IMSL linear operators

I’m having trouble using the linear operators defined in linear_operators.mod from the IMSL library. I’m working with PVF 11.0 on a Win7 OS. My project contains the file linear_operators.mod in the Include Files folder and, in Project Properties, Use IMSL is set to Yes. To try and get a handle on the problem, I’ve been experimenting with the following code, modified from the IMSL example subroutine operator_ex01.f90.

program OperatorEx01

	use linear_operators
	implicit none

	integer :: i
	real(kind=4), dimension(4,4) :: A, b, x
	real(kind=4), dimension(4) :: c

	A=rand(A); b=rand(b); c=rand(c)

	x=A .ix. b	!test linear operator .ix.

	A=eye(4)	  !test linear function eye()
	do i=1,4
		print *, A(i,1:4)
	end do
	x=diag(c)	 !test linear function diag()
	do i=1,4
		print *, x(i,1:4)
	end do

	end program OperatorEx01

When I try to compile it, I get the following error messages:

unresolved external symbol s_lin_sol_gen_ referenced in function operation_ix_zimzvp_
unresolved external symbol s_lin_sol_lsq_ referenced in function operation_ix_zimzvp_
unresolved external symbol d_lin_sol_gen_ referenced in function operation_ix_dimsmp_
unresolved external symbol d_lin_sol_lsq_ referenced in function operation_ix_dimsm_
unresolved external symbol c_lin_sol_gen_ referenced in function operation_ix_cimdmp_
unresolved external symbol c_lin_sol_lsq_ referenced in function operation_ix_cimdmp_
unresolved external symbol z_lin_sol_gen_ referenced in function operation_ix_zimcmp_
unresolved external symbol z_lin_sol_lsq_ referenced in function operation_ix_zimcmp_

I get comparable error messages if I try to use any of the other linear operators (e.g., .i., .t., .ix., …). Because all of the errors are tied to the IMSL modules lin_sol_gen_int.mod and lin_sol_lsq_int.mod, I tried adding these explicitly to the project, but this did not alter the result. If I comment out the line 'x=A .ix. b", the program compiles and runs fine. Note that the functions eye and diag both require linear_operators.mod, and that they compile and run correctly. So, I am able to use the functions defined in linear_operators.mod, but not the linear operators themselves. My question is, what should I be doing differently in order to use the linear operators? Many thanks for whatever insights you can provide.

I’ve found a solution to the problem. Rather than simply setting Use IMSL to Yes under Project Properties, I manually specified the paths for the compiler/linker to use for the IMSL routines (as per section 4 of the IMSL file Readme_windows.txt). After doing so, my programs that use lin_sol_gen_int.mod and the different linear operators compile and run without error. I am surprised at this result, however. Since the option to simply set Use IMSL to Yes was added, I’ve found this kind of manual path specification unnecessary. Am I missing something, or does this suggest a problem with the Use IMSL compiler option?


Thank you for reporting to us. This is certainly a bug at link time. libimslscalar.lib is missing at link line. We will get it fixed in 11.2 release.

To keep track of the bug, TPR 17609 has been filed.