An Interactive Tuning Support for Processor Allocation of Data-Driven Realtime Programs

This paper presents the effectiveness of an interactive support facility to tune processor allocation of data-driven realtime programs on CUE (Coordinating Users’ requirements and Engineering constraints)-series super-integrated dynamic datadriven processors. In tuning processor allocation of data-driven programs on CUE processors, one of the most critical issues is how to detect and remove sideeffects on processor loads. Especially, in multiple processing of some streams, it must be clearly identified how a peak of instantaneous load convergence comes into existence and which portion of program causes the peak. The authors have been proposed a development environment named Realtime Execution System for CUE-series data-driven processors (RESCUE). CUE processors have ideal multi-processing capability without any interference among independent processes running simultaneously. Hence, the runtime performance of CUE processor can be predicted by static analysis of the program as long as pipeline hardware resources are not problems. This paper first proposes a support facility to tune processor allocation of a program to prevent overloaded status. This facility supports a user to clearly know how the peak of load comes existence to identify which portion of program causes the peak. Then a user can change processor allocation of a portion of program using this facility. This paper then preliminary evaluates this facility by applying this facility to tune a realtime image processing program as an practical example. As a result, this paper shows that this facility can support not only detection of the bottleneck of a realtime system based on CUE processors, but also tuning of CUE processor architecture.