Q/A Solution 2

Q- What is the difference between the two solutions proposed for C?

A- In the first solution the parallel sections are opened inside the main, while in the second are restricted inside the functions. This makes the program easier to read and to modify for future improvements, since the wider a parallel section is, the most dangerous it can become when dealing with thread-safety.

Another change made in solution 2 is in the "evolve" function. Inside the first for loop, the temporary variables are declared inside the loop instead of outside, as in solution 1:

 #pragma omp parallel for
   for(int iy=1;iy<=NLY;++iy)
     for(int ix=1;ix<=NX;++ix){
       float temp0 = temp[((NX+2)*iy)+ix];
       temp_new[((NX+2)*iy)+ix] = temp0-(0.5*dt*(temp[((NX+2)*(iy+1))+ix]-temp[((NX+2)*(iy-1))+ix]+
       temp[((NX+2)*iy)+(ix+1)]-temp[((NX+2)*iy)+(ix-1)]))/dx;
     }

This may solve a thread-safety issue, since, if declared outside, the "temp0" variable is considered shared while it should be private (of course, another solution is to declare it as private in the "parallel for" pragma).

Notice that the problem wouldn't sussist if the parallel section began inside the main, as in solution 1 (in such case the temp0 variable is already declared inside the parallel section).

Warning: depending on the compiler of your choice, for solution 2 you may need to add the flag -std=c99 because of some variable definitions that require C99 standard.