The MULTICUBE Design Flow - Multi-objective Design Space Exploration of Multiprocessor SoC Architectures - page 14

benefits on the process can be measurable and tangible like the reduction of the overall
design process lead time, and qualitative or intangible like the streamlining and the
reduction of human error prone repetitive operations. These benefits are particularly
valuable for design problems where the number of configuration parameters to be
explored is quite large, like in MP-SoC designs.
The specific design process activities can be analyzed and classified to measure
the various performance indicators. In a general way, we can consider the following
steps as the basis for any manual design space exploration or optimization process:

￿ Model Setup: preparation of an initial model of the virtual platform;

￿ Simulation: execution of the simulation of the executable model with a single

configuration of parameters;

￿ Results Assessment: meaningful measures are extracted and compared with

historical and expected ones;

￿ Model Edit: the model is manually modified and resubmitted for a further analysis.

Figure 1.5 represents the steps of a typical manual exploration procedure. In a manual
approach, the exploration of the design space is done by subjective assumptions of
the human designer, who will modify at most one or two parameters per evaluation.
The model simulation corresponds to a minimal portion of the time of the whole
exploration procedure. A large amount of time is spent by the designer editing the
configuration parameters and analyzing the results. There is also an Idle Time (from
the point of view of the use of computational resources) that lasts from the end of
the simulation till the moment in which the human operator is informed about it
and handles the simulation tools to get the results. This idle time can be very short
if the designer is immediately informed about the end of the simulation or can be
significant if the designer is not on duty.
The automatic design space exploration process can be defined by identifying the
following basic steps:

￿ Model setup: the model must be correctly parameterized in order to be easily

managed by the automatic exploration tools;

Simulation Time

Results Assessment

Setup

Model Edit

Idle Time

Fig. 1.5 Manual exploration procedure