[GRASS-dev] Re: [GRASS-user] v.lidar.edgedetection very slow

[Hamish]

Jyothish Soman wrote:
> > I think that will be the parallel
> > version of the code given in the link. Sorry, I dont have
> > the resources right now to check the validity of the code,
> > my laptop is fried, working from a common terminal.
> > Just add -fopenmp flag in the make file for the code
Hamish:
> thanks!  small problem- `#pragma for` doesn't take shared() as an arg,
> for that do you have to do something like:
>  #pragma omp parallel for
> ????
Jyothish:
> Yes, shared is a default state, so that spawned an error.
> Also, you are right, it is parallel for.

with 
#pragma omp parallel for private(j,k,somma)
 or just
#pragma omp parallel for

I get a segfault:
Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7f11d6d96750 (LWP 4058)]
0x00007f11d6ab22ec in tcholDec.omp_fn.0 (.omp_data_i=0x7fffdeecf950) at TcholBand.c:37
37                      somma -= T[i - k][k] * T[i - k][j + k];
(gdb) bt f
#0  0x00007f11d6ab22ec in tcholDec.omp_fn.0 (.omp_data_i=0x7fffdeecf950)
    at TcholBand.c:37
No locals.
#1  0x00007f11d6ab2092 in tcholDec (N=0x7f11d6d47010, T=0x7f11c44f5010, n=40000, BW=401)
    at TcholBand.c:49
        i = -1
        j = 0
        k = 0
        end = 0
        somma = 6.9028182614708871e-310




also, is omp_set_num_threads(omp_get_num_procs()); really needed?
According to this article it sets default number of threads to number of
processors on system automatically:
http://software.intel.com/en-us/articles/threading-models-for-high-performance-computing-pthreads-or-openmp/

?



thanks,
Hamish