Vulkan Pipelines, Barrier, memory management

Hi!
Once again, I am going to present you some vulkan features, like pipelines, barriers, memory management, and all things useful for prior ones. This article will be long, but it will be separating into several chapters.

Memory Management

In Vulkan application, it is up to the developer to manage himself the memory. The number of allocations is limited. Make one allocation for one buffer, or one image is really a bad design in Vulkan. One good design is to make a big allocation (let’s call that a chunk), and manage it yourself, and allocate buffer or image within the chunk.

A Chunk Allocator

We need a simple object which has responsibility for allocations of chunks. It just has to select the good heap and call allocate and free from Vulkan API.

This piece of code is quite simple and easy to read.

Memory Pool

Memory pools are structures used to optimize dynamic allocation performances. In video games, it is not an option to use a memory pool. Ideas are the same I told in the first part. Allocate a chunk, and sub allocate yourself within the chunk. I made a simple generic memory pool.
There is a little scheme which explains what I wanted to do.

Memory Pool
Memory Pool

As you can see, video memory is separated into several parts (4 here) and each “Block” in the linked list describes one sub-allocation.
One block is described by :

  1. Size of the block
  2. Offset of the block relatively with the DeviceMemory
  3. A pointer to set data from the host (map)
  4. Boolean to know about the freeness of the block

A sub-allocation within a chunk is performed as follows :

  1. Traverse the linked list until we find a well-sized free block
  2. Modify the size and set the boolean to false
  3. Create a new block, set size, offset and put boolean to true and insert it after the current one.

A free is quite simple, you just have to put the boolean to true.
A good other method could be a “shrink to fit”. If there are some following others with the boolean set to true, we merge all blocks into one.

Buffers

Buffers are a well-known part in OpenGL. In Vulkan, it is approximately the same, but you have to manage yourself the memory through one memory pool.

When you create one buffer, you have to give him a size, an usage (uniform buffer, index buffer, vertex buffer, …). You also could ask for a sparse buffer (Sparse resources will be a subject of an article one day ^_^). You also could tell him to be in a mode concurrent. Thanks to that, you could access the same buffer through two different queues.

I chose to have a host visible and host coherent memory. But it is not especially useful. Indeed, to achieve a better performance, you could want to use a non coherent memory (but you will have to flush/invalidate your memory!!).
For the host visible memory, it is not especially useful as well, indeed, for indirect rendering, it could be smart to perform culling with the GPU to fill all structures!

Shaders

Shaders are Different parts of your pipelines. It is an approximation obviously. But, for each part (vertex processing, geometry processing, fragment processing…), shader associated is invoked. In Vulkan, shaders are wrote with SPIR-V.
SPIR-V is “.class” are for Java. You may compile your GLSL sources to SPIR-V using glslangvalidator.

Why is SPIR-V so powerful ?

SPIR-V allows developers to provide their application without the shader’s source.
SPIR-V is an intermediate representation. Thanks to that, vendor implementation does not have to write a specific language compiler. It results in a lower complexity for the driver and it could more optimize, and compile it faster.

Shaders in Vulkan

Contrary to OpenGL’s shader, it is really easy to compile in Vulkan.
My implementation keeps in memory all shaders into a hashtable. It lets to prevent any shader’s recompilation.

Pipelines

Pipelines are objects used for dispatch (compute pipelines) or render something (graphic pipelines).

The beginning of this part is going to be a summarize of the Vulkan’s specs.

Descriptors

Shaders access buffer and image resources through special variables. These variables are organized into a set of bindings. One set is described by one descriptor.

Descriptor Set Layout

They describe one set. One set is compound with an array of bindings. Each bindings are described by :

  1. A binding number
  2. One type : Image, uniform buffer, SSBO, …
  3. The number of values (Could be an array of textures)
  4. Stage where shader could access the binding.

Allocation of Descriptor Sets

They are allocated from descriptor pool objects.
One descriptor pool object is described by a number of set allocation possible, and an array of descriptor type / count it can allocate.

Once you have the descriptor pool, you could allocate from it sets (using both descriptor pool and descriptor set layout).
When you destroy the pool, sets also are destroyed.

