Due to the continuously increasing complexity of MPSoC
and tight time-to market requirements, virtual platforms have
become a widely adopted solution to achieve concurrent
hardware/software design for embedded architectures. These
platforms overcome the stagnation in developing
multiprocessor designs as well as the parallel applications on
multicore hardware. The market for virtual platform tools is
maturing. The abstraction of some levels from the design flow
is really important to integrate different architectures and
enables discovering the interactive exploration of MPSoC at
virtual prototyping level.
In the present work, we use the OVP technology. As
hardware platforms are the backbone of embedded systems,
this technology enables designers to make complex
architectures at early stage in the design of MPSoC debug and
test their architectures and software. To achieve our goal, we
developed homogeneous multiprocessor architectures using
different types of processors with two memory hierarchies.
Then, we made heterogeneous designs. Inter-processor
communication is assured by the shared memory as mapping
tasks on each core and synchronization between cores are
really important. We also proposed architectures that support
the integration of operating systems (AMP and SMP Linux).
These architectures are composed of several peripherals. In
addition, the system can handle the sharing of memory between
processors and the distribution of the workload. We also tried
the parallelism processing. This early integration of software
and hardware is the key solution to make various integration
alternatives, to select the best solution which meets constraints
and take decision of the optimal solution.
We plan in our ongoing work to integrate a multiprocessor
Real-time Operating System (RTOS) for the heterogeneous
multiprocessor system behavior. The system should be also
able to handle the distribution of the workload between the
different cores. As well, the ability to visualize the Central
Processing Unit (CPU) utilization and other key variables
across multiple cores provides a solid foundation to new
products.