The Embedded Java Benchmark Suite JemBench

TheEmbeddedJavaBenchmarkSuiteJemBench
Martin Schoeberl Thomas B. Preußer Sascha Uhrig
Department of Informatics and Department of Computer Department of Computer
Mathematical Modeling Engineering Science
Technical University of Technische Universität University of Augsburg,
Denmark Dresden, Germany Germany
masca@imm.dtu.dk thomas.preusser@tu- uhrig@informatik.uni-
dresden.de augsburg.de
ABSTRACT which perform data-parallel and pipelined execution.
As the benchmark suite does not target consumer electronics, butRequirements to embedded systems increase steadily. In parallel,
embedded control systems, the benchmarks are mainly designed inalso the performance of the processors used in these systems is
a WCET (Worst Case Execution Time) analyzable way. Hence, aimproved leading to multithreaded and/or multicore architectures.
target platform can be examined concerning their real-time capabil-Depending on the type of the embedded system, using Java is a
ity by performing a static WCET analysis and compare the resultsmore and more popular way for software development. In this pa-
to the actual execution. Of course, the real-time capability of theper, we present a Java benchmark suite that enables the comparison
used platform cannot be proven but an estimation of the WCETof different embedded Java platforms while solely assuming the
analysis’ tightness can be obtained.availability of a CLDC API, the minimal configuration defined for
Depending on the platform, the execution time of ...