Re: [Bug-gsl] gsl_blas_dgemv return a 0 vector when using a vector twice

[Alejandro Cruz Marcelo]

Hello Brian
Sorry for my late response.
When using the funciton gsl_blas_dgemv to perform the operation
x = Ax
it returns x equal to 0. I guess we are not supposed to use the same 
vector but that warning is not in the manual.
However, the operation
x = Ax + x
it is performed properly.
Here is some code to illustrate this.
Thank you,
Alejandro


//   Start code
#include <stdio.h>
#include <math.h>
#include <gsl/gsl_vector.h>
#include <gsl/gsl_matrix.h>
#include <gsl/gsl_blas.h>

int main()
{

gsl_vector * x = gsl_vector_calloc(2);
gsl_matrix * A = gsl_matrix_calloc(2,2);

gsl_matrix_set_identity(A);
gsl_vector_set(x,0,2);
gsl_vector_set(x,1,2);

gsl_blas_dgemv(CblasNoTrans,
                1.0, A, x,
                0.0, x);
printf( "The operation performed is x = (1.0)*A*x + (0.0)*x  \n ");
printf("Output should be the vector [2,2]^t ");
printf("but instead we get: ");
printf("%f \n ", gsl_vector_get(x,0));
printf("%f  ", gsl_vector_get(x,1)); 
 system("PAUSE");   

gsl_vector_set(x,0,2);
gsl_vector_set(x,1,2);

gsl_blas_dgemv(CblasNoTrans,
                1.0, A, x,
                1.0, x);
printf( "But if the operation performed is x = (1.0)*A*x + (1.0)*x  \n ");
printf("Output should be the vector [4,4]^t ");
printf("and that is what we get: ");
printf("%f \n ", gsl_vector_get(x,0));
printf("%f  ", gsl_vector_get(x,1));   

 system("PAUSE");     
 return 0;
}

// The end



Brian Gough wrote:
> At Fri, 29 Aug 2008 09:44:30 -0500,
> Alejandro Cruz-Marcelo wrote:
>   
> > gsl_blas_dgemv(CblasNoTrans,
> >                     1.0, A, y,
> >                     1.0, y)   
> > then y becomes a vector of 0's and no error is produced, even though the 
> > operation A*y + y makes perfect sense. I believe that in the 
> > documentation there is no warning about using the same vector twice in 
> > the function.
> >     
>
> Please can you send an example program that demonstrates the problem,
> thanks.
>
>