Minimal CUDA Solver

The minimal CUDA solver example.

Kind: basic
Builds on: simple-solver
Upstream source: examples/minimal-cuda-solver/minimal-cuda-solver.cpp in the Ginkgo repository.

Introduction

This is a minimal example that solves a system with Ginkgo. The matrix, right hand side and initial guess are read from standard input, and the result is written to standard output. The system matrix is stored in CSR format, and the system solved using the CG method, preconditioned with the block-Jacobi preconditioner. All computations are done on the GPU.

The easiest way to use the example data from the data/ folder is to concatenate the matrix, the right hand side and the initial solution (in that exact order), and pipe the result to the minimal_solver_cuda executable:

textcat data/A.mtx data/b.mtx data/x0.mtx | ./minimal-cuda-solver

The commented program

#include <iostream>

#include <ginkgo/ginkgo.hpp>

int main()
{

Instantiate a CUDA executor

    auto gpu = gko::CudaExecutor::create(0, gko::OmpExecutor::create());

Read data

    auto A = gko::read<gko::matrix::Csr<>>(std::cin, gpu);
    auto b = gko::read<gko::matrix::Dense<>>(std::cin, gpu);
    auto x = gko::read<gko::matrix::Dense<>>(std::cin, gpu);

Create the solver

    auto solver =
        gko::solver::Cg<>::build()
            .with_preconditioner(gko::preconditioner::Jacobi<>::build())
            .with_criteria(
                gko::stop::Iteration::build().with_max_iters(20u),
                gko::stop::ResidualNorm<>::build().with_reduction_factor(1e-15))
            .on(gpu);

Solve system

    solver->generate(give(A))->apply(b, x);

Write result

    write(std::cout, x);
}

Results

The following is the expected result when using the data contained in the folder data as input:

%%MatrixMarket matrix array real general
19 1
0.252218
0.108645
0.0662811
0.0630433
0.0384088
0.0396536
0.0402648
0.0338935
0.0193098
0.0234653
0.0211499
0.0196413
0.0199151
0.0181674
0.0162722
0.0150714
0.0107016
0.0121141
0.0123025

The plain program