Give buffer / image to sets

Now, we have descriptors, but we have to tell Vulkan where shaders can get data from.

Pipeline Layouts

Pipeline layouts are a kind of bridge between the pipeline and descriptor sets. They let you manage push constant as well (we’ll see them in a future article).

Implementation

Since descriptor sets are not coupled with pipelines layout. We could separate pipeline layout and descriptor pool / sets, but currently, I prefer to keep them coupled. It is a choice, and it will maybe change in the future.

The idea is quite easy. You create all your descriptor set layouts, then you allocate them through a pool.

Graphics Pipelines in a nutshell

Graphics Pipelines describe exactly what will happened on the rendering part.
They describe

  1. Shader stages
  2. Which kind of data you want to deal with (Position, normal,…)
  3. Which kind of primitive you want to draw (triangle, lines, points)
  4. Which operator you want to use for Stencil and Depth
  5. Multi sampling, color blending,…

The creation of a Graphic Pipeline is really easy, the main difficulty is the configuration.

I used a kind of builder design pattern to configure pipelines.

For the example, I configure my pipeline as follows :

  1. 2 stages : vertex shader and fragment shader
  2. Position 4D (x, y, z, w)
  3. No depth / stencil test
  4. An uniform buffer for one color

This code is a bit long, but it gives all the steps you have to follow to create simple pipelines.

Pipelines and descriptor sets give you an unmatched flexibility.

The main.cpp is this one

And now, we have our perfect triangle !!!!

Triangle using pipelines, shaders
Triangle using pipelines

Barrier and explanations for the main

I am going to explain quickly what memory barriers are.
The idea behind the memory barrier is ensured writes are performed.
When you performed one compute or one render, it is your duty to ensure that data will be visible when you want to re-use them.

In our main.cpp example, I draw a triangle into a frame buffer and present it.

The first barrier is :

Image barriers are compound with access, layout, and pipeline barrier with stage.
Since the presentation is a read of a framebuffer, srcAccessMask is VK_ACCESS_MEMORY_READ_BIT.
Now, we want to render inside this image via a framebuffer, so dstAccessMask is VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT.

We were presented the image, and now we want to render inside it, so, layouts are obvious.
When we submit image memory barrier to the command buffer, we have to tell it which stages are affected. Here, we wait for all commands and we begin for the first stage of the pipeline.

The second image memory barrier is

The only difference is the order and stageMasks. Here we wait for the color attachement (and not the Fragment one !!!!) and we begin with the end of the stages (It is not really easy to explain… but it does not sound not logic).

Steps to render something using pipelines are:

  1. Create pipelines
  2. Create command pools, command buffer and begin them
  3. Create vertex / index buffers
  4. Bind pipelines to their subpass, bind buffers and descriptor sets
  5. VkCmdDraw

References

Specification

It was a long article, I hope it was not unclear and that I didn’t do to much mistakes ^^.

Kiss !!!!

Lava erupting from Vulkan : Initialization or Hello World

Hi there !
A Few weeks ago, February 16th to be precise, Vulkan, the new graphic API from Khronos was released. It is a new API which gives much more control about the GPUs than OpenGL (API I loved before Vulkan ^_^).

OpenGL’s problems

Driver Overhead

Fast rendering problems could be from the driver, video games don’t use perfectly the GPU (maybe 80% instead of 95-100% of use). Driver overheads have big costs and more recent OpenGL version tend to solve this problem with Bindless Textures, multi draws, direct state access, etc.
Keep in mind that each GPU calls could have a big cost.
Cass Everitt, Tim Foley, John McDonald, Graham Sellers presented Approaching Zero Driver Overhead with OpenGL in 2014.

Multi threading

With OpenGL, it is not possible to have an efficient multi threading, because an OpenGL context is for one and only one thread that is why it is not so easy to make a draw call from another thread ^_^.

Vulkan

Vulkan is not really a low level API, but it provides a far better abstraction for moderns hardwares. Vulkan is more than AZDO, it is, as Graham Sellers said, PDCTZO (Pretty Darn Close To Zero Overhead).

