We use g++ 4.2.4 and I'm trying to track down some performance problems in my code.
I'm running gprof to generate the profile, and I'm getting the following "strangeness" in that the most expensive function is __tcf_0:
Each sample counts as 0.01 seconds.
% cumulative self self total
time seconds seconds calls ms/call ms/call name
40.00 0.04 0.04 1 40.00 95.00 __tcf_0
This function then appears to calls most of my user functions (ie. it is the one that's called from main). The nearest explanation that I found for this was here, but that link refers to static objects and atexit, and I don't think this applies in my case.
If it's helpful, I'm using Boost (program_options and fusion) and the HDF5 libraries.
UPDATE:
The command I use when building is:
g++ -Wreturn-type -Wunused -Winline -pg -DLINUX -DHAS_SETENV \
-DFUSION_MAX_MAP_SIZE=15 -DFUSION_MAX_VECTOR_SIZE=15 -g -O0 \
--param large-function-growth=300 --param inline-unit-growth=200
__tcf_0 seems indeed to be a function which calls destructor of static objects and which is registered for each static objects, to be called at exit (taking for granted what is said on this page)
Now, the result of your gprof is quite strange, since the function which takes most of the time only takes 0.04 seconds, which means the whole program takes 0.1 s to execute. If I'm not mistaken, my guess is that you didn't profile correctly. Did you compile your code with profiling enabled?
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With