David,

I've used Polyspace (it's called something else now it's been acquired by Matlab) for in depth static code analysis. It's strength is that it will find run time exceptions without the tester having to figure out a comprehensive set of test inputs.

More recently I've been using Tessy from Razorcat. This automatically generates a test harness and compares actual output with expected results, and it can do this on the target for some targets. If your target isn't support, you can run your code on the PC. The tester needs to work out test cases and expected results, but it does provide help in the form of a classification tree editor and an execution path trace. It's a pretty good tool to prove that your functions meet your requirements (you do have requirements for your functions, don't you?). As it can be automated, it's really, really useful for regression testing.

I know of people who use RTRT, but support seems pretty patchy these days and I haven't used it myself.

Whichever way you go, you will probably have to put some effort into your code to make it testable. Once your are familiar with the issues, you will probably start to code with a 'design for test' mentality, and you will get better producivity from your tools.

Although the cost of these tools look steep, they are generally very cost effective.