The SparseLinSol library for solving large sparse systems of linear algebraic equations is presented in the paper. The key implementation features for multicore CPUs and GPUs are discussed. Performance characteristics are compared for “solve” part of the methods for MPI and hybrid code implementations and GPU-accelerated implementation against the hypre library for a set of matrices of 41–99 mln. unknowns on up to 128 GPU-equipped compute nodes. Preliminary results on coupling the developed library with OpenFOAM package to speedup the hydrodynamic modelling problems are discussed.