《09_doc_IAC_2016_landrover》.pdf

Verifying and Validating Automatically Generated Code Andrew Burnard, Land Rover Introduction The embedded software, developed in the automotive industry, has become increasingly more complex whilst the time to develop the systems has been reducing. This has required the process used to develop electronic control systems to make more use of computer simulation. The simulation allows earlier validation of the system being developed before representative vehicles are available. But there is still a significant effort required to develop and validate the code. With the advent of production code generation technology it is now possible to develop the code easily. But how much can the automatic code generation technology be trusted? This paper covers issues raised in validating the code produced by the use of an automatic code generator. Previously, when using manual generation of code, verification and validation is required to assure the quality of the code. Techniques that have been used to verify and validate manual code include: reviewing, module testing, integration testing, system testing, and static analysis. These techniques help to find errors that engineers make when developing code. When using an automatic code generator, there should be no random errors that are typical of manual code generation, but any errors should be more systematic because the tool should always to the same thing. Production code generation for embedded automotive controllers is a relatively new technology. The first production code generators were available around the turn of the century, so it is reasonable to be cautious with the code produced by a tool. The V-model is a useful representation to illustrate what is being covered by the verification and validation (VV) of automatically generated code. The le