Sign in to follow this  
Followers 0
Lil_Lloyd

OpenGL
Naivete Challenged: Interleaving arrays causes slow performance?

7 posts in this topic

Ok, so I was thinking about how to speed up my opengl application, and remembered hearing something about interleaving array data. To my simple mind it seemed that when I pass attributes to a shader in separate buffer objects some kind of jumping around in the video cards' memory would occur, so if I interleaved the data for every vertex then this should provide the benefit of locality of reference like in a cache, and things should speed up.

 

However after some experimenting it cause a decline of 2 frames per second in my overall speed. So WHY does this happen? I'm clearly drinking from the wrong bottle of GL goodness. 

0

Share this post


Link to post
Share on other sites

2 frames per second

2 frames of what ? FPS is not the best measurement, 2 fps of 4 fps is much more than 2 fps of 10000 fps.

 

From what you describes I would say, that vertex perfomance is not your bottleneck and a drop of 2 fps of X is not a proof of interlaeving arrays are slower than not interleaved ones.

Edited by Ashaman73
0

Share this post


Link to post
Share on other sites

so if I interleaved the data for every vertex then this should provide the benefit of locality of reference like in a cache, and things should speed up.

Did you make sure that each vertex is a power-of-2 size? If your vertex is an odd size it will cross multiple cache lines - 16, 32 or 64 bytes are typically good sizes.

0

Share this post


Link to post
Share on other sites

As a general rule interleaving should be faster, but there is one case in which it may be slower, and that's if the per-vertex pipeline is either implemented in or being emulated by software.  In that case splitting off the position component of your arrays, and padding it to 4 floats to better allow for SIMD optimizations in your driver, may be the optimal solution.

 

You haven't really given much info about the kind of workload you're running or what kind of hardware you've got, but some circumstances that can cause per-vertex to run in software would include having older graphics hardware that doesn't support hardware T&L (some Intel integrateds from about 5/6 years or more ago would fall under this category), or exceeding a hardware limit/trying to do something that's allowed by the GL spec but is not actually supported in hardware (that may include using GL_UNSIGNED_INT for your GL_ELEMENT_ARRAY_BUFFER on older hardware - again, you have to go back at least 6 years for this - or using GL_UNSIGNED_BYTE on any hardware).  If any of this sounds familiar you may have just identified your cause.

0

Share this post


Link to post
Share on other sites

so if I interleaved the data for every vertex then this should provide the benefit of locality of reference like in a cache, and things should speed up.

Did you make sure that each vertex is a power-of-2 size? If your vertex is an odd size it will cross multiple cache lines - 16, 32 or 64 bytes are typically good sizes.

 

Aha! I had very odd vertex sizes, 12 bytes for a position, 12 bytes for a normal, 8 bytes for a texture or  just 12 bytes for a position, 8 bytes for a texture. So I will pad the vertices, textures and normals out to 16 bytes each but make each stride between a whole bunch of vertex data 64 bytes as 48 bytes isn't a power of two. Is this a good strategy?

 

Are there any useful links or articles concerning this kind of thing and the strategies you should use?

 

 

 

As a general rule interleaving should be faster, but there is one case in which it may be slower, and that's if the per-vertex pipeline is either implemented in or being emulated by software.  In that case splitting off the position component of your arrays, and padding it to 4 floats to better allow for SIMD optimizations in your driver, may be the optimal solution.

 

You haven't really given much info about the kind of workload you're running or what kind of hardware you've got, but some circumstances that can cause per-vertex to run in software would include having older graphics hardware that doesn't support hardware T&L (some Intel integrateds from about 5/6 years or more ago would fall under this category), or exceeding a hardware limit/trying to do something that's allowed by the GL spec but is not actually supported in hardware (that may include using GL_UNSIGNED_INT for your GL_ELEMENT_ARRAY_BUFFER on older hardware - again, you have to go back at least 6 years for this - or using GL_UNSIGNED_BYTE on any hardware).  If any of this sounds familiar you may have just identified your cause.

 

I'm using an AMD HD Radeon 4200 card. Is this a terrible Graphics card?

0

Share this post


Link to post
Share on other sites

Did you make sure that each vertex is a power-of-2 size? If your vertex is an odd size it will cross multiple cache lines - 16, 32 or 64 bytes are typically good sizes.

 
Aha! I had very odd vertex sizes, 12 bytes for a position, 12 bytes for a normal, 8 bytes for a texture or  just 12 bytes for a position, 8 bytes for a texture. So I will pad the vertices, textures and normals out to 16 bytes each but make each stride between a whole bunch of vertex data 64 bytes as 48 bytes isn't a power of two. Is this a good strategy?

Actually it isn’t about it being a power of 2, but a multiple of the cache size, which is usually 16 or 32. MSDN quotes 32 explicitly for certain cases of vertex buffers but my testing reveals this to improve the performance in all cases, not just what they list.  Of course, their documentation is for Direct3D, but cache issues and friends are a universal issue.