Series of articles about Lava

What is Lava ?

Lava is the name I gave to my new graphic (physics?) engine. It will let me learn how Vulkan work, play with it, implement some global illumination algorithms, and probably share with you my learnings and feelings about Vulkan. It is possible that I’ll make some mistakes, so, If I do, please let me know !

Why Lava ?

Vulkan makes me think about Volcano that make me think about Lava, so… I chose it 😀 .

Initialization

Now begins what I wanted to discuss, initialization of Vulkan.
First of all, you have to really know and understand what you will attend to do. For the beginning, we are going to see how to have a simple pink window.

Hello world with Vulkan
Hello world with Vulkan

When you are developing with Vulkan, I advise you to have specifications from Khronos on another window (or screen if you are using multiple screens).
To have an easier way to manage windows, I am using GLFW 3.2, and yes, you are mandatory to compile it yourself ^_^, but it is not difficult at all, so it is not a big deal.

Instance

Contrary to OpenGL, in Vulkan, there is no global state, an instance could be similar to an OpenGL Context. An instance doesn’t know anything about other instances, is utterly isolate. The creation of an instance is really easy.

Physical devices, devices and queues

From this Instance, you could retrieve all GPUs on your computer.
You could create a connection between your application and the GPU you want using a VkDevice.
Creating this connection, you have to create as well queues.
Queues are used to perform tasks, you submit the task to a queue and it will be performed.
The queues are separated between several families.
A good way could be use several queues, for example, one for the physics and one for the graphics (or even 2 or three for this last).
You could as well give a priority (between 0 and 1) to a queue. Thanks to that, if you consider a task not so important, you just have to give to the used queue a low priority :).

Image, ImageViews and FrameBuffers

The images represent a mono or multi dimensional array (1D, 2D or 3D).
The images don’t give any get or set for data. If you want to use them in your application, then you must use ImageViews.

ImageViews are directly relied to an image. The creation of an ImageView is not really complicated.

You could write into ImageViews via FrameBuffers. A FrameBuffer owns multiple imageViews (attachments) and is used to write into them.

The way to render something

A window is assigned to a Surface (VkSurfaceKHR). To draw something, you have to render into this surface via swapchains.

From notions of Swapchains

In Vulkan, you have to manage the double buffering by yourself via Swapchain. When you create a swapchain, you link it to a Surface and tell it how many images you need. For a double buffering, you need 2 images.

Once the swapchain was created, you should retrieve images and create frame buffers using them.

The steps to have a correct swapchain is :

  1. Create a Window
  2. Create a Surface assigned to this Window
  3. Create a Swapchain with several images assigned to this Surface
  4. Create FrameBuffers using all of these images.

Using swapchain is not difficult.

  1. Acquire the new image index
  2. Present queue

To notions of Render Pass

Right now, Vulkan should be initialized. To render something, we have to use render pass, and command buffer.

Command Buffers

Command buffer is quite similar to vertex array object (VAO) or display list (old old old OpenGL 😀 ).
You begin the recorded state, you record some “information” and you end the recorded state.
Command buffers are allocated from the CommandPool.

Vulkan provides two types of Command Buffer.

  1. Primary level : They should be submitted within a queue.
  2. Secondary level : They should be executed by a primary level command buffer.

Renderpass

One render pass is executed on one framebuffer. The creation is not easy at all. One render pass is componed with one or several subpasses.
I remind that framebuffers could have several attachments.
Each attachment are not mandatory to be used for all subpasses.

This piece of code to create one renderpass is not definitive at all and will be changed as soon as possible ^^. But for our example, it is correct.

In the same way as for command buffer, render pass should be began and ended!

Our engine in action

Actually, our “engine” is not really usable ^^.
But in the future, command pool, render pass should don’t appear in the user files !

If you want the whole source code :
GitHub

Reference

Approaching Zero Driver Overhead :Lecture
Approaching Zero Driver Overhead : Slides
Vulkan Overview 2015
Vulkan in 30 minutes
VkCube
GLFW with Vulkan

Blog talking about 3D rendering, Qt and C++