So what can be done? One way to validate algorithms is to implement them in isolation. This works great, unless you need another 10K lines of code to actually exercise your algorithm. In this case, you want some way to emulate the rest of the system, and create some type of test bench for your code. In this case, it would be an environment for implementing scheduling algorithms for RTEMS that has all of the interfaces to the scheduler. A similar project is LinSched, which is an infrastructure for trying out Linux scheduler algorithms. This is one of the goals of the GSoC project that I am proposing to do, and I will post more on this later. :)
For now, I'm going to try and validate that my code actually works, and doesn't just "run to completion."