Performance Optimizations

 

Also, we are not talking about padding each vertex-buffer element.  12-byte normals and positions (etc.) are completely normal.

This padding is between each vertex in the buffer, not each element of each vertex.

Normally once you interleave position, normals, UV’s, tangents, and bitangents, you have 56-byte vertices.  This should be padded out to 64.

 

 

L. Spiro

Edited by L. Spiro
2

Share this post


Link to post
Share on other sites

Actually it isn’t about it being a power of 2, but a multiple of the cache size, which is usually 16 or 32.

Sure. And a power of two will always be an even multiple/divisor of the cache size.

It's accurate enough for a rule of thumb *shrugs*

Aha! I had very odd vertex sizes, 12 bytes for a position, 12 bytes for a normal, 8 bytes for a texture

12 + 12 + 8 = 32 bytes, which is just the right size for cache efficiency.

If you weren't interleaving these you'd have to be more careful about the individual sizes, but for an interleaved vertex it's the structure as a whole that matters. Edited by swiftcoder
0

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  
Followers 0

  • Similar Content

    • By recp
      Hi,
      I'm working on new asset importer (https://github.com/recp/assetkit) based on COLLADA specs, the question is not about COLLADA directly
      also I'm working on a new renderer to render (https://github.com/recp/libgk) imported document.
      In the future I'll spend more time on this renderer of course, currently rendering imported (implemented parts) is enough for me
      assetkit imports COLLADA document (it will support glTF too),
      importing scene, geometries, effects/materials, 2d textures and rendering them seems working
      My actual confusion is about shaders. COLLADA has COMMON profile and GLSL... profiles,
      GLSL profile provides shaders for effects so I don't need to wory about them just compile, link, group them before render

      The problem occours in COMMON profile because I need to write shaders,
      Actually I wrote them for basic matrials and another version for 2d texture
      I would like to create multiple program but I am not sure how to split this this shader into smaller ones,

      Basic material version (only colors):
      https://github.com/recp/libgk/blob/master/src/default/shader/gk_default.frag
      Texture version:
      https://gist.github.com/recp/b0368c74c35d9d6912f524624bfbf5a3
      I used subroutines to bind materials, actually I liked it,
      In scene graph every node can have different program, and it switches between them if parentNode->program != node->program
      (I'll do scene graph optimizations e.g.  view frustum culling, grouping shaders... later)

      I'm going to implement transparency but I'm considering to create separate shaders,
      because default shader is going to be branching hell
      I can't generate shader for every node because I don't know how many node can be exist, there is no limit.
      I don't know how to write a good uber-shader for different cases:

      Here material struct:
      struct Material { ColorOrTexture emission; ColorOrTexture ambient; ColorOrTexture specular; ColorOrTexture reflective; ColorOrTexture transparent; ColorOrTexture diffuse; float shininess; float reflectivEyety; float transparency; float indexOfRefraction; }; ColorOrTexture could be color or 2d texture, if there would be single colorOrTex then I could split into two programs,
      Also I'm going to implement transparency, I am not sure how many program that I needed

      I'm considering to maintain a few default shaders for COMMON profile,
      1-no-texture, 2-one of colorOrTexture contains texture, 3-........

      Any advices in general or about how to optimize/split (if I need) these shaders which I provied as link?
      What do you think the shaders I wrote, I would like to write them without branching if posible,
      I hope I don't need to write 50+ or 100+ shaders, and 100+ default programs

      PS: These default shaders should render any document, they are not specific, they are general purpose...
             I'm compiling and linking default shaders when app launched

      Thanks
    • By CircleOfLight97
      Hi guys,
      I would like to contribute to a game project as a developer (open source possibly). I have some experiences in C/C++ in game development (perso projects). I don't know either unreal or unity but I have some knowledges in opengl, glsl and shading theory as I had some courses at university regarding to that. I have some knowledges in maths and basic in physics. I know a little how to use blender to do modelling, texturing and simple game assets (no characters, no animation no skinning/rigging). I have no game preferences but I like aventure game, dungeon crawler, platformers, randomly generated things. I know these kind of projects involve a lot of time and I'd be really to work on it but if there are no cleary defined specific design goals/stories/gameplay mechanics I would like to not be part of it x) and I would rather prefer a smaller but well defined project to work on that a huge and not 'finishable' one.
      CircleOfLight97
    • By gamesthatcouldbeworse
      Hi, I finally released KILL COMMANDO on gamejolt for free. It is a retro-funsplatter-shooter with C64 style. Give it a try.
    • By phil67rpg

      void TimerFunction(int value) {  glutPostRedisplay();  glutTimerFunc(1000, TimerFunction, 1); } void drawScene() {  glClear(GL_COLOR_BUFFER_BIT);      drawScene_bug();  TimerFunction(1);  eraseScene_bug(); // drawScene_bug_two(); // eraseScene_bug_two(); drawScene_ship(); drawScene_bullet();  glutSwapBuffers(); }
  • Popular Now