults.
通过集体智慧和协作来改善编程学习和解决问题的方式。致力于成为全球开发者共同参与的知识库，让每个人都能够通过互相帮助和分享经验来进步。
评论