Jump to content
  • Advertisement
Sign in to follow this  
Ubik

OpenGL Cases for multithreading OpenGL code?

This topic is 1495 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

I have wanted to support multiple contexts to be used in separate threads with shared resources in this convenience GL wrapper I've been fiddling with. My goal hasn't been to expose everything it can do but in a nice way, but the multi-context support has seemed like a good thing, to align the wrapper with a pretty big aspect of the underlying system. However, multithreaded stuff is hard, so I finally started to question if supporting multiple contexts with resource sharing is even worth it.

 

If the intended use is PC gaming - a single simple window and so on (a single person project too, to put things to scale, and currently targeting version 3.3 if that has any relevance), what reasons would there be to take the harder route? My understanding is that the benefits might actually be pretty limited, but my knowledge of all the various implementations and their capabilities is definitely limited.

Share this post


Link to post
Share on other sites
Advertisement

Alright, I honestly hadn't expected this to be so clear-cut. A bit sad and maybe a little ironic too that GPU rendering can't be parallelized from client side too, with OpenGL anyway.

 

Thank you for sharing the knowledge!

Share this post


Link to post
Share on other sites
Well, it's only a partly parallelisable problem as the GPU is reading from a single command buffer (well, in the GL/D3D model, the hardware doesn't work quite the same as Mantle shows giving you 3 command queues per device but still...) so at some point your commands have to get into that stream (be it by physically adding to a chunk of memory or inserting a jump instruction to a block to execute) so you are always going to a single thread/sync point going on.

However, command sub-buffer construction is a highly parallelisable thing, consoles have been doing it for ages, the problem is the OpenGL mindset seems to be 'this isn't a problem - just multi-draw all the things!' and the D3D11 "solution" was a flawed one because of how the driver works internally.

D3D12 and Mantle should shake this up and hopefully show that parallel command buffer construction is a good thing and that OpenGL needs to get with the program (or, as someone at Valve said, it'll get chewed up by the newer APIs).

Share this post


Link to post
Share on other sites

Doing texture streaming in parallel works great on all GPUs I've tested on, which include a shitload of Nvidia GPUs, at least an AMD HD7790 and a few Intel GPUs. It's essentially stutter-free.

Share this post


Link to post
Share on other sites

Here's a silly but related question. If I use a second context to upload data that takes, for example, a second to transfer over the PCI bus, will the main rendering thread stall while it waits for it's own per-frame data, or will the driver split the larger data into chunks, thereby allowing the two threads to interlace their data?

Share this post


Link to post
Share on other sites

Well, it's only a partly parallelisable problem as the GPU is reading from a single command buffer (well, in the GL/D3D model, the hardware doesn't work quite the same as Mantle shows giving you 3 command queues per device but still...) so at some point your commands have to get into that stream (be it by physically adding to a chunk of memory or inserting a jump instruction to a block to execute) so you are always going to a single thread/sync point going on.

However, command sub-buffer construction is a highly parallelisable thing, consoles have been doing it for ages, the problem is the OpenGL mindset seems to be 'this isn't a problem - just multi-draw all the things!' and the D3D11 "solution" was a flawed one because of how the driver works internally.

D3D12 and Mantle should shake this up and hopefully show that parallel command buffer construction is a good thing and that OpenGL needs to get with the program (or, as someone at Valve said, it'll get chewed up by the newer APIs).

Good clarification. Makes sense that even if the GPU has lots of pixel/vertex/computing units, the system controlling them isn't necessarily as parallel-friendly. For a non-hw person the number three sounds like a curious choice, but in any case it seems to make some intuitive sense to have the number close to a common number of CPU cores. That's excluding hyper-threading but that's an Intel thing so doesn't matter to folks at AMD. (Though there's the consoles with more cores...)

 

I'm wishing for something nicer than OpenGL to happen too, but it's probably going to take some time for things to actually change. Not on Windows here, so the wait is likely going to be longer still. Might as well use GL in the mean time.

