GSOC Ideas

Preamble: EasyBuild is an open source tool, which is very handy for building Scientific Computing software; as such, it is now becoming popular among HPC sites.

Here is a list of ideas that are desired to appear in next versions of EasyBuild, or worthy experiments: (their declared importance varies according to who you ask ;-)

Software Mixer

• Software Mixer; ie. provide pathways for software among different repositories/projects:
• .eb easyconfigs for EasyBuild
• pkgsrc Makefiles
• ports Makefiles
• macports
• portage
• (...you name it...)
• rpms/srpms/specfiles
• deb

CUDF

• Use a "formal" language to describe package info & dependencies, eg. CUDF syntax;
• http://blog.mancoosi.org
• http://www.mancoosi.org/papers/cbse11.pdf "MPM: a modular package manager"
• http://arxiv.org/abs/1109.0456 "Aligning component upgrades"
• ANSWER: may consider; TBD/postponed.
• Provide some smart way to handle osdependencies like "tcsh" which may have no meaning on eg. MacOSX # a particularly important case are the boost-devel vs boost-dev differences (rpm vs deb conventions)
• and for now allow escaping with --ignore-osdependencies ; See #102; FG update: this now is handled with --strict=ignore

Build easyconfigs for missing scientific software

• Plenty of work to be done in this area, related open issues are:
• https://github.com/hpcugent/easybuild-easyblocks/issues/86 # HPC-BC
• https://github.com/hpcugent/easybuild-easyblocks/issues/82 # PRACE
• https://github.com/hpcugent/easybuild-easyblocks/issues/39 # CBench
• https://github.com/hpcugent/easybuild-easyblocks/issues/123 # Eclipse with PTP
• http://hpcbios.readthedocs.org/en/latest/HPCBIOS_2012-99.html # GPUs
• .. or any other from the many at http://hpcbios.readthedocs.org/en/latest/

CDASH

• This is the application of ctest/dash while building software; sounds trivial, but there are >500 modules that now get built automatically, so there is some work here.
• dependencies make it interesting.
• Next step would be to combine it with some benchmarking graphs

EasyDoc

• Can we automate the generation of documentation for Scientific Software?
• Example: http://eniac.cyi.ac.cy/display/UserDoc/GROMACS
• The idea is about integrating EasyBuild with 3rd party wikis/docs/sites, preferably using markup-intelligent tools like pandoc.
• Caveat: we need to dynamically import information from the running system(s)!

env-modules cmd extension

• Extend one of the "environment-modules" implementations (tcl/c/lmod)
• supply features: hierarchies, categories, keywords, whatis, module -d
• extreme thought: yes, this is an extreme thought: consider module search/install extensions!

Extend auto-generated module namespace

• Give freedom to define the namespace format (eg. original Package names OR the lower case version)
• allow to export namespace in both UpperCase and LowerCase conventions # a post-install hook? TBD
• this is tracked here: https://github.com/hpcugent/easybuild-framework/issues/173
• Support more configurable formats of version strings (check below, how other HPC sites do it)
• nested levels (ie. split version string in sub sections)
• provide hooks for gnu/gcc/intel/pgi etc strings
• IDEA: For now, supply a "custom_namespace_function" variable, to define eg. modulename.lower();

Integrate hashdist/EasyBuild/HPCBIOS

• all of the above attack the reproducibility aspects from different, orthogonal angles
• nobody has tried yet to synthesize all idea into one unit... TBD!