FAQ

  • How can I contribute?

QMCPACK is fully open source. Contributions of new methods, tests, documentation etc. are all welcomed. Please post on the google groups site https://groups.google.com/forum/#!forum/qmcpack or discuss with an existing developer. Comments for suggested improvements are highly valued.

  • How do I install QMCPACK?

See the instructions in the Getting Started guide.

  • How can I get help?

After verifying that your problem is repeatable, please post a question giving full details on the google groups site https://groups.google.com/forum/#!forum/qmcpack

  • How often is a new version of QMCPACK released?

We are aiming for a release every six months, or when a significant new feature becomes available, whichever is more frequent.

  • I can’t install QMCPACK on new system XYZ.

We recommend to install QMCPACK in standard Linux environment before attempting to install on a new system. This will provide a reference to see which part of the configuration or compilation has failed.

In general problems occur when incompatible combinations of compilers and parallel libraries are picked up by CMake. In this case specify the compilers directly, either on the CMake line or via a toolchain file. See the Getting Started guide.

If you still have problems, please post a message on the google groups site  https://groups.google.com/forum/#!forum/qmcpack.

  • I can’t find any documentation for feature XYZ

While all the key QMCPACK features are documented, many of the less used features are currently undocumented. We are working to correct this and are prioritizing in order of perceived popularity. We are aiming to have documentation and an example for all features. Please post your request to the google group! We can probably get you a sample working input.

  • How much computer time will I need for my problem?

This is a question that can only be answered through experimentation. As a reference, QMC is generally several orders of magnitude more expensive than conventional cubic scaling density functional methods when run to typical statistical accuracies (few 0.1 eV). The Monte Carlo nature of QMC means that the statistical error bar can be increased to reduce the computational cost. However, obtaining a physical/chemical result to a given degree of confidence may require e.g. extensive costly testing of different wavefunctions.

It is not untypical for “research grade” solid-state investigations to require millions of computer hours, while molecular calculations can be much cheaper.