5.14. Platform Design

These same system-level design techniques can be used to develop reusable platform architectures. In particular, the ability to specify the structure and architecture of a platform subsystem, to simulate and analyze it, and to iterate on the number and types of processor(s) and the structure of the associated memory hierarchy and communications subsystems, are very helpful to SOC designers who have previously developed their system designs based more on gut feel and the mechanical selection of old, familiar system architectures rather than quantitative simulation results that produce more efficient, better optimized designs.

Often, what’s missing is the ability to drive platform design with a product’s actual application code, which may not be ready at such an early stage in the product’s development. Using the top-down design methodology discussed above, the final code isn’t essential for architectural development. Instead of the final application code, similar or related code or code kernels drawn from the general application domain (for example, audio or video encoding or decoding algorithms from a previous generation) can help determine many of the architectural requirements and processor optimizations.

Another possibility is to use artificial code sequences that generate idealized traffic patterns and consume processing and communications resources. Such traffic-generation code can be used to characterize the capabilities of a platform architecture. Once such an architecture has been defined using quantitative system-simulation results, systems and software applications specialists can map the applications onto the target platform and further develop the actual system implementation.