AMD ROCm Performance Primitives (RPP) library is a comprehensive, high-performance computer vision library for AMD processors that have HIP, OpenCL, or CPU backends.

The latest RPP release is:

Input (nifti1 .nii medical image)	fused_multiply_add_scalar (brightened 3D image)

You can build our documentation locally using the following code:

• Sphinx

  cd docs
  pip3 install -r .sphinx/requirements.txt
  python3 -m sphinx -T -E -b html -d _build/doctrees -D language=en . _build/html

• Doxygen

  doxygen .Doxyfile

Refer to our Linux GPU and OS support page to see if your system is supported.

To use RPP, you must have installed the following:

• ROCm For ROCm installation instructions, see Linux quick-start.

• Clang Version 5.0.1 and above

  • Ubuntu 20/22

    sudo apt-get install clang

  • CentOS 7

    sudo yum install llvm-toolset-7-clang llvm-toolset-7-clang-analyzer llvm-toolset-7-clang-tools-extra
    scl enable llvm-toolset-7 bash

  • RHEL 8/9

    sudo yum install clang

  • SLES 15-SP4 (use ROCm LLVM Clang)

    zypper -n --no-gpg-checks install clang
    update-alternatives --install /usr/bin/clang clang /opt/rocm-*/llvm/bin/clang 100
    update-alternatives --install /usr/bin/clang++ clang++ /opt/rocm-*/llvm/bin/clang++ 100

• CMake Version 3.5 and above

• IEEE 754-based half-precision floating-point library (half.hpp)

  • Use the half package with ROCm

    sudo apt-get install half

  Note that you must use the correct package management utility (zypper/yum).

  • Install from source

  wget https://sourceforge.net/projects/half/files/half/1.12.0/half-1.12.0.zip
  unzip half-1.12.0.zip -d half-files
  sudo mkdir /usr/local/include/half
  sudo cp half-files/include/half.hpp /usr/local/include/half

• Compiler with support for C++ Version 17 and above

• OpenMP

• Threads

• OpenCV 3.4.0/4.5.5

  • Install OpenCV prerequisites:

    sudo apt-get update
    sudo -S apt-get -y --allow-unauthenticated install build-essential libgtk2.0-dev libavcodec-dev libavformat-dev libswscale-dev python-dev python-numpy
    sudo -S apt-get -y --allow-unauthenticated install libtbb2 libtbb-dev libjpeg-dev libpng-dev libtiff-dev libdc1394-22-dev unzip wget

  • Download OpenCV 3.4.0 /4.5.5:

    wget https://github.com/opencv/opencv/archive/3.4.0.zip
    unzip 3.4.0.zip
    cd opencv-3.4.0/

    OR

    wget https://github.com/opencv/opencv/archive/4.5.5.zip
    unzip 4.5.5.zip
    cd opencv-4.5.5/

  • Install OpenCV:

    mkdir build
    cd build
    cmake -D WITH_GTK=ON -D WITH_JPEG=ON -D BUILD_JPEG=ON -D WITH_OPENCL=OFF -D WITH_OPENCLAMDFFT=OFF -D WITH_OPENCLAMDBLAS=OFF -D WITH_VA_INTEL=OFF -D WITH_OPENCL_SVM=OFF -D CMAKE_INSTALL_PREFIX=/usr/local ..
    sudo -S make -j128 <Or other number of threads to use>
    sudo -S make install
    sudo -S ldconfig

• Install TurboJpeg:

  sudo apt-get install nasm
  sudo apt-get install wget
  git clone -b 2.0.6.1 https://github.com/rrawther/libjpeg-turbo.git
  cd libjpeg-turbo
  mkdir build
  cd build
  cmake -DCMAKE_INSTALL_PREFIX=/usr \
        -DCMAKE_BUILD_TYPE=RELEASE  \
        -DENABLE_STATIC=FALSE       \
        -DCMAKE_INSTALL_DOCDIR=/usr/share/doc/libjpeg-turbo-2.0.3 \
        -DCMAKE_INSTALL_DEFAULT_LIBDIR=lib  \
        ..
  make -j$nproc
  sudo make install

• Libsndfile installation

  sudo apt-get update
  sudo apt-get install libsndfile1-dev
  

• Imagemagick

  sudo apt-get install imagemagick
  

• Nifti-Imaging nifti_clib

  git clone [email protected]:NIFTI-Imaging/nifti_clib.git
  cd nifti_clib
  mkdir build
  cd build
  cmake ..
  sudo make -j$nproc install
  

The ROCm Performance Primitives (RPP) library has support for three backends: HIP, OpenCL, and CPU:

• Building RPP with the HIP backend (default):

$ git clone https://github.com/GPUOpen-ProfessionalCompute-Libraries/rpp.git
$ mkdir build && cd build
$ cmake -DBACKEND=HIP ../rpp
$ make -j8
$ sudo make install

• Building RPP with the OPENCL backend:

$ git clone https://github.com/GPUOpen-ProfessionalCompute-Libraries/rpp.git
$ mkdir build && cd build
$ cmake -DBACKEND=OCL ../rpp
$ make -j8
$ sudo make install

• Building RPP with the CPU backend:

$ git clone https://github.com/GPUOpen-ProfessionalCompute-Libraries/rpp.git
$ mkdir build && cd build
$ cmake -DBACKEND=CPU ../rpp
$ make -j8
$ sudo make install

To build and install RPP, run the code shown for your backend:

• HIP (default)

  git clone https://github.com/GPUOpen-ProfessionalCompute-Libraries/rpp.git
  mkdir build && cd build
  cmake -DBACKEND=HIP ../rpp
  make -j8
  sudo make install

• OpenCL

  git clone https://github.com/GPUOpen-ProfessionalCompute-Libraries/rpp.git
  mkdir build && cd build
  cmake -DBACKEND=OCL ../rpp
  make -j8
  sudo make install

• CPU

  git clone https://github.com/GPUOpen-ProfessionalCompute-Libraries/rpp.git
  mkdir build && cd build
  cmake -DBACKEND=CPU ../rpp
  make -j8
  sudo make install

To test the functionalities of RPP, run the code shown for your backend:

• HIP

    cd rpp/utilities/rpp-unittests/HIP_NEW
    ./testAllScript.sh

• OpenCL

    cd rpp/utilities/rpp-unittests/OCL_NEW
    ./testAllScript.sh

  • CPU

    cd rpp/utilities/rpp-unittests/HOST_NEW
    ./testAllScript.sh

MIVisionX RPP extension vx_rpp supports RPP functionality through the OpenVX Framework.

For RPP questions and feedback, you can contact us at [email protected].

To submit feature requests and bug reports, use our GitHub issues page.

All notable changes for each release are added to our changelog.

• Linux distribution
  • Ubuntu - 20.04 / 22.04
  • CentOS - 7
  • RedHat - 8 / 9
  • SLES - 15-SP4
• ROCm: rocm-core - 5.7.0.50700-63
• OpenCV - 4.6.0