Skip to Main Content
Download | Contact Us | Index |
Requirements
Download ECCE
Install ECCE
Release Notes

ECCE Version 1.4 Release Notes

NOTE : All existing ECCE users upgrading to V1.4 are required to perform the action described in the first item of the "What's new in V1.4" section.

The intent of this page is to provide information specific to the Version 1.4 release of ECCE. This includes a brief description of new features, known bugs, and recently fixed items. You should also see the ECCE Version 1.3 Release Notes if you haven't already.

What's new?      Known bugs      What's fixed?      Gaussian 94 Registration Release Notes

What's new?

  1. Preference files have been renamed.
  2. Time limit within which jobs must start has been changed.
  3. Control over which version of a chemistry code is run has been added.
  4. Paths no longer include the ECCE version.
  5. Deletion of databases.
  6. Dialog for updating database server password.
  7. Depth cueing support added for visualization.
  8. The submit script timeout has been increased.
  9. User/database/calculation path is appended to the scratch directory specified in the Launcher.
  10. Launcher displays jobs that have already been submitted.
  11. install_ecce has been modified to create a generic link if a "Production" version is being installed.
  12. Web based support form replaces Motif support tool.
  13. Input file generation is now proactive about checking available disk space and write permission.
  14. ECCE tool command line names have been changed.
  15. New ECCE logo in Gateway.
  16. Calculation Editor "splash" screen eliminated.
  17. Blinking feedback message text areas for warnings and errors.
  18. Better recognition of basis set name when importing output files.
  19. Setting chemical system for calculations in the Calculation Manager no longer overrides theory spin restriction.

Known bugs

  1. If you submit a job to mpp1 and it takes longer than three days to start running, ECCE will consider the job to have failed due to a timeout.
  2. Geometries not parsed for Gaussian-94 molecules without Standard orientation.
  3. Security hole with ToolTalk.
  4. Mulliken total charges for NWChem have incorrect sign.

What's fixed?

  1. Correctly handles end of calculation output in the midst of parsing a property.
  2. The polarization properties are now parsed more robustly.
  3. Obviation of the need to re-drop calculations in Job Launcher.
  4. Job Launcher parameters no longer being ignored.
  5. NWChem geometry optimization Stepper algorithm gradient convergence default corrected.
  6. Remote shells now get DCE credentials.
  7. Machine Browser remote shell now works.
  8. Calculation and scratch directories cannot be NFS mounted disks.
  9. Fonts used in GUIs are now correct for DFS users.
  10. Calculation Editor no longer locks up X server after selecting from the Quick Basis Set List before selecting chemical system.
  11. In the calculation viewer the code name and launch information have been improved.

