/cuda-reduce-max-with-index

Primary LanguageCMIT LicenseMIT

Introduction

This sample was created to demonstrate and measure speed of searching for max element and its index in CUDA with different algorithms comparing to C++ STL max_element function.

Build

  • Download and install Visual Studio 2017
  • Download and install CUDA
    • Download CUDA Toolkit 10.0
    • Launch the downloaded installer package
    • Read and accept the EULA
    • Select "Next" to download and install all components
    • Once the download completes, the installation will begin automatically
    • Follow the steps of installation wizard
    • Finish installation
  • Download and install CMake
  • Get the sources
    • git clone https://github.com/apriorit/cuda-reduce-max-with-index.git into c:\cuda-reduce-max-with-index (destination dir can be anything, c:\cuda-reduce-max-with-index is used here just for the reference)
  • Generate Visual Studio project files
    • Launch CMake (cmake-gui) from the start menu
    • Set "Where is the source code" to c:\cuda-reduce-max-with-index
    • Set "Where to build the binaries" to c:\cuda-reduce-max-with-index\bin
    • Click "Generate"
    • Choose "Visual Studio 15 2017 Win64"
    • Click "Finish"
    • Click "Open Project" or launch (may not be available for CUDA projects) c:\cuda-reduce-max-with-index\bin\cuda-reduce-max-with-index.sln
  • Build
    • Set Solution Configurations
    • Choose from the main menu "Build->Build solution"
    • Navigate to c:\cuda-reduce-max-with-index\bin\<configuration output folder> and get ReduceMaxWithIndex.exe
  • Build with msbuild.exe
    • msbuild.exe bin/ReduceMaxWithIndex.exe.sln /t:Build /p:Configuration=Release