how to use openmp in a c++ program

Question

I would like to use OpenMP to parallelize some of my functions in a c++ program.

I am using ubuntu 12.04, on intel i5 with 4 cores. But after following certain steps, I do not see any improve in performance. I could see that only one CPU core is being used. (system monitor in ubuntu )

What I did..

added #include <omp.h>

added these two lines before a for loop

omp_set_num_threads(4);

#pragma omp parallel for

in CMakeLists.txt,

I added target_link_libraries (executable -fopenmp -lgomp ${PCL_LIBRARIES} )

Can you please help me in getting the parallelism ?

Thank you!

Answer 1

As mentioned by @Mikael Persson , I removed -fopenmp from the target link libraries and added find_package(OpenMP) if (OPENMP_FOUND) set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${OpenMP_C_FLAGS}") set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OpenMP_CXX_FLAGS}") endif()

This is working.

The error r *** double free or corruption (!prev): 0x00007ff424006b20 *** was arising because in the for loop, i was trying to use push_back() function to load a stack and the index was dependant on the for loop index. I think that was the cause of the problem.

I learnt one more thing that one can have nested for loops and OMP can work without errors if their variables are independent and produce valid results