LR>Low level tweaking - and I presume here you are speaking of using ass'y 
lang
LR>code to replace compiler generated code MAY speed up certain parts of the
Among other things.  Mostly though, tuning the type of code you write
for the particular compiler you are using.  In other words you are
changing the way you write code for the specific way your compiler
writes executable.  A very dangerous thing if the code is supposed to be
portable or you might end up compiling it with a different brand
compiler, or version of the same brand.
A profiler can certainly help you spot areas of your program that taking
too long, and the solution might be better algorithms, a better
compiler, etc.
If you need the speed, and have the free time to do it, and don't care
too much about portability, and have extremely limited systems it'll run
on, lower level optimizations can work.
But for portability, if you don't have twice as much time to spend
optimizing as you did writing, need it to run on multiple types of
processors or OS, or compiler under different compilers, etc., then it's
not worth it.
LR>Conclusion:  Low level tweaking is not worth it -
Yup.
--- QScan/PCB v1.19b / 01-0162
---------------