The implicit parallelism is an active domain of computer- science to hide intricate details of parallelization from the end-user. Some solutions are specific to a precise domain while others are more generic, however, the purpose is always to find the adapted level of abstraction to ease the high performance and parallel programming. We present Skel- GIS, a header-only implicit parallelism C++ library to solve mesh-based scientific simulations. In this paper is detailed the implementation of SkelGIS for the specific case of network simulations, where the space domain can be represented as a directed acyclic graph (DAG). This im- plementation is based on a modified, optimized and parallelized version of the Compressed Sparse Row format, which is completely described in this paper. Finally, experiments on different kinds of clusters and differ- ent sizes of DAGs are evaluated.