What's new?     Top

  1. In order to improve consistency with application names and have more logical preference file names while getting rid of obsolete files, many have been renamed. Before running V1.4 of ECCE for the first time users need to run the update_prefs script to convert their existing preference files, including their configured machines, over to V1.4 names. This is done simply by typing update-perfs at the shell prompt.
  2. The time limit within which jobs must start (or else be considered failed) has been changed from one day to three days on LoadLeveler systems.
  3. When a job is launched via ECCE, the user can control which installation of the chemistry code is used. This is a feature for advanced users only since we cannot guarantee that ECCE and the version of the code chosen to run are compatible. To run an individual version of a code, use the environment variables ECCE_NWCHEM and ECCE_GAUSSIAN94. These should be set in the .mycshrc file to the full path of the code. For example setenv ECCE_NWCHEM /opt/local/nwchem. An exit or restart is not needed when the variables have been set or changed in the .mycshrc of the launch machine.
  4. In previous versions, user databases were found in a path that included a version such as /ECCE/V1.3/users/ecceuser. With V1.4 the paths will no longer include the ECCE version. The internal changes to databases should be transparent to users but this is something to be aware of when navigating the database selection tree. Shared reference databases (such as the basis set library) are still located under the versioned path.
  5. The database selection dialog now supports an option to delete the selected database.
  6. There is now a popup dialog that will prompt you for your database password if you have never entered a password to ECCE or if your password has changed. Previously, this information had to be updated manually in a preference file. A command line argument to the "ecce" startup script will always bring up the dialog allowing you to authenticate as a different user than previous ECCE sessions . The last successful login name and password is saved and will be used the next time ECCE is started provided the dialog is not requested with the command line argument. The argument is "-l" for "login" or, equivalently, "-a" for authenticate.
  7. A depth cueing option has been added to the Builder. This is essentially a fog effect where the fog increases linearly with the distance of the atoms. This technique provides some depth perspective and is especially useful for larger molecules.
  8. The submit script timeout has been increased from 60 to 120 seconds to provide for long submission times. In addition, users can set the submit time explicitly by using the environment variable ECCE_SUBMIT_TIMEOUT. ECCE must be exited and restarted for it to recognize a change in this variable.
  9. A user/database/calculation path is appended to the scratch directory specified in the Launcher. In V1.3, calculations use the same scratch directory and could write over files for other calculations, leading to failure. The calculation directories already have a unique path appended.
  10. The Launcher now displays jobs that have already been submitted, so launch parameters may be viewed for a previously run calculation. Imported calculations are not supported since they have no launch information.
  11. install_ecce has been overhauled for improved readability and it now creates a generic link if a "production" version is being installed so users won't need to modify their .myschrc file with each new ecce version. The new /.mycshrc logic for ECCE is:
    1. #setup to run ECCE for V1.4 or later
    2. if (-e /opt/local/ecce/scripts/runtime_setup)then
    3. source /opt/local/ecce/scripts/runtime_setup
    4. endif
    The generic link is /opt/local/ecce/scripts -> /opt/local/ecce/V1.4/scripts (or whatever the ECCE destination directories actually are.)
  12. The "Feedback . . ." help menu item in each of the ECCE tools now links to a web based form for submitting support requests. The first time the form is accessed the user must enter name, e-mail, and phone number which is saved for subsequent requests.
  13. Since most users keep their ECCE UNIX project tree directories under their home accounts which are usually on inadequately sized AFS or DFS volumes for EMSL users, more effort was made to catch when input file generation was likely to fail. To compound the problem users lose write permissions to these volumes when authentication expires. Before input file generation is done such as for a Calculation Editor final edit or when a job is launched, an attempt is made to create a 2 Megabyte file in the same directory as the input file. This is immediately removed prior to generating an input file so no disk space is actually lost. If this test file can not be generated then an error is displayed to the user indicating insufficient disk space or lack of write permission. The 2 Megabyte file size was a fairly arbitrary choice based on chances of success for both input file generation and then later copying output files back to the UNIX project tree.
  14. To make meaningful and consistent names several ECCE tool executables have been renamed. For users that start ECCE by typing "ecce" at the shell prompt this will not make a difference. It is only of interest to those who source the ecce_env script and start individual tools at the shell prompt.
  15. The V1.3 EMSL logo has been integrated with the ECCE "wave" animated logo on the gateway. This shrinks the gateway back to its pre-V1.3 size.
  16. To improve startup time, the Calculation Editor startup "splash" screen animation has been removed. The Calculation Editor main window is now displayed immediately consistent with other tools although it takes a few seconds for initialization to complete while the busy watch cursor is displayed.
  17. The background of the feedback message text areas at the bottom of the main window in most ECCE tools now blinks momentarily when a warning or error is reported. This is done in order to draw the user's attention since it had been observed that many users ignored warnings and errors previously even when accompanied by a beep. Warnings are distinguished from errors by the color of background--pale yellow flashes for warnings and red for errors.
  18. To assign a name to an imported gaussian calculation's basis set, a user can place the basis set name within the route card after the '/'. If a name is not specified, the keyword 'GEN' should be present in it's place. (This feature is not supported for checkpoint files where the keyword 'chkbas' is within the route card.)
  19. Although it is a little known feature of ECCE, it is possible to set the chemical system subject of a calculation or even all calculations under a project within the Calculation Manager by dropping the chemical system (middle mouse button over the Builder drag icon) over the desired calculation or project. Prior to V1.4 this would set the theory spin restriction from unrestricted to restricted if the new chemical system had an even number of electrons. With V1.4 an unrestricted theory selection is left that way. This does not apply when setting the chemical system within the Calculation Editor. Here ECCE behaves as it previously did by overriding the spin restriction to restricted if the number of electrons changes from odd to even. This is seen as the most sensible choice to set as default. It is always possible for the user to override the automatic setting and force an unrestricted spin restriction. The differentiation between the Calculation Manager and Calculation Editor was made to allow setting up entire sets of calculations such as for testing or benchmarking and to be guaranteed that the theories remained the same when the chemical system changed. It has always been the case in all tools that a restricted spin theory would be overriden when the number of electrons changed from even to odd.

