|
From: David F. L. <d.f...@gm...> - 2014-12-01 11:33:16
|
Hello everyone, I'm trying to profile my code using callgrind. I use this command: valgrind --dsymutil=yes --tool=callgrind --trace-children=yes --separate-threads=yes --instr-atstart=no --collect-atstart=no --collect-systime=yes <my_program> The results show a lot of functions with 0 CPU cycles, but these functions are called a lot of times. For example: fn=(13976) aes_icm_encrypt cfn=(13978) aes_icm_encrypt_ismacryp calls=2883 0 0 0 1 13 But, the output gives correct information about functions in the same library. Theses functions have a high overhead and they are the bottleneck of my application. Do you have any clue why I don't have CPU cycles information about these functions? Thanks in advance. Regards. -- David Fernández López |