Create a Uniform Buffer

Code file for this section is 07-init_uniform_buffer.cpp

Since you created buffers in the recent samples, you might as well work on another one now.

A uniform buffer is a buffer that is made accessible in a read-only fashion to shaders so that the shaders can read constant parameter data.

This is another example of a step that you have to perform in a Vulkan program that you wouldn't have to do in another graphics API. In GLES, you would simply make an API call to set the contents of uniform variables that are sent to the shader. Here, you have to allocate the memory and fill it in.

Setting up the Uniform Data

The sample uses the uniform buffer to pass the MVP (Model-View-Projection) matrix to the shader so that the shader can use it to transform each vertex.

The sample sets it up like this:

info.Projection = glm::perspective(glm::radians(45.0f), 1.0f, 0.1f, 100.0f);
info.View = glm::lookAt(
    glm::vec3(-5, 3, -10), // Camera is at (-5,3,-10), in World Space
    glm::vec3(0, 0, 0),    // and looks at the origin
    glm::vec3(0, -1, 0)    // Head is up (set to 0,-1,0 to look upside-down)
    );
info.Model = glm::mat4(1.0f);
// Vulkan clip space has inverted Y and half Z.
info.Clip = glm::mat4(1.0f,  0.0f, 0.0f, 0.0f,
                      0.0f, -1.0f, 0.0f, 0.0f,
                      0.0f,  0.0f, 0.5f, 0.0f,
                      0.0f,  0.0f, 0.5f, 1.0f);

info.MVP = info.Clip * info.Projection * info.View * info.Model;

Note that the glm library is used here to simplify the code. info.MVP is a 4x4 matrix.

Creating the Uniform Buffer Object

Creating this buffer is fairly similar to how you created the depth buffer in a previous sample, with just a change to the usage:

VkBufferCreateInfo buf_info = {};
buf_info.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
buf_info.pNext = NULL;
buf_info.usage = VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT;
buf_info.size = sizeof(info.MVP);
buf_info.queueFamilyIndexCount = 0;
buf_info.pQueueFamilyIndices = NULL;
buf_info.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
buf_info.flags = 0;
res = vkCreateBuffer(info.device, &buf_info, NULL, &info.uniform_data.buf);
assert(res == VK_SUCCESS);

Allocating the Uniform Buffer Memory

Like the depth buffer, you need to explicitly allocate the memory for the uniform buffer.

VkMemoryRequirements mem_reqs;
vkGetBufferMemoryRequirements(info.device, info.uniform_data.buf,
                              &mem_reqs);

VkMemoryAllocateInfo alloc_info = {};
alloc_info.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO;
alloc_info.pNext = NULL;
alloc_info.memoryTypeIndex = 0;

alloc_info.allocationSize = mem_reqs.size;
pass = memory_type_from_properties(info, mem_reqs.memoryTypeBits,
                                   VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
                                       VK_MEMORY_PROPERTY_HOST_COHERENT_BIT,
                                   &alloc_info.memoryTypeIndex);

res = vkAllocateMemory(info.device, &alloc_info, NULL,
                       &(info.uniform_data.mem));

The VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT communicates that the memory should be mapped so that the CPU (host) can access it.

The VK_MEMORY_PROPERTY_HOST_COHERENT_BIT requests that the writes to the memory by the host are visible to the device (and vice-versa) without the need to flush memory caches. This just makes it a bit simpler to program, since it isn't necessary to call vkFlushMappedMemoryRanges and vkInvalidateMappedMemoryRanges to make sure that the data is visible to the GPU.

Mapping and Setting the Uniform Buffer Memory

You didn't need to initialize the contents of the depth buffer memory when you allocated it in a previous sample. This is because the GPU takes care of reading and writing it. But with the uniform buffer, you need to populate it with the data that you want the shader to read. In this case, the data is the MVP matrix. In order to get CPU access to the memory, you need to map it:

res = vkMapMemory(info.device, info.uniform_data.mem, 0, mem_reqs.size, 0,
                  (void **)&pData);

It is then a pretty simple matter to copy the MVP into the uniform buffer and then unmap it:

memcpy(pData, &info.MVP, sizeof(info.MVP));

vkUnmapMemory(info.device, info.uniform_data.mem);

You can unmap the memory now since we're drawing a single static frame and there is no need to update the MVP matrix again. But if you plan to extend this example to change the view, it is more efficient to leave it mapped.

Finally, you associate the memory you just allocated with the buffer object:

res = vkBindBufferMemory(info.device, info.uniform_data.buf,
                         info.uniform_data.mem, 0);

And you are done.