Runtime overheads of C++ exceptions

Runtime overheads of C++ exceptions

[Dhananjai M. Rao]

Hello,

I am trying to figure out if C++ exceptions have any runtime performance
overheads.  In other words, let's say an exception is never thrown in my
c++ program, then would my c++ code run slower when compared with a
different version that does not use any exceptions (but does exactly the
same job).

thanks

with regards
-DJ

Re: Runtime overheads of C++ exceptions

[Eljay Love-Jensen]

Hi DJ,

Run your tests via...
g++ -fno-exceptions test.cpp -o test_without_exception_overhead
g++ test.cpp -o test_with_exceptions
...and compare the runtime performance overhead that is pertinent to your project.

Note that exception handling provides a mechanism that reduces program complexity and avoids often neglected unfilled contract situations that often go undetected because the return code indicating failure is ignored.  To be fair, you should factor in both of those issues into your analysis.

Generally speaking with most modern C++ compilers, there is NO runtime overhead for the exception framework in terms of CPU cycles wasted -- UNLESS an exception occurs.  Exceptions are "expensive" and should NEVER be used for mere flow control. 

There is some overhead in terms of memory footprint -- exception frames on the stack and such, used for unwinding, and the exception tables that take up some memory.  I presume those kinds of overhead costs are not much of a concern.

HTH,
--Eljay

Re: Runtime overheads of C++ exceptions

[Alex Vinokur]

"Dhananjai M. Rao" <dmadhava@ececs.uc.edu> wrote in message news:Pine.GSO.4.43.0309091132200.25498-100000@gatekeeper.ececs.uc.edu...
> Hello,
>
> I am trying to figure out if C++ exceptions have any runtime performance
> overheads.  In other words, let's say an exception is never thrown in my
> c++ program, then would my c++ code run slower when compared with a
> different version that does not use any exceptions (but does exactly the
> same job).
>
> thanks
>
> with regards
> -DJ
>
>

Look at
http://groups.google.com/groups?selm=bh5nja%24uqf37%241%40ID-79865.news.uni-berlin.de


   =====================================
   Alex Vinokur
     mailto:alexvn@connect.to
     http://mathforum.org/library/view/10978.html
   =====================================