Schedule Syllabus Resources

Particle simulation

The Particles code in your repository is simple example of General Purpose GPU computing. We will use the GPU to solve a problem that is not graphics: particle simulation. This is done storing the particle data on the GPU and using shaders to read and update the data. We are going to do this using textures, framebuffers, and shaders. In general, OpenGL supports these kind of computations with a "Transform Feedback" pipeline. Since we are using OpenGL, there are still limitations related to its focus on graphics. The parallel compute language OpenCL can be used when more flexibility is needed.

We can store particle position and velocity in textures. We can the textures as input to the shaders. The shaders can then read the current particle position and velocity. The shaders will then need to compute the new positions and velocities and write the output. The new values are output to the framebuffer. So, the framebuffer must be setup to have two color buffers, one for position and one for velocity. Given these requirements, setup:

• Source framebuffer with two color attachments
  • Source position texture
  • Source velocity texture
• Destination framebuffer with two color attachments
  • Destination position texture
  • Destination velocity texture

A shader reads the input textures, then outputs the new values to the output buffers. Then, the source and destination buffers are switched for the next update cycle.

Code

• particleUpdate.frag

  //TODO update p and v
  p = p + v*t + 0.5*a*t*t;
  v = v + a*t;
  
  if(increaseVelocity == 1u)
  	v = v*(1+t);
  

• Model.h : You can adjust particle by changing PARTICLE_SIZE_X and PARTICLE_SIZE_Y