The MULTICUBE Design Flow - Multi-objective Design Space Exploration of Multiprocessor SoC Architectures - page 13

9

< item value= "2lev" />

10

< item value= "comb" />

11

</ parameter >

12

...

</ parameters > 13

For each parameter an unique name must be provided. The parameters types can be
divided into two categories: scalar types, variable vector types. Scalar types can be
integer , boolean (a subset of integers), exp2 (progression of power of 2) and string
(a type for defining categorical variables). Vector types can be used to describe
combination of boolean values ( on-off-masks or permutations ). In particular, on-off-
masks can be useful for describing the space of active processors while permutations
can be used to describe the mapping of tasks on the available processors.

1.3.1.3 System Metrics Definition

The < system_metrics > section is used by the use case and simulator provider to specify
the names, the types and the units of the system metrics that can be estimated by the
simulator:
1

< system_metrics >

2

< system_metric name= "cycles" type= "integer" unit= "cycles" />

3

< system_metric name= "instructions" type= "integer" unit= "insts" />

< system_metric name= "powerconsumption" type= "float" unit= "W" /> 4

< system_metric name= "area" type= "float" unit= "mm2" />

5

</ system_metrics >

6

A complex expression of the system metrics can be defined as one of the objective
of the exploration algorithm.

1.3.2 Simulator Input/Output XML Interface

The simulator input file contains a preamble and a sequence of < parameter > sections
where, for each parameter, the name and the value is specified. The number of
< parameter > sections and the name of the parameters should be the same as defined in
the XML Design Space description file. Similarly the simulator output file contains
a preamble and a sequence of < system_metric > sections where, for each metric, the
name and the value is specified. Besides, an appropriate error reporting syntax has
been described in the specification.

1.4 Advantages of Automatic DSE

The procedure to assess the benefits of the introduction of an Automatic Design
Space Exploration (or Optimization Methodology) has to address, not only the final
objective quality and the improvement of the target design but also the impact of such
a technology on the entire design process of embedded computing platforms. The