Invasive Computing - Common Terms and Granularity of Invasion
share
Future MPSoCs with 1000 or more processor cores on a chip require new means for resource-aware programming in order to deal with increasing imperfections such as process variation, fault rates, aging effects, and power as well as thermal problems. On the other hand, predictable program executions are threatened if not impossible if no proper means of resource isolation and exclusive use may be established on demand. In view of these problems and menaces, invasive computing enables an application programmer to claim for processing resources and spread computations to claimed processors dynamically at certain points of the program execution. Such decisions may be depending on the degree of application parallelism and the state of the underlying resources such as utilization, load, and temperature, but also with the goal to provide predictable program execution on MPSoCs by claiming processing resources exclusively as the default and thus eliminating interferences and creating the necessary isolation between multiple concurrently running applications. For achieving this goal, invasive computing introduces new programming constructs for resource-aware programming that meanwhile, for testing purpose, have been embedded into the parallel computing language X10 as developed by IBM using a library-based approach. This paper presents major ideas and common terms of invasive computing as investigated by the DFG Transregional Collaborative Research Centre TR89. Moreoever, a reflection is given on the granularity of resources that may be requested by invasive programs.
READ FULL TEXT