SyncChecker: Detecting Synchronization Errors between MPI Applications and Libraries

Abstract

We have implemented a prototype of SyncChecker on Linux and evaluated it with seven bug cases, i.e., five introduced by the original developers and two injected, in four different MPI applications. Our experiments show that SyncChecker detects all the evaluated synchronization errors and provides helpful diagnostic information. Moreover, our experiments with seven NAS Parallel Benchmarks demonstrate that SyncChecker incurs moderate runtime overhead, 1.3-9.5 times with an average of 5.2 times, making it suitable for software testing.

Synchronization Error Detection of MPI Programs by Symbolic Execution

“…Another kind of tools, such as SyncChecker [1], Marmot [16], the Intel Trace Analyzer and Collector [17] and MUST [18], intercept the runtime MPI calls and record the running information of an MPI program. Then, using runtime information, runtime errors, synchronization errors, deadlocks or performance bottlenecks can be analyzed.…”

“…However, it also introduces synchronization errors [1], which frequently happen during optimizing MPI programs. The following situation shows a typical synchronization error: the application reuse the buffer of a send/receive call when the send/receive operation has not copied out/in the data in the buffer.…”

“…Most existing work uses dynamic program analysis methods [2] [1]. For example, Umpire [2] computes the checksum of the message data in a non-blocking send, and recomputes it in the corresponding completion check calls to detect synchronization errors.…”

“…For example, Umpire [2] computes the checksum of the message data in a non-blocking send, and recomputes it in the corresponding completion check calls to detect synchronization errors. SyncChecker [1] checks the temporal order of the related operations of a message buffer at runtime via instrumenting MPI programs. However, dynamic methods need to run the programs with some concrete inputs, which would miss input-related synchronization errors since they do not guarantee input space coverage.…”

Estudo e definição do teste de integração de software para o contexto de sistemas robóticos móveis

“…Algumas abordagens de teste de integração em programas concorrentes exploram comunicação e sincronização entre processos (BEN-ASHER et al, 2006;SOUZA et al, 2011;CHEN et al, 2012). Essas abordagens foram revisitadas, verificando similaridades com o teste de sistemas robóticos móveis que usam o modelo publish/subscribe e, assim, a abordagem aqui apresentada é inspirada nos critérios de teste para programas concorrentes (SOUZA; SOUZA; ZALUSKA, 2014).…”