Ginkgo Integrations#

Ginkgo is a high-performance numerical linear algebra library for many-core systems, with a focus on the solution of sparse linear systems. It is implemented using modern C++ with GPU kernels for NVIDIA (CUDA), AMD (HIP), and Intel (SYCL) GPUs. The library core contains architecture-agnostic algorithms – iterative solvers, preconditioners, and matrix formats – while runtime polymorphism selects the right architecture-optimized kernel depending on the target hardware.

Ginkgo architecture

This site collects self-contained integration examples that demonstrate how Ginkgo can be used as a solver backend inside established scientific computing frameworks. Each example ships with its own build system, source code, and documentation so you can study the approach, build it locally, and adapt it to your own project.


Browse by library#

deal.II

Finite-element library for adaptive meshes.

Poisson Equation with deal.II + Ginkgo
MFEM

Scalable finite-element library with high-order support.

Poisson Equation with MFEM + Ginkgo
NekRS

GPU-accelerated spectral-element CFD solver.

Pressure Poisson with NekRS + Ginkgo
SUNDIALS

Suite of nonlinear and differential/algebraic equation solvers.

Steady-State Poisson with SUNDIALS KINSOL + Ginkgo

Compare by problem#

Poisson equation#

The Poisson equation appears in nearly every scientific domain. Compare how each framework formulates the same problem and uses Ginkgo to solve the resulting linear system:

Framework

Example

Approach

deal.II

Poisson

Continuous Galerkin FEM on adaptive meshes

MFEM

Poisson

High-performance FEM with flexible backends

NekRS

Poisson

Spectral-element discretization on GPUs

SUNDIALS

Poisson

Nonlinear solve (KINSOL) with Ginkgo preconditioner