A Generalized Reduction Construct for Deterministic OpenMP

In parallel programming, a reduction is an operation that combines values across threads into a single result, and can be designed and implemented so as to enforce determinism, not only on the result, but also on the intermediate values and evaluation sequence. These features make the reduction an attractive feature for a language-based approach to deterministic parallelism, especially one that, like OpenMP, provides annotations to parallelize legacy serial code. Hence reductions are a planned feature of the Deterministic OpenMP (DOMP) project. To enable DOMP to help programmers eliminate nondeterministic code wherever possible, we propose a generalized reduction that supports arbitrary data types and user-defined operations—a generalization rich in challenges for language design and implementation.