(Updated Jan 9, 2019) The current stable version is v.6.3, released on July 10, 2018. A bugfix release is available as a git branch "qe-6.3-backports" on github and gitlab. Releases are typically packaged every 3-4 months.
- The next version will be released on February 28, 2019
The following is an incomplete list of major projects related to Quantum ESPRESSO development, with even more incomplete information on their status and future directions.
- New GPU-enabled code
- New diagonalization algorithms with reduced communications
- Automatic estimate of parallelization parameters
- Refactoring of existing machinery for XC functionals, more general libxc interface
- More robust self-consistency ior global minimization algorithm, or joint optimization of density and atomic positions
- Speeding up phonon calculations for large systems
- DFT+U+V and refactoring of DFT+U codes
- Electron-phonon coefficients with DFT+U
Things you can do to help QE
The following list includes simple projects, mostly computational rather than physical, suitable for an internship or for summer students.
- Improve performances: analyse performances of FFTs and linear algebra operations on some relevant architecture (e.g. the more recent intel cpus), locate and remove bottlenecks; analyze OpenMP parallelization and improve it
- Improve build mechanism: clean up and simplify the current "configure" script, make it more robust and able to recognize more cases and more external libraries
- Improve packaging: help with existing Debian packages (available in DebiChem), produce binary packages in other formats: RPM, Mac OS-X, Windows executables
- Improve interfacing with scripting languages (or even non-scripting languages): write a QE library that is easily callable from python or other languages
- Improve pseudopotential format: implement the new xml format, produce conversion tools to such format for as many other formats as possible
- Tools for verification: extend automated tests to cover more QE packages and more kinds of calculations; make it less sensitive to small numerical errors