Wave simulation using CUDA



s1170143 Ryo Kawamukai
Computer Graphics laboratory
University of AIZU
Supervised by Prof.Gennadiy Nikishkov
Outline
●   Introduction
      -Background
      -Purpose
●   Methd
      -CUDA technology
      -Wave Simulation
●   Summary
●   Future Work
Introduction -Background
In recent years, the processing speed of current graphics hardware
is equal to the ten years ago super computer.


A Graphics Processing Unit (GPU) is a dedicated graphics
rendering hardware.


The processing speed of the GPU is higher than that of the CPU.


Now, the GPU is used for general purpose computations not only
for graphics rendering.
Introduction -Purpose
The purpose of this research is:


Perform wave simulation using CUDA technology.


Investigate processing speed of the GPU and CPU, and
find the situations when the GPU processing speed
considerably exceeds CPU speed in wave simulation.


If I could, I make computing speed of the simulation
faster.
Method -CUDA technology

  Device Memory f_d       Host Code           Host Memory f_h
                       //Memory Pointer
                       float f_d,f_h;
                         CUDA API
                       cudaMalloc(&f_d);
                       cudaMemcpy(f_h,f_d);




     Device Code      //Kernel function
      __global__      func<<<Dg,Db>>>(f_d);
      func(f_d){ }
Method -Wave Simulation
 Result of wave simulation are rendered by OpenGL as
 a real time 3D visualization.
 This simulation are performed by the GPU and the
 CPU.
 Processing speed is measured to compare
 performance.
Summary
 In this research, wave simulation is performed using CUDA
 technology.
 Processing time for computation and total time of particle
 simulation using the GPU and the CPU is measured.


 Comparison shows that processing speed of the GPU is
 considerably higher than processing speed of the CPU.
 Larger number of particles leads to more efficient
 simulation on the GPU in comparison to the CPU.
Future Work

 To give proving data that it became early.



 If I could, I make computing speed of the
 simulation faster.

S1170143 2

  • 1.
    Wave simulation usingCUDA s1170143 Ryo Kawamukai Computer Graphics laboratory University of AIZU Supervised by Prof.Gennadiy Nikishkov
  • 2.
    Outline ● Introduction -Background -Purpose ● Methd -CUDA technology -Wave Simulation ● Summary ● Future Work
  • 3.
    Introduction -Background In recentyears, the processing speed of current graphics hardware is equal to the ten years ago super computer. A Graphics Processing Unit (GPU) is a dedicated graphics rendering hardware. The processing speed of the GPU is higher than that of the CPU. Now, the GPU is used for general purpose computations not only for graphics rendering.
  • 4.
    Introduction -Purpose The purposeof this research is: Perform wave simulation using CUDA technology. Investigate processing speed of the GPU and CPU, and find the situations when the GPU processing speed considerably exceeds CPU speed in wave simulation. If I could, I make computing speed of the simulation faster.
  • 5.
    Method -CUDA technology Device Memory f_d Host Code Host Memory f_h //Memory Pointer float f_d,f_h; CUDA API cudaMalloc(&f_d); cudaMemcpy(f_h,f_d); Device Code //Kernel function __global__ func<<<Dg,Db>>>(f_d); func(f_d){ }
  • 6.
    Method -Wave Simulation Result of wave simulation are rendered by OpenGL as a real time 3D visualization. This simulation are performed by the GPU and the CPU. Processing speed is measured to compare performance.
  • 7.
    Summary In thisresearch, wave simulation is performed using CUDA technology. Processing time for computation and total time of particle simulation using the GPU and the CPU is measured. Comparison shows that processing speed of the GPU is considerably higher than processing speed of the CPU. Larger number of particles leads to more efficient simulation on the GPU in comparison to the CPU.
  • 8.
    Future Work Togive proving data that it became early. If I could, I make computing speed of the simulation faster.