[Help-glpk] [Fwd: glpk tuning]

[Andrew Makhorin]

-------- Forwarded Message --------
From: Karol Rydzewski <address@hidden>
To: address@hidden
Subject: glpk tuning
Date: Mon, 21 Sep 2015 23:37:03 +0200

Hi,

I would like to make use of glpk library in my C programme. I have made
some tests of it in Octave and it works great, giving an expected
results. However using "raw" glpk via C gives different, wrong, results.
I suspect, that it's a matter of proper parametrisation of the library,
but I am a novice and a real problem may be hiding somewhere else. I
have tried to set the same defaults that glpk has in octave but it
turned out to be a lot of fiddling around and the way to success seems
vague. Can you provide me steps I should perform to make glpk work well
in C as it does in Octave?

Bellow I am providing my sample code in Octave and its output(correct)
followed by C code and its output(wrong).

Best regards,
Karol

        OCTAVE CODE:
                c=[1,0,0,0,0];
                A=[1, 0, 1, 1, 0;
                0,1,1,1,0;
                0,1,0,0,1;
                1,1,1,0,1];
                b=[1.0969,0.8416,0.2007,0.5986]';
                lb=[0;0;0;0;0];
                s=-1;
                ctype=["S","S","S","S"];
                vartype=["C","C","C","C","C"];
                param.msglev=3;
                
[xmin,fmin,status,extra]=glpk(c,A,b,lb,ub,ctype,vartype,s,param);
        
        OCTAVE OUTPUT:
                GLPK Simplex Optimizer, v4.55
                4 rows, 5 columns, 12 non-zeros
                Preprocessing...
                4 rows, 5 columns, 12 non-zeros
                Scaling...
                 A: min|aij| = 1.000e+000  max|aij| = 1.000e+000  ratio
                = 1.000e+000
                Problem data seem to be well scaled
                Constructing initial basis...
                Size of triangular part is 3
                      0: obj =  3.979000000e-001  infeas = 1.426e-001
                (1)
                *     1: obj =  3.979000000e-001  infeas = 0.000e+000
                (0)
                OPTIMAL LP SOLUTION FOUND
                fmin= 0.25530
                xmin=[  0.25530; 0.00000; 0.14260;  0.69900;  0.20070 ]
        
        ##############################
        C CODE:
            int i,j;
            int X[13],Y[13];
            double *pathPerf, coeff[13];
            double o1,o2,o3,o4,o5;
            glp_smcp smcp;
        
            glp_init_smcp(&smcp);
            smcp.msg_lev=GLP_MSG_ALL;
        
            //glp problem structure
            glp_prob *problem;
        
            //init problem structure
            problem=glp_create_prob();
        
            glp_set_obj_dir(problem, GLP_MIN);
        
            //define number of rows and columns in the matrix
            glp_add_rows(problem, 4);
            glp_add_cols(problem,5);
        
            //seting bounds
            for(i=1;i<5;i++)
                glp_set_col_bnds(problem,i,GLP_LO,0.0,0);
        
            //right-hand values of each equation
            glp_set_row_bnds(problem,1,GLP_FX,1.0969,0);
            glp_set_row_bnds(problem,2,GLP_FX,0.8416,0);
            glp_set_row_bnds(problem,3,GLP_FX,0.2007,0);
            glp_set_row_bnds(problem,4,GLP_FX,0.5986,0);
        
            //objective func. coefficients
            glp_set_obj_coef(problem, 1, 1.0);
            glp_set_obj_coef(problem, 2, 0.0);
            glp_set_obj_coef(problem, 3, 0.0);
            glp_set_obj_coef(problem, 4, 0.0);
            glp_set_obj_coef(problem, 5, 0.0);
        
            X[1]=1,Y[1]=1,coeff[1]=1.0;
            X[2]=1,Y[2]=3,coeff[2]=1.0;
            X[3]=1,Y[3]=4,coeff[3]=1.0;
            X[4]=2,Y[4]=2,coeff[4]=1.0;
            X[5]=2,Y[5]=3,coeff[5]=1.0;
            X[6]=2,Y[6]=4,coeff[6]=1.0;
            X[7]=3,Y[7]=2,coeff[7]=1.0;
            X[8]=3,Y[8]=5,coeff[8]=1.0;
            X[9]=4,Y[9]=1,coeff[9]=1.0;
            X[10]=4,Y[10]=2,coeff[10]=1.0;
            X[11]=4,Y[11]=3,coeff[11]=1.0;
            X[12]=4,Y[12]=5,coeff[12]=1.0;
        
            glp_load_matrix(problem, 12, X, Y, coeff);
            glp_simplex(problem, &smcp);
        
            o1=glp_get_col_prim(problem,1);
            o2=glp_get_col_prim(problem,2);
            o3=glp_get_col_prim(problem,3);
            o4=glp_get_col_prim(problem,4);
            o5=glp_get_col_prim(problem,5);
        
            printf("fmin:%f\n",glp_get_row_prim(problem,  1));
            printf("xmin:%f,%f,%f,%f,%f\n",o1,o2,o3,o4,o5);
        
            //cleanup
            glp_delete_prob(problem);
            glp_free_env();
        
        C OUTPUT:
                GLPK Simplex Optimizer, v4.55
                4 rows, 5 columns, 12 non-zeros
                      0: obj =   0.000000000e+00  infeas =  2.738e+00
                (4)
                      4: obj =   3.979000000e-01  infeas =  5.810e-02
                (1)
                LP HAS NO PRIMAL FEASIBLE SOLUTION
                fmin:1.038800
                xmin:0.397900,0.200700,0.000000,0.640900,0.000000