You might want to write the whole thing in assembly if you want exactly 100 cycles.
Or better yet if you don't want the CPU doing anything during these delay times just create a custom instruction that takes 99 cycles to complete. This should be a simple as loading a down counter with 100 when the 'start' bit is asserted and when the counter reaches '1' assert the 'done' bit. Then when the counter hits 0 hold it until the next time the custom instruction is run.
By the way what are you trying to do by stalling the CPU for 100 cycles?