I've noticed that changing the c language dialect to C99 from GNU99 roughly doubles the compilation speed.
We also noticed that using macros extensively (define statements) especially nested macros are REALLY slow on xcode. Removing as much of those as possible also helped tremendously and our compilation is now 2-3 times faster.
Update: Since that we also discovered that includes within headers and in general over-use of include statements was also a contributor to performance issues. By reducing the number of includes we were able to further increase the compilation speed significantly well beyond the original project. I had incorrectly assumed that due to the usage of an SSD and caching having the typical C #ifdef guards against double include would be enough to make compilation fast but it seems LLVM is really slow with such includes for some reason.