CompuServe Messages

You're right, it's a hole in the design. However, you can approximate the latency by building some code around a forced interrupt call to see what the overhead in the kernel is for a simple task switch. Not accurate, but it will get you within an order of magnitude.