FastROCS TK 1.6.0

New features

  • The performance of FastROCS when loading a database into memory has been significantly improved when the appropriate caching has been done to the input OEB file. OEPrepareFastROCSMol function has been added to allow preparing a molecule to be written to OEB so that it can be read very quickly by OEShapeDatabase::Open. It is highly recommended to pre-process input OEB files with this function to achieve faster load times. This works especially well in combination with OEPRECompress, resulting in uncompressed OEB files that are comparable to their gzip-compressed counterparts, even with extra cached information inside them.

    Warning

    OEPrepareFastROCSMol strips out information that OMEGA generates (i.e., hydrogen coordinates and conformer energies). Care must be taken when using these OEB files for alternative use cases. Please refer to the OEPrepareFastROCSMol documentation for a full accounting of how the molecules are changed.

  • OEFastROCSCacheStuff was never officially supported and has been replaced by the OEPrepareFastROCSMol function. It now throws a deprecation warning and will be removed in a future release.

  • FastROCS TK should be binary compatible with more NVidia driver versions as of this release. Previously, FastROCS TK could only target a single NVidia driver version. Now, several can be targeted in a single toolkit build. This version of FastROCS TK will work on the following NVidia driver versions: 340.x, 346.x, and 352.x. This should significantly reduce FastROCS TK installation headaches.

  • OEShapeDatabase::SetNumOpenThreads method has been added to control how many CPU operating-system threads are created when calling OEShapeDatabase::Open on an OEMolDatabase object.

  • All OEShapeDatabase instances now share the same GPU context. Previously, creating many OEShapeDatabase instances would result in many operating system threads created per instance. This would often hit resource limits around the number of threads allowed in a single process.

Minor bug fixes

Python-specific changes

  • All FastROCS TK Python examples now use /usr/bin/env python to determine the appropriate Python interpreter to use. Previously, some examples were erroneously hardcoded to use /usr/bin/python.

  • karma.py shape clustering example had a minor bug that affected how invalid file names were handled. This has been fixed.

Documentation changes

  • The FastROCS TK Installation section is now more general and defers to the NVidia website for installation instructions. A list of supported NVidia drivers for FastROCS TK can also be found in the Installation section.