Scilab Home page | Wiki | Bug tracker | Forge | Mailing list archives | ATOMS | File exchange
Please login or create an account
Change language to: English - Português - Русский - 日本語

Please note that the recommended version of Scilab is 6.0.0. This page might be outdated.
See the recommended documentation of this function

Aide de Scilab >> Optimisation et Simulation > qp_solve


linear quadratic programming solver builtin

Calling Sequence

[x [,iact [,iter [,f [,info]]]]] = qp_solve(Q, p, C, b, me)



real positive definite symmetric matrix (dimension n x n ).


real (column) vector (dimension n)


real matrix (dimension (me + md) x n). This matrix may be dense or sparse.


RHS column vector (dimension m=(me + md) )


number of equality constraints (i.e. x'*C(:,1:me) = b(1:me)' )


optimal solution found.


vector, indicator of active constraints. The non zero entries give the index of the active constraints. The entries of the iact vector are ordered this way: equality constraints come first, then come the inequality constraints.


2x1 vector, first component gives the number of "main" iterations, the second one says how many constraints were deleted after they became active.


integer, error flag. If it is present and qp_solve encounters an error, then a warning is issued. The current results are returned, so in this case they are probably inaccurate.


This function requires Q to be symmetric positive definite. If this hypothesis is not satisfied, one may use the contributed quapro toolbox.


// Find x in R^6 such that:
// x'*C1 = b1 (3 equality constraints i.e me=3)
C1= [ 1,-1, 2;
     -1, 0, 5;
      1,-3, 3;
      0,-4, 0;
      3, 5, 1;
      1, 6, 0];

// x'*C2 >= b2 (2 inequality constraints i.e md=2)
C2= [ 0 ,1;
     -1, 0;
      1, 0];
b2=[ 1;-2.5];

// and minimize 0.5*x'*Q*x - p'*x with
p=[-1;-2;-3;-4;-5;-6]; Q=eye(6,6);

[x,iact,iter,f]=qp_solve(Q,p,[C1 C2],[b1;b2],me)
// Only linear constraints (1 to 4) are active

See Also

The contributed toolbox "quapro" may also be of interest, in particular for singular Q.

Memory requirements

Let r be


Then the memory required by qp_solve during the computations is

2*n+r*(r+5)/2 + 2*m +1


  • Goldfarb, D. and Idnani, A. (1982). "Dual and Primal-Dual Methods for Solving Strictly Convex Quadratic Programs", in J.P. Hennart (ed.), Numerical Analysis, Proceedings, Cocoyoc, Mexico 1981, Vol. 909 of Lecture Notes in Mathematics, Springer-Verlag, Berlin, pp. 226-239.

  • Goldfarb, D. and Idnani, A. (1983). "A numerically stable dual method for solving strictly convex quadratic programs", Mathematical Programming 27: 1-33.

  • QuadProg (Quadratic Programming Routines), Berwin A Turlach,

Used Functions

qpgen2.f and >qpgen1.f (also named QP.solve.f) developped by Berwin A. Turlach according to the Goldfarb/Idnani algorithm


5.5.0 Fifth output argument info added for error information.
Scilab Enterprises
Copyright (c) 2011-2017 (Scilab Enterprises)
Copyright (c) 1989-2012 (INRIA)
Copyright (c) 1989-2007 (ENPC)
with contributors
Last updated:
Thu Oct 02 13:54:34 CEST 2014