Sign in to follow this  
Relfos

OpenGL How to optimize scene rendering

Recommended Posts

I was building an map editor for our RPG game, but as the features and complexity of the maps raise, some more optimizations are needed. As you can see in the screenshot, the polycount of the scene is rather low for today standarts. Also, basic boundingbox-frustum test are used to cull invisible objects out of the rendering. The "Chunks" is the number of mesh chunks passed to the renderer, the "Batches" the number of actual glDrawElements calls for 3d geometry. The "Spr.Batches" is the number of glDrawElements calls for 2d geometry. About 200 sprites are passed in, but they're sorted into batches and only 6 calls are done, so I think this dont need more optimizations for now. Ignore the "Mirrors" and "Lights" for now, as they're not used in this scene. The "Shaders" is the number of shader switches per frame, and the "Transfer" the ammount of texture memory swapped around per second. About the current optimizations. The renderer sort objects by alpha, then by texture, then by distance. (Shaders are not taken into the sort function, should I include them?) All static meshes are rendered using VBOs, however the animated ones are updated in the CPU (both skeletal, morphs and keyframe animation). A cache is used per instance, and only modified when needed (eg: The updates are only done at 24fps). I need to store the modified geometry myself since then its used for generating both mesh outlines and shadow volumes. About the shadow volume itself, its done in the CPU, and even if I disable it in this scene, only gain about 5 or 6 more fps. All textures and shaders are cached, so as long as sort the objects correctly, minimal changes are done. Also, every shader uniform is cached, to minimize changes. The sky itself is a dynamic skydome, however its also rendered fast in one GL call. The whole scene is rendered using a modified parallax mapping algoritm that uses 3 texture units but its very simple and lightweight. Finally, theres a post process shader that is rendered as a fullscreen quad, however its quite fast, disabling it doesnt change the fps it seems. So I need tips to optimize this, please. How to sort objects for minimizing gl status changes? Its more expensive swapping a shader or a texture? Its there any openGL tools to check where the rendering bottlenecks are? The terrain geometry is completely unoptimized, even a completely flat surface will have lots of polys generated. Should I try to implement some form of geo-mipmapping or even LOD for the trees and stuff?

Share this post


Link to post
Share on other sites
Have you profiled to see where the bottleneck is? I love the debug info you have in your HUD, but that isn't really enough to go on. In particular, are you GPU bound or CPU bound (I would assume the former)?

First off, I am pretty sure that switching shaders ought to be more expensive than switching textures - most engines sort by material rather than texture.

Secondly, by my estimation you are only rendering about 30 triangles per batch, which is very low for modern cards. Can you increase this at all?

Share this post


Link to post
Share on other sites
No, I dont have much experience profiling code, and I'm using Delphi so I really dont have any idea how to do it.
But probably CPU bound, as the scene geometry is far from complex for my gfx card.
Already changed the sorting to include shaders and the fps increased a little.

"Secondly, by my estimation you are only rendering about 30 triangles per batch, which is very low for modern cards. Can you increase this at all?"

Hmmm, as you can, most of the meshes in the scene are very low poly trees.
How can I batch more than one mesh per call, since every mesh has it own transformation matrix?

Share this post


Link to post
Share on other sites
Quote:
Original post by Relfos
How can I batch more than one mesh per call, since every mesh has it own transformation matrix?
Either by using the instancing extension (only available on the newest nVidia cards) or by pseudo-instancing. Interestingly, pseudo-instancing seems to be no slower than "real" instancing (at least not measurable).

For pseudo-instancing, you pass your transform matrix (or something else) as color or texture coordinate using the glTexCoord or glColor functions (or glVertexAttrib ). Note that you do not have to add something extra to your vertex buffer data, but you use these functions to set the state once before rendering each object. It is the same as you would do in immediate mode (although it is not immediate mode!). This is very little overhead and effectively lets you draw hundreds of trees in one batch.

In your vertex shader, you figure out what to make of that data. It is pretty much up to you what you want to do. For example, if you want to be able to scale your trees and place them at different locations, but you don't care about orientation, that will fit neatly into one attribute. Construct a matrix from it, and multiply.

Share this post


Link to post
Share on other sites
Quote:
Original post by Relfos

About the shadow volume itself, its done in the CPU, and even if I disable it in this scene, only gain about 5 or 6 more fps.



Only gain 5 or more fps? That's a lot of gain for this particular scene.

12 FPS = 83 ms.
12 + 5 = 17 FPS = 58 ms.
Speed gain = (83 - 58) / 83 = 30%

If the framerate with shadows was 30 FPS, it would probably be around 44 FPS with shadows turned off. Please stop using FPS to measure framerate and use milliseconds per frame instead. Speed gains are more obvious that way.

Share this post


Link to post
Share on other sites
Yeah, it's really hard to say where the slowdown is without a little profiling. I believe the newer high end versions of Visual C++ come with a profiler, although I moved to Linux before that was the case so I don't know too much about it. I don't know any free Windows profilers off the top of my head, but since you're doing OpenGL you could always install some flavor of Linux and use Valgrind, whish is about as powerful as they come ;) Seriously though...

My first observation though, is that 14k polygons is a LOT for a scene of the complexity I'm seeing in that screenshot. Is that how many are being passed to the renderer, or are a lot of those being culled out? I wasn't too clear on how you described that.

One thing I would try is simply turning off ALL of the special effects in the scene, then turning them on again one by one to see where the biggest performance hits come in. That is, first start by rendering it as a completely static scene with no shadows, no fog, and only a simple passthrough shader that doesn't do anything fancy. Then activate all those features individually, and test the speed that way (i.e. test with ONLY shadows, ONLY fog, ONLY animation, ONLY advanced shaders). With those results you should be in a better position to try out specific optimizations or come back with more detailed information. Also, I don't know what video card you have, but my wild guess is the parallax mapping you're doing would be very slow on anything that isn't fairly recent.

