how to optimize this function using Tabu search

hello every body, i m trying to solve a GPU(x) problem using Tabu search i use F the cost function for evaluation:

 for k1=1:3;
 for k2=1:3;  
 for e1=1:10;
 for e2=1:10;
e1~=e2;
tabaffes(k1,e1)=1;
    tabaffee=bsxfun(@eq,SB(:),SB);
    tabaffee(k1,e1)<=tabaffes(k1,e1);
      tabaffee(k1,e1)<=tabaffes(k2,e2);
       tabaffee(k1,e1)=tabaffes(k1,e1)+tabaffes(k2,e2)-1;
       tabaffee(k1,e1)>=0;
      y(e1,e2)=sum(tabaffee(k1,e1)); 
F=sum(sum(coutl(k1,e1))*tabaffes(k1,e1))+sum(sum((couth(e1,e2))*(1-tabaffee(e1,e2))));

and for generating the nighberhood space i use a permutation function wich give me every possible permutation of the vector S:

j=1:10;
S=[1,1,1,2,2,2,2,3,3,3]
V=perm1(S,j);

and G the function of gain wich genearte the gain of each move :

G=sum(couth(e1,e2)+couth(e2,e1))*(tabaffes(k,e1))-sum(couth(e1,e2)+couth(e2,e1))*(tabaffes1(k,e1));

i need help for the implimentation of Tabu search programme wich give me the best S which minimise the cost function. thank you for helping me