race condition.......

[Intel_C_Intel]

I am not sure if its a right forum thread, if not let me 
know I will move my question to appropriate one. I need some help to 
detect race condition inside my code. Here is the part which is 
parallelized.

!$omp parallel default(none)&
!$omp shared(GKL, IEL, NR, MAXBWL, NRS) 
!$omp do lastprivate(L, CP, I, J, K, N, M, PIVOT, MR) DO N=1,NRS 
M = N - 1 
MR = MIN0(MAXBWL,NR-M) 
PIVOT = GKL(IEL,N,1) DO L=2,MR 
CP = GKL(IEL,N,L)/PIVOT 
I = M + L 
J = 0 DO K=L,MR 
J = J + 1 
GKL(IEL,I,J) = GKL(IEL,I,J) - CP*GKL(IEL,N,K) END DO
GKL(IEL,N,L) = CPEND DOEND DO
!$omp end do
!$omp end parallel
I am having problem with 2D matrix, and its really hard to detect 

because when I am using small example something like 8x8 matrix everything 
works great, but when I am running bigger case at some point it start 
miscalculating decimal numbers, for example instead of 0.3122880E-04 it 
gives me 0.3125670E-04. 
If someone could help me figure this out I would be highly appreciated.
Thanks in advance.

[Intel_C_Intel]

Wow I cant believe that no one can help me here.

[TimP]

You could help yourself by running Intel Thread Checker, if you don't care to clarify your example. If the question were more specifically about threading, you could try the threading forum. If you are using single precision, it may not be surprising that changes in threading produce numerical changes such as you mention.