 

Doing texture streaming in parallel works great on all GPUs I've tested on, which include a shitload of Nvidia GPUs, at least an AMD HD7790 and a few Intel GPUs. It's essentially stutter-free.

Creating resources or uploading data on a second context is what I've mostly had in mind earlier. I did try to find info on this, but probably didn't use the right terms because I got the impression that actually parallel data transfer isn't that commonly supported.

 

I've now thought that if I'm going to add secondary context support anyway, it will be in a very constrained way, so that the other context (or wrapper for it to be specific) won't be a general purpose one but targeting things like resource loading specifically. That could allow me to keep the complexity at bay.

Share this post


Link to post
Share on other sites
As we've pretty much got the answer to the original question I'm going to take a moment to quickly (and basically) cover a thing smile.png

Good clarification. Makes sense that even if the GPU has lots of pixel/vertex/computing units, the system controlling them isn't necessarily as parallel-friendly. For a non-hw person the number three sounds like a curious choice, but in any case it seems to make some intuitive sense to have the number close to a common number of CPU cores. That's excluding hyper-threading but that's an Intel thing so doesn't matter to folks at AMD. (Though there's the consoles with more cores...)


So, the number '3' has nothing to do with CPU core counts; when it comes to GPU/CPU reasoning very little of one directly impacts the other.

A GPU works by consuming 'command packets'; the OpenGL calls you make get translated by the driver into bytes the GPU can natively read and understand, in the same way a compiler transforms your code to binary for the CPU.

The OpenGL and D3D11 model of a GPU presents a case where the command stream is handled by a single 'command processor' which is the hardware which decodes the command packets to make the GPU do it's work. For a long time this was probably the case too so the conceptual model 'works'.

However, a recent GPU, such as AMD's Graphics Core Next series is a bit more complicated than that as the interface which deals with the commands isn't a single block but in fact 3 which can each consume a stream of commands.

First is the 'graphics command processor'; this can dispatch graphics and compute workloads to the GPU hardware to work - glDraw/glDispatch family of functions - and is where your commands end up.

Secondly there is the 'compute command processors' - these can handle compute only workloads. Not exposed via GL, I think OpenCL can kind of expose them but with Mantle it is a separate command queue. (The driver might make use of them as well behind the scenes)

Finally 'dma commands' which is a separate command queue to move data to/from the GPU which is handled in OpenGL behind the scenes by the driver (but in Mantle would allow you to kick your own uploads/downloads as required.

So the command queues as exposed by Mantle more closely mirror the operation of the hardware (it still hides some details) which explains why you have three, to cover the 3 types of command work the GPU can do.

If you are interested AMD have made a lot of this detail available which is pretty cool.
(Annoyingly NV are very conservative about their hardware details which makes me sad sad.png)

To be clear, you don't need to know this stuff although I personally find it interesting - this is also a pretty high level overview of the situation so don't take it as a "this is how GPUs work!" kinda thing smile.png

Share this post


Link to post
Share on other sites
Sign in to follow this  

  • Advertisement
  • Advertisement
  • Popular Tags

  • Similar Content

    • By nOoNEE
      hello guys , i have some questions  what does glLinkProgram  and  glBindAttribLocation do?  i searched but there wasnt any good resource 
    • By owenjr
      Hi, I'm a Multimedia Engineering student. I am about to finish my dergree and I'm already thinking about what topic to cover in my final college project.
      I'm interested in the procedural animation with c++ and OpenGL of creatures, something like a spider for example. Can someone tell me what are the issues I should investigate to carry it out? I understand that it has some dependence on artificial intelligence but I do not know to what extent. Can someone help me to find information about it? Thank you very much.
       
