Timing Measurements for a Context Switch

I am making a simulator for which i  need to do some performance
analysis . The measurements i need to take are :

1. The time a context switch takes for execution. Does this vary
greatly depending on the machine arch. ?

2. There is a block with if statements within a for loop . Is there a
way to measure them to a finer resolution so that the time obtained is
in some comparison to the context switch ?

Actually doing a comparison between a tradeoff between 1 and 2. Are
there any built in linux system calls for this ? Do i have to use a
tool ? Have heard of a tool called lmbench, but not used it .

Thanks,
vivekian