Jump to content

Buffers in Vulkan

Josh

530 views

I've now got the Vulkan renderer drawing multiple different models in one single pass. This is done by merging all mesh geometry into one single vertex and indice buffer and using indirect drawing. I implemented this originally in OpenGL and was able to translate the technique over to Vulkan. This can allow an entire scene to be drawn in just one or a few draw calls. This will make a tremendous improvement in performance in complex scenes like The Zone. In that scene in Leadwerks the slow step is the rendering routine on the CPU churning through thousands of OpenGL commands, and this design effectively eliminates that entire bottleneck.

Untitled.thumb.png.924b12d05a17e7a611ffdd44c2676cec.png

There is no depth buffer in use in the above image, so some triangles appear on top of others they are behind.

Vulkan provides a lot of control when transferring memory into VRAM, and as a result we saw an 80% performance improvement over OpenGL in our first performance comparison. I have set up a system that uses staging buffers to transfer bits of memory from the CPU into shared memory buffers on the GPU. Another interesting capability is the ability to transfer multiple chunks of data between buffers in just one command.

However, that control comes at a cost of complexity. At the moment, the above code works fine on Intel graphics but crashes on my discrete Nvidia card. This makes sense because of the way Vulkan handles memory. You have to explicitly synchronize memory yourself using a pipeline barrier. Since Intel graphics just uses system memory I don't think it will have any problems with memory synchronization like a discrete card will.

That will be the next step, and it is really a complex topic, but my usage of it will be limited, so I think in the end my own code will turn out to be pretty simple. I expect Vulkan 2.0 will probably introduce a lot of simplified paths that will become the default, because this stuff is really just too hard for both beginners and experts. There’s no reason for memory to not be synced automatically and you’re just playing with fire otherwise.

  • Like 3


3 Comments


Recommended Comments

To my surprise, my first attempt at pipeline barriers seems to have worked perfectly.

Share this comment


Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Add a comment...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

  • Blog Entries

    • By Josh in Josh's Dev Blog 2
      Documentation in Leadwerks 5 will start in the header files, where functions descriptions are being added directly like this:
      /// <summary> /// Sets the height of one terrain point. /// </summary> /// <param name="x">Horizontal position of the point to modify.</param> /// <param name="y">Vertical position of the point to modify.</param> /// <param name="height">Height to set, in the range -1.0 to +1.0.</param> virtual void SetHeight(const int x, const int y, const float height); This will make function descriptions appear automatically in Visual Studio, to help you write code faster and more easily:

      Visual Studio can also generate an XML file containing all of the project's function descriptions as part of the build process. The generated XML file will serve as the basis for the online documentation and Visual Studio Code extension for Lua. This is how I see it working:

      I am also moving all things private to private members. I found a cool trick that allows me to create read-only members. In the example below, you can access the "position" member to get an entity's local position, but you cannot modify it without using the SetPosition() method. This is important because modifying values often involves updating lots of things in the engine under the hood and syncing data with other threads. This also means that any method Visual Studio displays as you are typing is okay to use, and there won't be any undocumented / use-at-your-own risk types of commands like we had in Leadwerks 4.
      class Entity { private: Vec3 m_position; public: const Vec3& position; }; Entity::Entity() : position(m_position) {} It is even possible to make constructors private so that the programmer has to use the correct CreateTerrain() or whatever command, instead of trying to construct a new instance of the class, with unpredictable results. Interestingly, the constructor itself has to be added as a friend function for this to work.
      class Terrein { private: Terrain(); public: friend shared_ptr<World> CreateTerrain(shared_ptr<World>, int, int, int) }; The only difference is that inside the CreateTerrain function I have to do this:
      auto terrain = shared_ptr<Terrain>(new Terrain); instead of this, because make_shared() doesn't have access to the Terrain constructor. (If it did, you would be able to create a shared pointer to a new terrain, so we don't want that!)
      auto terrain = make_shared<Terrain>(); I have big expectations for Leadwerks 5, so it makes sense to pay a lot of attention to the coding experience you will have while using this. I hope you like it!
    • By Josh in Josh's Dev Blog 8
      An often-requested feature for terrain building commands in Leadwerks 5 is being implemented. Here is my script to create a terrain. This creates a 256 x 256 terrain with one terrain point every meter, and a maximum height of +/- 50 meters:
      --Create terrain local terrain = CreateTerrain(world,256,256) terrain:SetScale(256,100,256) Here is what it looks like:

      A single material layer is then added to the terrain.
      --Add a material layer local mtl = LoadMaterial("Materials/Dirt/dirt01.mat") local layerID = terrain:AddLayer(mtl) We don't have to do anything else to make the material appear because by default the entire terrain is set to use the first layer, if a material is available there:

      Next we will raise a few terrain points.
      --Modify terrain height for x=-5,5 do for y=-5,5 do h = (1 - (math.sqrt(x*x + y*y)) / 5) * 20 terrain:SetElevation(127 + x, 127 + y, h) end end And then we will update the normals for that whole section, all at once. Notice that we specify a larger grid for the normals update, because the terrain points next to the ones we modified will have their normals affected by the change in height of the neighboring pixel.
      --Update normals of modified and neighboring points terrain:UpdateNormals(127 - 6, 127 - 6, 13, 13) Now we have a small hill.

      Next let's add another layer and apply it to terrain points that are on the side of the hill we just created:
      --Add another layer mtl = LoadMaterial("Materials/Rough-rockface1.json") rockLayerID = terrain:AddLayer(mtl) --Apply layer to sides of hill for x=-5,5 do for y=-5,5 do slope = terrain:GetSlope(127 + x, 127 + y) alpha = math.min(slope / 15, 1.0) terrain:SetMaterial(rockLayerID, 127 + x, 127 + y, alpha) end end We could improve the appearance by giving it a more gradual change in the rock layer alpha, but it's okay for now.

      This gives you an idea of the basic terrain building API in Leadwerks 5, and it will serve as the foundation for more advanced terrain features. This will be included in the next beta.
    • By Josh in Josh's Dev Blog 0
      A new update is available for beta testers.
      Terrain
      The terrain building API is now available and you can begin working with it, This allows you to construct and modify terrains in pure code. Terrain supports up to 256 materials, each with its own albedo, normal, and displacement maps. Collision and raycasting are currently not supported.
      Fast C++ Builds
      Precompiled headers have been integrated into the example project. The Debug build will compile in about 20 seconds the first run, and compile in just 2-3 seconds thereafter. An example class is included which shows how to add files to your game project for optimum compile times. Even if you edit one of your header files, your game will still compile in just a few seconds in debug mode! Integrating precompiled headers into the engine actually brought the size of the static libraries down significantly, so the download is only about 350 MB now.
      Enums Everywhere
      Integer arguments have been replaced with enum values for window styles, entity bounds, and load flags. This is nice because the C++ compiler has some error checking so you don't do something like this:
      LoadTexture("grass.dds", WINDOW_FULLSCREEN); Operators have been added to allow combining enum values as bitwise flags.
      A new LOAD_DUMP_INFO LoadFlags value has been added which will print out information about loaded files (I need this to debug the GLTF loader!).
      Early Spring Cleaning
      Almost all the pre-processor macros have been removed from the Visual Studio project, with just a couple ones left. Overall the headers and project structure have been massively cleaned up.
×
×
  • Create New...