Known bugs      Top

  1. If you submit a job to mpp1 and it takes longer than three days to start running, ECCE will consider the job to have failed due to a timeout.
  2. For any Gaussian-94 calculations, ECCE will not parse geometry steps if they are not put into the output file in Standard orientation. This will happen in certain cases such as symmetric tops and atomic systems where only the Input orientation is output by Gaussian. For the same reasons, these calculations cannot currently be imported. If you encounter a case like this, please contact ecce-support@emsl.pnl.gov for assistance with importing the results.
  3. There is a potentially serious security hole in the current version of ToolTalk used by ECCE. A technical description follows:
    • Due to an implementation fault in ToolTalk's rpc.ttdbserverd, it is possible for a remote client to formulate an RPC message that will cause the server to overflow an automatic variable on the stack. By overwriting activation records stored on the stack, it is possible to force a transfer of control into arbitrary instructions provided by the attacker in the RPC message, and thus gain total control of the server process. This implementation problem can only be resolved completely by applying patches to or replacing affected ToolTalk software.
    Sun plans to release patches the week of 9/7/98 - 9/11/98 for SunOS 5.6, 5.5.1,amd 5.5. The Security Labs team at Network Associates has confirmed that SGI IRIX 6.3 is vulnerable to this attack. SGI's security team has been contacted and informed of the vulnerability. No repair information has been made available from Silicon Graphics regarding this problem.
  4. Mulliken total charges for NWChem have incorrect sign.
    The Mulliken charge values supplied by NWChem to ecce have the wrong sign. This shows up in the the Mulliken charge interface of the Calculation Viewer. To get the correct values, multiply each total charge by -1.0. This will be fixed with the next release of NWChem.

What's fixed?     Top

  1. If the end of calculation output is reached in the midst of parsing a property, it is now treated as an error.
  2. The polarization runtype properties are now parsed more robustly.
  3. The Launcher correctly updates if the current calculation is reset for restart, obviating the need to re-drop the calculation.
  4. It was previously possible in any numeric or text entry field to enter a value that was not immediately validated. Validation in ECCE is triggered by changing a value and then moving the mouse pointer outside the field. It was possible to get "input focus" in the field so that it accepts keyboard input and then move the mouse pointer outside the field while retaining input focus. If the user then changed the value and immediately hits the "Launch" button in the Job Launcher the last change wasn't recognized. ECCE now forces any keyboard input to be done while the mouse pointer is within the field. This enables ECCE to recognize when to perform validation without resorting to the Motif default of requiring the user to hit a carriage return every time they make a change.
  5. NWChem geometry optimization Stepper algorithm gradient convergence default corrected.
  6. Previously remote xterm shells did not get DCE credentials due to a particularly devious bug in the secure shell communications. This was also known to cause large scale havoc on the remote machine blocking further shells from connecting.
  7. In Version 1.3 selecting the "Open Shell..." menu item in the Machine Browser resulted in nothing happening. This was a problem that wasn't noticed when the remote communications were first migrated to ssh/rsh.
  8. In addition to restricting remote calculation and scratch directories from being AFS and DFS volumes they are also not allowed to be NFS disks. The reason for this are timing problems when writing to NFS files from different nodes of a parallel machine. The order of output in a file once it is completely written can actually differ from the order it is received by a process polling for output to the file as it is being written. Only locally mounted disks are immune from this problem. ECCE itself checks that a disk is not AFS, DFS or NFS as it does machine configuration and launch validation. As a practical point the write access time to NFS disks, although much better than AFS and DFS, is still unreasonable for any real world computation.
  9. Due to an EMSL CANS policy change in maintaining a common distribution of X Window software between AFS and DFS the fonts were incorrect for DFS users. Instead of picking up the Adobe version of Helvetica fonts, Linotype, was being picked up. ECCE has been changed to explicitly use Adobe Helvetica fonts.
  10. In previous versions it was possible to bring up the Quick Basis Set List (right mouse over "Basis Set..." button) on the Calculation Editor main window even though the "Basis Set..." button was disabled prior to selecting a chemical system. Actually making a selection from the list would lock the X server display completely. It is no longer possible to make a selection from the list until the chemical system is specified.
  11. In the Calculation Viewer, the code name and launch information (such as machine, queue...) now have their own entries in the outline. Previously this information was located under Setup Parameters. In addition, the Setup Parameters window now displays only non-default values.