The best suggestion I can think of off the top of my head would be to implement some kind of space partitioning structure (assuming you aren't using one) for the entire scene. A basic quadtree for example would be a simple one to implement, and would probably give a nice boost to both the culling and depth sorting.

Also, an instancing technique, as mentioned earlier, might help a little, although with that particular scene it doesn't look like its effect would be that great (just on the trees really, and there aren't all that many of them).

I don't know if you're doing any kind of dynamic level-of-detail work on your terrain, but that might be something to look into to reduce the polygon count being passed to the renderer. I would definitely save this for last though, because dynamic LoD on terrain is pretty complex just by itself.

And finally, is that compiled in debug or release mode?

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this  

  • Forum Statistics

    • Total Topics
      627778
    • Total Posts
      2979025
  • Similar Content

    • By lonewolff
      Hi guys,
      With OpenGL not having a dedicated SDK, how were libraries like GLUT and the likes ever written?
      Could someone these days write an OpenGL library from scratch? How would you even go about this?
      Obviously this question stems from the fact that there is no OpenGL SDK.
      DirectX is a bit different as MS has the advantage of having the relationship with the vendors and having full access to OS source code and the entire works.
      If I were to attempt to write the most absolute basic lib to access OpenGL on the GPU, how would I go about this?
    • By DelicateTreeFrog
      Hello! As an exercise for delving into modern OpenGL, I'm creating a simple .obj renderer. I want to support things like varying degrees of specularity, geometry opacity, things like that, on a per-material basis. Different materials can also have different textures. Basic .obj necessities. I've done this in old school OpenGL, but modern OpenGL has its own thing going on, and I'd like to conform as closely to the standards as possible so as to keep the program running correctly, and I'm hoping to avoid picking up bad habits this early on.
      Reading around on the OpenGL Wiki, one tip in particular really stands out to me on this page:
      For something like a renderer for .obj files, this sort of thing seems almost ideal, but according to the wiki, it's a bad idea. Interesting to note!
      So, here's what the plan is so far as far as loading goes:
      Set up a type for materials so that materials can be created and destroyed. They will contain things like diffuse color, diffuse texture, geometry opacity, and so on, for each material in the .mtl file. Since .obj files are conveniently split up by material, I can load different groups of vertices/normals/UVs and triangles into different blocks of data for different models. When it comes to the rendering, I get a bit lost. I can either:
      Between drawing triangle groups, call glUseProgram to use a different shader for that particular geometry (so a unique shader just for the material that is shared by this triangle group). or
      Between drawing triangle groups, call glUniform a few times to adjust different parameters within the "master shader", such as specularity, diffuse color, and geometry opacity. In both cases, I still have to call glBindTexture between drawing triangle groups in order to bind the diffuse texture used by the material, so there doesn't seem to be a way around having the CPU do *something* during the rendering process instead of letting the GPU do everything all at once.
      The second option here seems less cluttered, however. There are less shaders to keep up with while one "master shader" handles it all. I don't have to duplicate any code or compile multiple shaders. Arguably, I could always have the shader program for each material be embedded in the material itself, and be auto-generated upon loading the material from the .mtl file. But this still leads to constantly calling glUseProgram, much more than is probably necessary in order to properly render the .obj. There seem to be a number of differing opinions on if it's okay to use hundreds of shaders or if it's best to just use tens of shaders.
      So, ultimately, what is the "right" way to do this? Does using a "master shader" (or a few variants of one) bog down the system compared to using hundreds of shader programs each dedicated to their own corresponding materials? Keeping in mind that the "master shaders" would have to track these additional uniforms and potentially have numerous branches of ifs, it may be possible that the ifs will lead to additional and unnecessary processing. But would that more expensive than constantly calling glUseProgram to switch shaders, or storing the shaders to begin with?
      With all these angles to consider, it's difficult to come to a conclusion. Both possible methods work, and both seem rather convenient for their own reasons, but which is the most performant? Please help this beginner/dummy understand. Thank you!
    • By JJCDeveloper
      I want to make professional java 3d game with server program and database,packet handling for multiplayer and client-server communicating,maps rendering,models,and stuffs Which aspect of java can I learn and where can I learn java Lwjgl OpenGL rendering Like minecraft and world of tanks
    • By AyeRonTarpas
      A friend of mine and I are making a 2D game engine as a learning experience and to hopefully build upon the experience in the long run.

      -What I'm using:
          C++;. Since im learning this language while in college and its one of the popular language to make games with why not.     Visual Studios; Im using a windows so yea.     SDL or GLFW; was thinking about SDL since i do some research on it where it is catching my interest but i hear SDL is a huge package compared to GLFW, so i may do GLFW to start with as learning since i may get overwhelmed with SDL.  
      -Questions
      Knowing what we want in the engine what should our main focus be in terms of learning. File managements, with headers, functions ect. How can i properly manage files with out confusing myself and my friend when sharing code. Alternative to Visual studios: My friend has a mac and cant properly use Vis studios, is there another alternative to it?  
    • By ferreiradaselva
      Both functions are available since 3.0, and I'm currently using `glMapBuffer()`, which works fine.
      But, I was wondering if anyone has experienced advantage in using `glMapBufferRange()`, which allows to specify the range of the mapped buffer. Could this be only a safety measure or does it improve performance?
      Note: I'm not asking about glBufferSubData()/glBufferData. Those two are irrelevant in this case.
  • Popular Now