[CP2K:3071] Matrix Operations

[Urban Borštnik]

Hello,

to be able to release() a matrix, you should first create() it in some
manner.  The cp_dbcsr_init() call acts only as something similar to a
nullify().  So between lines 7 & 8 you should call some thing like

cp_dbcsr_create (exx_kernel(ispin), template=exx_energy, ...)

(or use another routine to populate the matrix, such as cp_dbcsr_copy or
similar.

Cheers,
Urban


On Sun, 2011-01-23 at 09:19 -0800, Andrey wrote:
> Dear cp2k developers,
> 
> I have the code, which works without errors:
> 1)
>    TYPE(cp_dbcsr_p_type), DIMENSION(:),  POINTER  :: exx_kernel
>    TYPE (qs_energy_type), POINTER           :: exx_energy
> 
> .....
> 1            NULLIFY(exx_energy, exx_kernel)
> 2             CALL allocate_qs_energy(exx_energy)
> 3
> 4            CALL
> cp_dbcsr_allocate_matrix_set(exx_kernel,nspins,error=error)
> 5             DO ispin=1,nspins
> 6                ALLOCATE(exx_kernel(ispin)%matrix)
> 7                 CALL cp_dbcsr_init(exx_kernel(ispin)
> %matrix,error=error)
> 8            END DO
> 9
> 10          CALL deallocate_qs_energy(exx_energy)
> ......
> 
> 2)  When CALL cp_dbcsr_deallocate_matrix_set(exx_kernel,error=error)
> added,
>               .....
> 9                CALL
> cp_dbcsr_deallocate_matrix_set(exx_kernel,error=error)
> 10              CALL deallocate_qs_energy(exx_energy) ASSERTION
> FAILED: F
> 
> program finishes well, but noncritical error messages appear.
> 
>  dbcsr_error_print_stack
>  error%msg=
>  Matrix not
> initialized
> 
>  error%routine=
>  
> dbcsr_release
> 
>  error%etype=           1
>  error%level=           1
>  error%line=         413
>  error%stack%level=           0
>  dbcsr_error_stack_size=          10
> 
> 
> 
> 3)  If CALL cp_dbcsr_set(exx_kernel(ispin)%matrix,0.0_dp,error=error)
> is set (between lines 7 and 8)
> then program does not pass it and crashes at this point.
> 
>  forrtl: severe (174): SIGSEGV, segmentation fault occurred
> 
> Could you, please, help me to understand what is wrong in my code.
>