Random Number Simulated Annealing

Accelerated Computing CUDA CUDA Programming and Performance
13

I took some time to test the code you suggested, but finally I did. Indeed when using the: x = LFG175shm() % 50 + a, in each iteration, the numbers generated are the same:

LFG175gmLoadToShm (gfibreg);

while (T > 0.0001){

	while ((pos1 == 0) || (pos2 == 0) || (pos1 == *municipios) || (pos2 == *municipios) || (pos1 == pos2)){

		pos1 = LFG175shm() % 50 + 1;

		pos2 = LFG175shm() % 50 + 1;

	}

	LFG175gmSaveFromShm(gfibreg);

	

	int temp;

	temp = solutionM[pos1];

	solutionM[pos1] = solutionM[pos2];

	solutionM[pos2] = temp;

	(...)

	T = T * 0.9999;

}

LFG175gmSaveFromShm(gfibreg);

The way it is, the code simply change the same two positions when it had to change two random positions. So I put the code:

while ((pos1 == 0) || (pos2 == 0) || (pos1 == *municipios) || (pos2 == *municipios) || (pos1 == pos2)){

	pos1 = LFG175shm()/(0x8000000>>6) + 1;

	pos2 = LFG175shm()/(0x8000000>>6) + 1;

}

but the numbers are greater than 64, and are always the same. There’s a way to fix it, sigismondo?

Again, thanks for the answers!