I saw an article this morning that came from VDC, a research company that has been compiling statistics about the embedded systems space. The article, titled Virtual Platforms – White Knight or Red Herring, was an interesting read. One paragraph in particular stood out to me.

Our research shows that projects using VSPS tools were reported to be late at approximately the same rate as within the embedded market as a whole (a dismal 43.4% versus 42.7%, respectively).  While it is certainly possible that projects using VSPS could have been more late without their use due to the level of complexity attributable to other project factors and requirements, the difficulty associated with using another, new tool could have also unnecessarily delayed the project as well.

The key in that paragraph is complexity. When ESL tools first became available, most designs were single processor with a few peripherals on a bus. There were few surprises at the system level, performance was predictable and software could be executed just as easily on an alternative host most of the time without having to worry much about timing or functional differences. Or in other words – the systems were too simple to warrant using a whole new class of tools, and let’s face it most of the ESL tools at that time could only handle these simple systems.

When we look at the users of ESL tools today, we find that the systems they are working on are multiprocessor, multiple busses, lots of communications going on around the system, lots of asynchronous activities. These systems are almost completely unpredictable without running simulations or designing the software in such a way that they can be made to behave in a predictable manner. So when VDC says it is surprising that users of system level virtual prototypes see no improvement in the rate at which their project are late – I see that as a HUGE gain. Given the enormous increase in complexity of these systems and the number of additional and unexpected ways to get caught out, then one would expect the bug rates to spike through the roof. These tools have managed to put a lid on it, and over time I am sure will reduce it, plus as more software engineers get used to thinking about concurrency, then some of the problems will go away as best practices become normally accepted coding practices.

---------------------------------------------------
Brian Bailey – keeping you covered