ExaHyPE-DSL A Python-based domain specific language (DSL) for the open source ExaHyPE hyperbolic PDE simulation engine.

ExaHyPE is an open-source simulation engine to write solvers for hyperbolic PDE systems. In the UK, this code is primarily used to study gravitational waves. However, next generation simulations, such as those of Black Holes, require faster and more optimal compute kernels. Whilst ExaHyPE ships with several different numerical kernels, optimised for different use-cases, it is impossible to provide optimal kernel implementations for every situation as these often change over the runtime of the code. This poster will report progress made in the ExaHyPE-DSL eCSE project which avoids the need to manually develop, maintain and tune multiple kernels across CPUs and GPUs by the use of a Domain Specific Language (DSL). DSLs enable the programmer to express their intentions in a manner that is close to the problem in hand. Then, using this rich source of information the compiler is able to make informed decisions around the tricky, low level details of parallelism. In ExaHyPE-DSL we have developed a Python based DSL which, using the ExCALIBUR xDSL toolkit, generates optimal variants of the kernels and autotunes them at runtime. This work means that the scientific community will be able to use ExaHyPE to run hitherto unfeasible simulations and study novel physics; notably alternative models of gravitation.