      Examples: 
      - Procedural multi-legged walking animation
      - Procedural Locomotion of Multi-Legged Characters in Dynamic Environments
    • By Lewa
      So, i'm still on my quest to unterstanding the intricacies of HDR and implementing this into my engine. Currently i'm at the step to implementing tonemapping. I stumbled upon this blogposts:
      http://filmicworlds.com/blog/filmic-tonemapping-operators/
      http://frictionalgames.blogspot.com/2012/09/tech-feature-hdr-lightning.html
      and tried to implement some of those mentioned tonemapping methods into my postprocessing shader.
      The issue is that none of them creates the same results as shown in the blogpost which definitely has to do with the initial range in which the values are stored in the HDR buffer. For simplicity sake i store the values between 0 and 1 in the HDR buffer (ambient light is 0.3, directional light is 0.7)
      This is the tonemapping code:
      vec3 Uncharted2Tonemap(vec3 x) { float A = 0.15; float B = 0.50; float C = 0.10; float D = 0.20; float E = 0.02; float F = 0.30; return ((x*(A*x+C*B)+D*E)/(x*(A*x+B)+D*F))-E/F; } This is without the uncharted tonemapping:
      This is with the uncharted tonemapping:
      Which makes the image a lot darker.
      The shader code looks like this:
      void main() { vec3 color = texture2D(texture_diffuse, vTexcoord).rgb; color = Uncharted2Tonemap(color); //gamma correction (use only if not done in tonemapping code) color = gammaCorrection(color); outputF = vec4(color,1.0f); } Now, from my understanding is that tonemapping should bring the range down from HDR to 0-1.
      But the output of the tonemapping function heavily depends on the initial range of the values in the HDR buffer. (You can't expect to set the sun intensity the first time to 10 and the second time to 1000 and excpect the same result if you feed that into the tonemapper.) So i suppose that this also depends on the exposure which i have to implement?
      To check this i plotted the tonemapping curve:
      You can see that the curve goes only up to around to a value of 0.21 (while being fed a value of 1) and then basically flattens out. (which would explain why the image got darker.)
       
      My guestion is: In what range should the values in the HDR buffer be which then get tonemapped? Do i have to bring them down to a range of 0-1 by multiplying with the exposure?
      For example, if i increase the values of the light by 10 (directional light would be 7 and ambient light 3) then i would need to divide HDR values by 10 in order to get a value range of 0-1 which then could be fed into the tonemapping curve. Is that correct?
    • By nOoNEE
      i am reading this book : link
      in the OpenGL Rendering Pipeline section there is a picture like this: link
      but the question is this i dont really understand why it is necessary to turn pixel data in to fragment and then fragment into pixel could please give me a source or a clear Explanation that why it is necessary ? thank you so mu
       
       
    • By Inbar_xz
      I'm using the OPENGL with eclipse+JOGL.
      My goal is to create movement of the camera and the player.
      I create main class, which create some box in 3D and hold 
      an object of PlayerAxis.
      I create PlayerAxis class which hold the axis of the player.
      If we want to move the camera, then in the main class I call to 
      the func "cameraMove"(from PlayerAxis) and it update the player axis.
      That's work good.
      The problem start if I move the camera on 2 axis, 
      for example if I move with the camera right(that's on the y axis)
      and then down(on the x axis) -
      in some point the move front is not to the front anymore..
      In order to move to the front, I do
      player.playerMoving(0, 0, 1);
      And I learn that in order to keep the front move, 
      I need to convert (0, 0, 1) to the player axis, and then add this.
      I think I dont do the convert right.. 
      I will be glad for help!

