Cosmic dust is a fundamental build block of stars and planets. It is also a primary source of emissions observed in astronomy, which enables us to map the universe. Dust comprises clumps of atoms and molecules in sizes ranging from sub-nanometer scales to a few hundred nanometers, created as part of the ejecta of supernovae and also formed in the turbulent structures of the ISM. Dust is also destroyed in the blast waves of supernovae and heating by stellar UV radiation. The processes of formation and destruction are complicated and the correlation of dust to gas in the ISM varies considerably due to the environment and depending on the dust size, making modelling the dust, modelling the gas and interpreting the observations of them highly challenging. Recent simulations of dust destruction in the turbulent ISM have revealed that the conservation of dust is improved by the turbulence and strong inhomogeneity in the density distribution, but, surprisingly, also due to the Lorentz forces acting on the dust from a turbulent magnetic field embedded within the ISM. We have, so far modelled these processes by applying 2D postprocessing of the dust using Paperboats onto 3D MHD simulations of supernova driven turbulence provided by the Pencil Code, due to the numerical challenges of modelling the dust processing and MHD in 3D simultaneously. With this project we aim to couple the simulations of the dust processing and the simulations of the MHD by coupling two implementations of the Pencil Code with itself to extract subdomains of the MHD to execute simultaneously the dust processing with accelerated timestepping at smaller scales. The use of GPU acceleration of the Pencil Code has also been developed at Aalto University enabling the simulations to be controlled and managed by the CPU based Pencil Code, while the heavy duty RHS integrations are handled by coupling with Astaroth. The scientific goals are a mjor focus of this proposal, but also is the continued development of the GPU acceleration to include dust modelling and the extension of coupling code techniques to improve code integration and versatility.