The creator of the LinFu-library is working on the ‘world’s fastest IOC container’, called Hiro. Basically Hiro reads once the configuration and then it compiles everything to a fixed, immutable container. The compiled container has then ‘burned in’ the wiring of all components in IL-Code. So there’s no state, no map lookups and no locking required. This makes Hiro a whole magnitude faster than traditional IOC-Containers like AutoFac, Ninject etc.
Of course, due the container is compiled, it isn’t as flexible as a non-compiled container. But I think it will serve its purpose well. Most of the fancy features aren’t that important.
That said, I think most of the projects don’t really care about the speed of their IOC-container. The average application doesn’t create hundreds or thousands of service-instances per second. So for the average application there’s no need for a faster container. However I can image application which will benefit massively from a faster one.
So for my projects I’ll continue to use the wonderful AutoFac-container.
More info on Hire: