Parallel C++

Building HPX applications

Fri, 06 Dec 2024 14:34:25 +0000

In this blog post, we will provide more details on how to build your own HPX application using CMake. In the previous blog post we showcased how on compile and install HPX.

Building a HPX application

We assume that we have a CMakeLists.txt and the main.cpp in the same folder. Below the main CMakeLists.txt is listed

project(MyHPXApp LANGUAGES CXX)
cmake_minimum_required(VERSION 3.18)

# Find the HPX package
find_package(HPX REQUIRED)

# Compile a HPX application
add_hpx_executable(app SOURCES main.cpp)

Compiling a HPX application

To compile the HPX applicaiton, we generate a new folder build, and change the director, and call cmake. Note that $HPX_INSTALL_DIR is the path where you installed HPX.

$ mkdir build
$ cd build
$ cmake -DHPX_DIR=$HPX_INSTALL_DIR/hpx/lib64/cmake/HPX ..
$ make

Addendum

Thu, 15 Aug 2024 14:34:25 +0000

It does not matter how often one proofreads the book, there will be always bugs in the code or some issues. We will document the issues here and update the code on GitHub. Of course, we will update the potential new release of the book accordingly.

Chapter 3

Listing 3.3 the cases for even and odd are wrong in the code and should be chnaged to

if (values.size() % 2 == 0) {
  median = *mid; |\label{container:median:list:defref}|
}  else {
  auto mid_one = values.begin();
  std::advance(mid_one, mid_index + 1);
  median = 0.5 * (*mid + *mid_one);
}

Listing 3.6 the C++20 range-based for statements with initializer

for (int index = 0; const int value : values) {
std::cout << "Index=" << index++ << " Value= " << value
}

Chapter 15

Listing 15.7 on page 175 of the printed version the lines 38 to 40 should be removed.

38 std::vector<hpx::future<void>> futures;
39 
40 hpx::when_all(futures).get();

Appendix

Listing A.10 is correct in source file, however, the pipes used for the ranges are not correctly rendered in the book.

#include <cmath>
#include <iostream>
#include <ranges>
#include <vector>

int main() {
  std ::vector<int> values = {0, 1, 2, 3, 4, 5, 6};

  auto tr_values = values
    |  std::views::filter([](int value) { return value % 2 == 0; }) 
    |  std::views::transform([](int value) -> double { return std::sqrt(value);});
  for (int i : tr_values)
      std::cout << double(i) << ' ';
  std::cout << std::endl;
}

Installing HPX

Tue, 13 Aug 2024 14:34:25 +0000

In this blog post, we will provide more details on how to install HPX on your system.

Fedora Linux

On Fedora we mantain the HPX package and the package can be installed as follows:

# Shared memory version
sudo dnf install hpx-devel
# Distributed memory version
sudo dnf install hpx-mpich-devel

Mac OS

On Mac OS, we recommend to install the depencies with brew and compile HPX using CMake as follows:

brew install hwloc cmake boost
git clone https://github.com/STEllAR-GROUP/hpx.git
cd hpx
mkdir build
cd build
cmake -DHPX_WITH_FETCH_ASIO=ON ..
make -j 8

Spack

On supercomputers and clusters, the HPX package mananger Spack can be used to install HPX.

# Install spack (Update v0.22 with the latest Spack release)
git clone --depth=100 --branch=releases/v0.22 https://github.com/spack/spack.git ~/spack
# Activate Spack
. ~/spack/share/spack/setup-env.sh
# Find compilers
spack compiler find
# Install hpx 
spack install hpx
# Install distributed hpx
spack external find openmpi
spack install hpx networking=mpi ^openmpi