Tutorials

Long-form, narrative walkthroughs designed to teach Ginkgo by building something concrete. Each tutorial is a single executable with copy-paste code, a CMake recipe, and expected output you can verify against your own run.

Available tutorials

• Poisson on CPU — Build a CG + block-Jacobi solver for a 1D Poisson problem on the CPU. Walks through CSR assembly, the RHS with boundary conditions, the solver factory, and a mesh-refinement study that exhibits the textbook \(\mathcal{O}(h^2)\) convergence rate.

• Poisson on GPU — The same problem, ported to the GPU with a three-line diff. Covers the host/device executor pattern, cross-executor data movement (gko::clone), and when GPU pays off vs. when CPU is faster.

• Heat equation on a distributed grid — Port the upstream 2D heat-conduction example to Ginkgo’s distributed types. Implicit-Euler time stepping, partition-by-stripes, Schwarz preconditioner, and a per-rank-stripes output pattern that a short Python script stitches into an animated GIF.