This paper will give an overview of multicore in automotive applications, covering the trends, benefits, challenges, and implementation scenarios.
The automotive silicon industry has been building multicore and multiprocessor systems for a long time. The reasons for this choice have been: increased performance, safety redundancy, increased I/O & peripheral, access to multiple architectures (performance type e.g. DSP) and technologies. In the past, multiprocessors have been mainly considered as multi-die, multi-package with simple interconnection such as serial or parallel busses with possible shared memories. The new challenge is to implement a multicore, micro-processor that combines two or more independent processors into a single package, often a single integrated circuit (IC). The multicores allow a computing device to exhibit some form of thread-level parallelism (TLP).
The automotive industry is also increasing complexity and safety with new standards such as IEC61508 and ISO 26262 being implemented. This will enable new systems X-by-wire. To achieve the certification, the electronic architecture will have to be modified to be SIL3 (safety integrity level) compliant. Dual-core is a good candidate with the possibilities of asymmetrical, symmetrical and lockstep configurations.
Developing automotive applications is also bound by specific processes and development methodologies. It requires following guide-lines, recommendations, best-practices and standards e.g. AUTOSAR and OSEK.
The amount of software that is built by auto-code generation may reach more than 50% in some applications. The use of multicore processors requires re-inventing tools: performance modeling, benchmarking of multiprocessor systems, automatic load-balancing, multiprocessor debugging and on chip instrumentation, calibration and fast prototyping.