The REAPER Stack: Compiling a Chemical Mechanism at Run Time

A chemistry model repeats the same reaction bookkeeping in many cells and at many times. The concentrations change; the identities of the reactants usually do not. REAPER uses that fixed structure to write a small set of specialized programs after the model has loaded its mechanism. A system compiler turns them into a library that the solver attaches and later runs reuse. We explain the construction, and six appendices explain the computer science from the ground up, after Feynman’s lectures on computation. TurboMICM provides working examples of the generated kernels, and has measured such kernels at five to six times the generic path inside one solver. The complete run-time loading and caching path remains a design; its own startup cost has not been measured.