      Here is part of my PlayerAxis class:
       
      //player coordinate float x[] = new float[3]; float y[] = new float[3]; float z[] = new float[3]; public PlayerAxis(float move_step, float angle_move) { x[0] = 1; y[1] = 1; z[2] = -1; step = move_step; angle = angle_move; setTransMatrix(); } public void cameraMoving(float angle_step, String axis) { float[] new_x = x; float[] new_y = y; float[] new_z = z; float alfa = angle_step * angle; switch(axis) { case "x": new_z = addVectors(multScalar(z, COS(alfa)), multScalar(y, SIN(alfa))); new_y = subVectors(multScalar(y, COS(alfa)), multScalar(z, SIN(alfa))); break; case "y": new_x = addVectors(multScalar(x, COS(alfa)), multScalar(z, SIN(alfa))); new_z = subVectors(multScalar(z, COS(alfa)), multScalar(x, SIN(alfa))); break; case "z": new_x = addVectors(multScalar(x, COS(alfa)), multScalar(y, SIN(alfa))); new_y = subVectors(multScalar(y, COS(alfa)), multScalar(x, SIN(alfa))); } x = new_x; y = new_y; z = new_z; normalization(); } public void playerMoving(float x_move, float y_move, float z_move) { float[] move = new float[3]; move[0] = x_move; move[1] = y_move; move[2] = z_move; setTransMatrix(); float[] trans_move = transVector(move); position[0] = position[0] + step*trans_move[0]; position[1] = position[1] + step*trans_move[1]; position[2] = position[2] + step*trans_move[2]; } public void setTransMatrix() { for (int i = 0; i < 3; i++) { coordiTrans[0][i] = x[i]; coordiTrans[1][i] = y[i]; coordiTrans[2][i] = z[i]; } } public float[] transVector(float[] v) { return multiplyMatrixInVector(coordiTrans, v); }  
      and in the main class i have this:
       
      public void keyPressed(KeyEvent e) { if (e.getKeyCode()== KeyEvent.VK_ESCAPE) { System.exit(0); //player move } else if (e.getKeyCode()== KeyEvent.VK_W) { //front //moveAmount[2] += -0.1f; player.playerMoving(0, 0, 1); } else if (e.getKeyCode()== KeyEvent.VK_S) { //back //moveAmount[2] += 0.1f; player.playerMoving(0, 0, -1); } else if (e.getKeyCode()== KeyEvent.VK_A) { //left //moveAmount[0] += -0.1f; player.playerMoving(-1, 0, 0); } else if (e.getKeyCode()== KeyEvent.VK_D) { //right //moveAmount[0] += 0.1f; player.playerMoving(1, 0, 0); } else if (e.getKeyCode()== KeyEvent.VK_E) { //moveAmount[0] += 0.1f; player.playerMoving(0, 1, 0); } else if (e.getKeyCode()== KeyEvent.VK_Q) { //moveAmount[0] += 0.1f; player.playerMoving(0, -1, 0); //camera move } else if (e.getKeyCode()== KeyEvent.VK_I) { //up player.cameraMoving(1, "x"); } else if (e.getKeyCode()== KeyEvent.VK_K) { //down player.cameraMoving(-1, "x"); } else if (e.getKeyCode()== KeyEvent.VK_L) { //right player.cameraMoving(-1, "y"); } else if (e.getKeyCode()== KeyEvent.VK_J) { //left player.cameraMoving(1, "y"); } else if (e.getKeyCode()== KeyEvent.VK_O) { //right round player.cameraMoving(-1, "z"); } else if (e.getKeyCode()== KeyEvent.VK_U) { //left round player.cameraMoving(1, "z"); } }  
      finallt found it.... i confused with the transformation matrix row and col. thanks anyway!
    • By Lewa
      So, i'm currently trying to implement an SSAO shader from THIS tutorial and i'm running into a few issues here.
      Now, this SSAO method requires view space positions and normals. I'm storing the normals in my deferred renderer in world-space so i had to do a conversion and reconstruct the position from the depth buffer.
      And something there goes horribly wrong (which has probably to do with worldspace to viewspace transformations).
      (here is the full shader source code if someone wants to take a look at it)
      Now, i suspect that the normals are the culprit.
      vec3 normal = ((uNormalViewMatrix*vec4(normalize(texture2D(sNormals, vTexcoord).rgb),1.0)).xyz); "sNormals" is a 2D texture which stores the normals in world space in a RGB FP16 buffer.
      Now i can't use the camera viewspace matrix to transform the normals into viewspace as the cameras position isn't set at (0,0,0), thus skewing the result.
      So what i did is to create a new viewmatrix specifically for this normal without the position at vec3(0,0,0);
      //"camera" is the camera which was used for rendering the normal buffer renderer.setUniform4m(ressources->shaderSSAO->getUniform("uNormalViewMatrix"), glmExt::createViewMatrix(glm::vec3(0,0,0),camera.getForward(),camera.getUp())//parameters are (position,forwardVector,upVector) ); Though i have the feeling this is the wrong approach. Is this right or is there a better/correct way of transforming a world space normal into viewspace?
    • By HawkDeath
      Hi,
      I'm trying mix two textures using own shader system, but I have a problem (I think) with uniforms.
      Code: https://github.com/HawkDeath/shader/tree/test
      To debug I use RenderDocs, but I did not receive good results. In the first attachment is my result, in the second attachment is what should be.
      PS. I base on this tutorial https://learnopengl.com/Getting-started/Textures.


