[Bug binutils/747] Profiling of nested functions

[laurent at guerby dot net]

------- Additional Comments From laurent at guerby dot net  2005-03-15 20:20 
-------
Thanks Nick for your answer. I missed the -l flag in the documentation, very
useful indeed!

The per "line" profile looks indeed correct, but
the call graph suffers from this nested=parent assimilation and shows p2 as a
recursive function (if I understand correctly):

index % time    self  children    called     name
                0.12    0.00       1/1           main (cn.c:21 @ 80485dc) [6]
[5]      0.7    0.12    0.00       1         p3 (cn.c:14 @ 804856d) [5]
-----------------------------------------------
                0.00    0.00       1/20000001     main (cn.c:21 @ 80485dc) [6]
                0.00    0.00 20000000/20000001     p2 (cn.c:12 @ 8048518) [2]
[11]     0.0    0.00    0.00 20000001         p2 (cn.c:7 @ 804850d) [11]
-----------------------------------------------
                0.00    0.00       1/1           main (cn.c:20 @ 80485d7) [38]
[12]     0.0    0.00    0.00       1         p1 (cn.c:3 @ 80484c4) [12]
-----------------------------------------------

I feel that treating nested functions as normal functions would be more helpful
for GNU Ada and Pascal for example, but this is a new feature request no longer
a bug. I don't know if it's easy to do in gprof, GCC 4.0 transforms all nested
functions to top level functions (with arguments) during code generation, it
wasn't doing so before so may be it helps.

I will try "-l" on the larger original Ada software and report back in a few 
days.

Thanks again for your time.

Laurent

-- 


http://sources.redhat.com/bugzilla/show_bug.cgi?id=747

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.