FastROCS TK 1.5.1¶
OEFastROCSHistogram has been added to retrieve FastROCS TK’s score distribution for the entire database, not just the hits, in real-time during a FastROCS TK query. The ShapeDatabaseServer.py example was extended with the
OEFastROCS::QueryHistogramfunction, as well as the ShapeDatabaseProxy.py. The ShapeDatabaseClientHistogram.py is an example of a client program to retrieve the histogram from the server in real time and print it to the terminal.
Major bug fixes¶
ShapeDatabaseOEThrowSetLevel.py will now work to effectively turn on additional performance debugging information. The additional operating system threads started by the OEShapeDatabase will now inherit the
OEErrorLevelbeing used by the thread to execute the query through
FastROCS TK will now be compatible with NVidia compute architecture 2.0 cards. FastROCS TK beta in 2015.Jun.5 only worked on compute capability 3.x cards. The 2015.Jun.6 release was a bug fix release to fix this oversight. FastROCS TK in 2015.Oct should now work with any compute capability 2.0 card (the Fermi architecture) or greater.
Minor bug fixes¶
ShapeDatabaseServer.py will now release its port number upon termination more quickly. Previously, the server would sometimes fail because when a new process was started the port number was already reserved by the previous process.
The Installation section now recommends installing the NVidia driver with the
--disable-nouveauflag to ensure that the open source
nouveaudisplay driver kernel module is turned off. The NVidia CUDA documentation is a much better source of this installation information: http://docs.nvidia.com/cuda/cuda-getting-started-guide-for-linux/#runfile-nouveau