Mail Archives: djgpp/1997/06/03/12:04:30
> AFAIK, `gcov' is not a profiling tool.
If I used my own language, I wouldn't describe it as a profiling
tool either, since it does not include CPU time.
Hence my reference to BFA, which seems a better description.
However, the manuals for gcov do list it as a profiling tool,
which is why I presented it as such..
info gcc gcov -> gcov intro
Introduction to 'gcov'
<snip some intro stuff>
You can use 'gcov' as a profiling tool to help discover where
your optimization efforts will best affect your code. You can
also use 'gcov' along with the other profiling tool, 'gprof'.
Profiling tools help help you analyze your code's performance. Using
a profiler such as 'gcov' or 'gprof' ...
> It is a program that shows you
> the coverage (hence its name) of your program's code by the recent
> run(s). Using it, you can tell which parts of your code has never been
> executed (due to combination of if's that never fired), and devise test
> cases which will excercise these code fragments (since code that has
> never been tested tends to have bugs). `gcov' generates a report which
> shows how many times was each line executed, which might give you some
> idea about where the hot spots might be.
This is correct AFAIK, except it believe you can specify the resolution
that you want. Either per line of C code or per function.
Unfortunately, I can't get the thing to run. Gcov.exe itself if fine
(untested, but should be OK),
its just that the tooled executable itself does not produce a .ba results file
when run. I think libgcc.a is OK, and that it is the actual code
itself that is wrong, which means waiting for a new compiler snapshot.
Andrew
- Raw text -