Frequently answered questions about Dalton

Reporting problems

Please discuss problems on dalton-users. In particular, do not send requests for help to dalton-admin - bug fixes and patches are still welcome, though. An archive of dalton-users mailing list is available.

Dalton 2.0 manual

Hypertext Dalton 2.0 manual (with internal links and all) is available in PDF form.

Compilation and setup

Intel fortran compiler issues

Intel fortran compilers generate admiteddly fastest code but such a complex code as dalton exposes many bugs in the compiler. Always run the test suite! We know of following issues:

• several 7.x compilers miscompile coupled-cluster code.
• some 8.1.x compilers in particular 8.1.025 and 8.1.26 have problems with character comparisons resulting in failed keyword recognition leading to following abnormal termination:

  1043: --> ERROR (OPENDX) INVALID STATUS KEYWORD: NEW
  1063:  --- SEVERE ERROR, PROGRAM WILL BE ABORTED ---
  1067: Reason: ERROR (OPENDX) INVALID STATUS KEYWORD
  

  You can find out whether your compiler has still this bug by running following program:

        PROGRAM x
        CHARACTER*3 s
        s = 'NEW'
        open(42,STATUS='NEW')
        if(s(1:2).NE.'NE') THEN
           print *, "Bug in the compiler"
        ELSE
           print *, "this compiler has no ifort-8.1.025 bug"
        END IF
        CLOSE(42,STATUS='DELETE')
        END
  

Version	x86	Itanium
8.1.024	OK	Fail
8.1.036		OK
9.0.021	Fail	Fail
9.0.027	OK

GNU compilers issues

• g77-3.3.3 (Fedora Core 2 and SuSE) on Opteron (64-bit version) has problems with walk_vibave2 test and all SO-related tests that claim no MOLECULE label has been found in a temporary file:

   622: ---
  SEVERE ERROR, PROGRAM WILL BE ABORTED --- 626: Reason: ERROR (MOLLAB)
  MOLECULE label not found on file 

• g77-3.4.2 (Fedora Core 3) miscompiles DLAMC3 routine from gp_zlapack.F file leading to apparent hang in rsp_zfs test. It is recommended to link against a working lapack library or add -ffloat-store compilation option. See also relevant Bugzilla entries: 138683 and in particular 138447 for possible workarounds.

Dalton on OSX 10.4 (Tiger)

Currently, linking Dalton on OSX 10.4 (Tiger) fails. Two suggestions were made. One workaround is to use the older 3.3 compiler. It can be selected as follows:

 sudo gcc_select 3.3

The other - probably better - alternative is to manually add -lSystemStubs to LIBS variable in Makefile.config.

Portland Compilers

We have received an yet unconfirmed report that PGI 5.2.4 generates incorrect code that gives rise to wrong results in response calculations (iterative solver may not converge).

Runtime problems

..limited to 255 orbitals

You get message similar to:

TRACTL_1-ERROR: TRACTL_1 is limited to 255 orbitals

The solution is to run the calculation in direct mode, as enabled with .DIRECT keyword:

**DALTON INPUT
.DIRECT
...

Sometimes presorting of integrals helps. It can be enabled with .PRESORT keyword:

**DALTON
.PRESORT
...

| Dalton home page | This page written by Pawel Salek and hosted by Theoretical Chemistry, KTH.