The MULTICUBE Design Flow - Multi-objective Design Space Exploration of Multiprocessor SoC Architectures - page 16

structured database. The designer, not only will not be stuck on repetitive operations,
but can focus his/her attention (and profit from his/her experience) in analyzing the
designs database in a statistical manner.
Moreover, the automatic exploration is driven by an optimization engine based on
several optimization algorithms, whereas the manual exploration is based on designer
ability and experience to assess the results and to move towards the next instance of
the model to be simulated.
Figure 1.7 presents a direct comparison between manual and automatic optimiza-
tion. The global lead time (Tmo) of the manual design exploration is determined
by the number of manual iterations that are needed to reach the expected design
improvement. The global lead time (Tao) in the automatic procedure case is dom-
inated by the effective simulation time needed by the optimization strategy, given
the overhead due to the automatic extraction and processing of measures, and com-
munication latencies between the design exploration tool and the simulator. Another
difference concerns the type of results produced by both types of optimization: the
automatic optimization generates a set of designs that are likely to belong to the
Pareto front, while the manual optimization generates just a set of designs that the
designer found interesting. On the average, the manual approach suffers from the
fact that it is affected by personal views, experience and background. This decreases
the likelihood of finding points that belong to the Pareto front.
The reduction of the lead time of the design exploration phase is frequently re-
flected in the reduction of the overall time-to-market for the final product. The lead
time must be always considered as one of the objectives for the introduction of the
automatic exploration methodologies in industrial design processes, together with
the more specific product related ones.
The general automatic DSE procedure described in this Chapter represented the
common basis for the validation of each of the use cases of the MULTICUBE project.
In order to apply this procedure, the responsible for each of the use cases adapted

Tao

Automatic Optimization

Setup

Pareto curve

Design
Space

Setup

Manual Optimization

Solutions set

Tmo

Fig. 1.7 Comparison between manual and automatic exploration procedures