Installation Guide

Prerequisites

Before installing Trilinos, ensure the following dependencies are available:

  • CMake: Required for configuring the build system.
  • Compilers: Compatible C++ compilers such as GCC, Clang, or Intel.
  • MPI: For parallel computing support (optional but recommended).
  • Libraries: Additional libraries like BLAS, LAPACK, and Boost may be required depending on the configuration.

Download Trilinos

You can download Trilinos from the official repository or the Download page.

Build Instructions

Follow these steps to configure, build, and install Trilinos:

  1. Clone the Trilinos repository:

    git clone https://github.com/trilinos/Trilinos.git
cd Trilinos

  2. Configure with CMake:

    mkdir build
cd build
cmake .. -DCMAKE_INSTALL_PREFIX=/path/to/install
    • Replace /path/to/install with the desired installation directory.
    • Add additional CMake options to enable specific packages or features.

  3. Build and install:

    make -j4
make install
    • Use -jN to specify the number of parallel build threads (N).

For detailed instructions, refer to the Quick configure, build and install hints for Trilinos.

Prebuilt Packages

If you prefer not to build Trilinos from source, you can explore prebuilt binaries or package managers like Spack Trilinos Package.

Troubleshooting

  • Common installation issues and solutions.

  • FAQ or support forum.

  • Common Issues: Check for missing dependencies, incorrect compiler versions, or MPI configuration errors.

  • Support: Submit an issue on the Trilinos GitHub Issues page.

Getting Started with Trilinos

Basic Concepts

  • Overview of Trilinos: Learn about Trilinos' modular architecture and how packages interact to solve complex problems. See Overview of Trilinos.
  • Modular Architecture: Trilinos is composed of individual packages, each designed for specific tasks (e.g., Tpetra for linear algebra, Panzer for finite element assembly). These packages can be used independently or combined to build custom solutions.

Simple Example

A simple example demonstrating basic Trilinos functionality:

#include <iostream>
#include <Tpetra_Core.hpp>
#include <Tpetra_Vector.hpp>

int main(int argc, char* argv[]) {
    Tpetra::ScopeGuard tpetraScope(&argc, &argv);           // Initialize Tpetra (and MPI if enabled)

    auto comm = Tpetra::getDefaultComm();                   // Get the default communicator (MPI or serial)

    const Tpetra::global_size_t globalSize = 10;            // Define the global size of the vector

    Tpetra::Map<> map(globalSize, 0, comm);                 // One-to-one distribution of elements across processes)

    Tpetra::Vector<> vec(map);                              // Create a Tpetra Vector

    vec.putScalar(1.0);                                     // Fill the vector with scalar value 1.0

    vec.sync<Kokkos::HostSpace>();                          // Ensure data is synchronized to the host
    auto localView = vec.getLocalView<Kokkos::HostSpace>(); // Access local data

    // Print local values for each process
    for (size_t i = 0; i < localView.extent(0); ++i) {
        std::cout << "Process " << comm->getRank() << ", vec[" << map.getGlobalElement(i) 
                  << "] = " << localView(i) << std::endl;
    }

    return 0;
}

This example demonstrates:

  • Initializing Trilinos and MPI.
  • Creating and manipulating a distributed vector using Tpetra.