Gaussian 94 Registration Release Notes      Top

The entire registration of Gaussian 94 within ECCE has been revamped. This includes options and defaults in the Calculation Editor, generation of the input file, and parsing of output properties that are viewable in the Calculation Viewer and/or Browser. See below for more details. To test the registration, a test suite of approximately 700 calculations was developed and run. This test suite included runs for every combination of theory and runtype that can be selected as well as runs for every detail that can be set in the Calculation Editor. Calculations were run for singlet, doublet, and triplet systems. Within the test suite, there were a number of runs which caused Gaussian to fail but these appear to be due to the nature of the calulation itself and not because of mistakes in setting up the input file or because an unsupported theory/runtype combination was allowed to run. Users that discover combinations that will not run under any circumstances or who know of theory/runtype combinations that will work, at least in some circumstances, that have been disabled within ECCE, are encouraged to contact the ECCE development team.

Input File Generation

  1. ECCE generated input files now contain minimal information. Redundant or irrelevant keywords have been eliminated from the input file as much as is possible.
  2. ECCE always requests the generation of MO's through the inclusion of the Punch=(MO) and Pop=(Full) keywords in the input file. This causes Gaussian 94 to produce the MO's and orbital occupations necessary to support the MO feature of the Calculation Viewer.
  3. The Density keyword is used as needed to cause Gaussian to generate properties using the highest applicable theory (i.e. an MP2 calculation will use the MP2 density instead of the SCF density).
  4. The disk options (direct/in core/to disk) for the SCF calculation have been fixed so that these directives actually get written out to the input file if the non-default values are selected. The disk options for the MP2 calculation have also been fixed so that the appropriate directives are written into the input file if the non-default values are selected.
  5. The DFT input has been fixed so that if the VWN 5 (local) correlation functional is selected, the right option is put into the input deck.

Details Windows and Defaults

  1. The Delta Density defaults under the SCF convergence window have been fixed so that the correct default values are shown in the fields. The manual is not very clear on what some of these defaults are, and the values have been extracted from the code output in these cases. The input field has also been fixed so that it only accepts powers of 1/10.
  2. The Delta Energy defaults for the Coupled Cluster Configuration Interaction calculations have been fixed to match the code ouput. According to the manual, the default values are 10**-7 for single point energy calculations and 10**-8 for everything else, but according to the code output, the defaults are 10**-8 for all calculations. The input field has also been fixed so that it only accepts powers of 1/10.
  3. The defaults for the maximum number of iterations have been fixed for the SCF calculations and the Coupled Cluster calculations. The default value for the number of SCF iterations now changes from 64 to 512 when the convergence algorithm is set to quadratic. The default maximum number of iterations in the Coupled Cluster calculations has been set to 50, which matches the Gaussian 94 reference manual value.
  4. The options of the SCF convergence algorithms have been fixed to match the availability and default values for the different theories.

Runtype Options

Most of the major problems with the Gaussian 94 registration are associated with the different runtypes. The gradients, geometry optimization, and polarizability have been fixed so that the available options are enabled if selected (not generally true for V1.3) and so that the availability and defaults correspond to the Gaussian 94 manual. The Magnetic/NMR runtype appears to be OK and no changes were made to this runtype. The Vibration calculation fields have not been modified yet although there appear to be several problems in terms of correctly describing algorithm availability and defaults. If the Vibration runtype is selected without changing any of the defaults, the calculation will probably run, but the Gaussian 94 output file should be checked to find out what algorithm was actually used. The major changes to the runtype windows are described below.

