Service Communautaire d'Information sur la Recherche et le Développement - CORDIS

Case studies - methodology

In this result we have gathered a collection of benchmarks from the pervasive domain. These benchmarks can be used by people to try and automatically generate "optimal" code, code that uses few instructions, few memory cells, few milli-amps when executing, etc.

These benchmarks can be compared to benchmarks such as the embassy benchmarks (for multimedia operations), linpack (numerical algorithms) and the splash benchmarks (for parallel programming). The unique selling point is that the programs are programs at the heart of pervasive systems.

The commercial value of the set of benchmarks is hard to measure, but previous sets of benchmarks have accelerated processor and compiler design, as benchmarks can be used as a standard to compare systems, compilers, or processors. (Having said that, benchmarks become obsolete once system designers start to target specific benchmarks, such as Dhrystone, and figures become meaningless).

Our target audience is developers of compilers, whether free software (gcc), academics, or commercial (.net, java). Over the past few years a steady stream of experimental compiler work has found its way into mainstream compilers, for example Java generics. We hope that making available source code of some basic algorithms, together with a statement as to what can be achieved, compiler writers will attempt to implement some of the specialisation algorithms in mainstream compilers.

Reported by

Department of Computer Science, University of Bristol
Merchant Venturers Building, Woodland Road
BS8 1UB Bristol
United Kingdom
See on map