Convolution Using CUDA

A project to compare time taken by CPU and GPU to carry out convolution operation

CUDA Installation

• Installer: CUDA Toolkit 10.1 Update 2 from https://developer.nvidia.com/cuda-downloads?target_os=Windows&target_arch=x86_64&target_version=7&target_type=exelocal
  • Install CUDA and follow on screen prompt
  • Once installed:
    • To verify correct configuration of hardware and software, build and run the deviceQuery sample program.
      • Solution path: C:\ProgramData\NVIDIA Corporation\CUDA Samples\v10.1\1_Utilities\deviceQuery\deviceQuery_vs2015.sln
    • To ensure that the system and the CUDA-capable device are able to communicate correctly, build and run the bandwidthTest program.
      • Solution path: C:\ProgramData\NVIDIA Corporation\CUDA Samples\v10.1\1_Utilities\bandwidthTest\bandwidthTest_vs2015.sln
    • To see a graphical representation of what CUDA can do, run the sample Particles program
      • Solution path: C:\ProgramData\NVIDIA Corporation\CUDA Samples\v10.1\5_Simulations\particles\particles_vs2015.sln
    • Output exe will be in path: C:\ProgramData\NVIDIA Corporation\CUDA Samples\v10.1\bin\win64\Release

Project Setup in Visual Studio 2015 Professional

• File->New->Project
• Select Templates->NVIDIA->CUDA 10.1->CUDA 10.1 Runtime
• Give project a name. E.g. 2D_Convolution_Using_Shared_Memory
• Go to "Properties" of the project:
  • Set "Output Directory" and "Intermediate Directory" under "General" tab as:
    • Output Directory : $(SolutionDir)output\$(Configuration)\
    • Intermediate Directory : $(SolutionDir)build\$(ProjectName)_$(Configuration)\
  • Set "Code Generation" under "CUDA C/C++->Device" tab as:
    • Code Generation : compute_30,sm_30;compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;