  1. The Gradient runtype has been fixed so that the finite difference option can be selected for those theories where it is not the default and the finite difference step size has been fixed so that a directive is included in the input deck if a value other than the default is selected. At present, the theories in Gaussian for which ECCE is assuming that analytic gradients are available are:
    • all SCF
    • all DFT
    • MP2, MP3, MP4(SDQ)
    • CI
    • QCISD
    • CCD
    • all Semi-Empirical
    The Gaussian 94 manual is not very clear on this point so if anyone knows differently, let the ECCE development team know as soon as possible. There is also a bug in the Gaussian 94 code that causes it to fail if a gradient calculation has been selected for the CCSD theory type (apparently, this bug has no effect on geometry optimizations). The gradient calculation has been disabled for these theory types. The finite difference calculation of the gradient is not available for semiempirical theories and has been removed from the options list. Finally, the gradient calculation fails completely for the restricted open-shell semi-empirical theories and the gradient runtype has been removed as an option for these theories, along with the geometry optimization and vibrational frequency runtypes.
  2. The optimizations have been modified so that the CalcFC keyword is included in the input if a transition state search is specified. However, the transition state searches do not appear to work very well under any circumstances, possibly because ECCE does not use the Z-matrix representation of the geometry.
  3. The Polarization runtype has been fixed so that the options in the details list correctly match the availability of different algorithms for each theory and the correct default algorithms are given. Again, the Gaussian 94 manual is not clear as to exactly what algorithms for calculating the polarizability are available for each theory, so if anyone finds an error in this regard, please let the ECCE development team know as soon as possible. The following is the current breakdown for the availability of different polarizability calculations:
    • Analytic polarizabilities and hyperpolarizabilities:
      • RHF, UHF
      • DFT
      • MP2
      Analytic first derivatives of energy with respect to electric field, numerical polarizabilities and hyperpolarizabilities:
      • ROHF
      • MP3, MP4(SDQ)
      • CCD
      • QCISD
      • CIS, CISD
    The remaining theories can calculate the polarizability as the numerical second derivative of the energy with respect to the electric field. Polarizability calculations are not available at all for any of the semi-empirical theories. There is a bug in Gaussian 94 that causes the code to generate an input error if the user tries to specify a finite difference step size in the case of numerical evaluation of the gradients. For this reason the finite difference step size value option has been turned off.
  4. Running a test suite of calculations on Gaussian 94 revealed a number of theory/runtype calculations that do not appear to work. These are currently believed to be due either to bugs in Gaussian 94 or because these theory/runtype combinations are not supported. The theory/runtype combinations that have been disabled are:
    • MP4 : Gradients, Optimizations, Opt. + Vib
    • MP5 : Everything but Energy
    • ROHF : Magnetic and Polarizability
    • ROMP2 : Everything but Energy and Vibration
    • QCISD(T): Gradients, Optimizations, Opt. + Vib
    • CCSD : Gradients
    • CCSD(T) : Gradients
    • Semi-Empirical Restricted Open-shell : Everything but Energy

Output Parsing

  1. The S**2 data is now parsed and available in the Calculation Viewer/Browser.
  2. Orbital occupations are from the Pop=(Full) option instead of the Pop=(NOAB) option.
  3. The highest available energy is labeled the Total energy and is defined for every calculation. Intermediate energy values are parsed only for optimization runs. For calculations involving finite difference evaluations of derivatives of the energy, a large number of converged energy values are generated but at present there is no way to present these in a meaningful way.
  4. The calculated gradient is included in the Calculation Viewer if the Gradient runtype was selected. This was not previously the case.

Remaining Problems

  1. As mentioned above, the Vibration calculation only performs the default calculation. The default values listed in the Runtype Details window of the Calculation Editor may not accurately reflect what Gaussian 94 actually does. Resetting any of the buttons in the Details window has no effect on the input file. The user should check the output file if a Vibration calculation is performed to determine what algorithms were actually used to obtain the frequencies.
  2. ECCE does not correctly recognize when many Gaussian 94 calculations are unconverged and lists these as failed calculations instead. The user should check the output files on failed calculations to see if they are merely unconverged.