    • By norman784
      I'm having issues loading textures, as I'm clueless on how to handle / load images maybe I missing something, but the past few days I just google a lot to try to find a solution. Well theres two issues I think, one I'm using Kotlin Native (EAP) and OpenGL wrapper / STB image, so I'm not quite sure wheres the issue, if someone with more experience could give me some hints on how to solve this issue?
      The code is here, if I'm not mistaken the workflow is pretty straight forward, stbi_load returns the pixels of the image (as char array or byte array) and you need to pass those pixels directly to glTexImage2D, so a I'm missing something here it seems.
      Regards
    • By Hashbrown
      I've noticed in most post processing tutorials several shaders are used one after another: one for bloom, another for contrast, and so on. For example: 
      postprocessing.quad.bind() // Effect 1 effect1.shader.bind(); postprocessing.texture.bind(); postprocessing.quad.draw(); postprocessing.texture.unbind(); effect1.shader.unbind(); // Effect 2 effect2.shader.bind(); // ...and so on postprocessing.quad.unbind() Is this good practice, how many shaders can I bind and unbind before I hit performance issues? I'm afraid I don't know what the good practices are in open/webGL regarding binding and unbinding resources. 
      I'm guessing binding many shaders at post processing is okay since the scene has already been updated and I'm just working on a quad and texture at that moment. Or is it more optimal to put shader code in chunks and bind less frequently? I'd love to use several shaders at post though. 
      Another example of what I'm doing at the moment:
      1) Loop through GameObjects, bind its phong shader (send color, shadow, spec, normal samplers), unbind all.
      2) At post: bind post processor quad, and loop/bind through different shader effects, and so on ...
      Thanks all! 
    • By phil67rpg
      void collision(int v) { collision_bug_one(0.0f, 10.0f); glutPostRedisplay(); glutTimerFunc(1000, collision, 0); } void coll_sprite() { if (board[0][0] == 1) { collision(0); flag[0][0] = 1; } } void erase_sprite() { if (flag[0][0] == 1) { glColor3f(0.0f, 0.0f, 0.0f); glBegin(GL_POLYGON); glVertex3f(0.0f, 10.0f, 0.0f); glVertex3f(0.0f, 9.0f, 0.0f); glVertex3f(1.0f, 9.0f, 0.0f); glVertex3f(1.0f, 10.0f, 0.0f); glEnd(); } } I am using glutTimerFunc to wait a small amount of time to display a collision sprite before I black out the sprite. unfortunately my code only blacks out the said sprite without drawing the collision sprite, I have done a great deal of research on the glutTimerFunc and  animation.
  • Advertisement
  • Popular Now

  • Forum Statistics

    • Total Topics
      631392
    • Total Posts
      2999738
×

Important Information

By using GameDev.net, you agree to our community Guidelines, Terms of Use, and Privacy Policy.

Participate in the game development conversation and more when you create an account on GameDev.net!

Sign me up!