Known issues in 0.8.1
***Error! Exclusion list full for particle ... when adding more than 4 bonds on a single particle
Due to a limitation in the way non-bonded interactions exclude bonded particle pairs, there can be a maximum of up to 4 bonds on each particle. Release 0.9.0 will finally solve this issue - ticket:86
***Error! unspecified launch failure after .../BinnedNeighborListGPU.cc:628 when running on compute 1.1 devices
- Note: The line number specified may be different from 628 if you are not running with the --gpu_error_checking option enabled
- Typically, this error occurs after several hours of run time.
- Again, this problem only occurs on compute 1.1 devices.
***Error! unspecified launch failure after ...... when running on compute 1.3 devices
There appear to be some stability problems with HOOMD 0.8.1 on compute 1.3 devices. This issue is already fixed in svn. To work around the problem, execute the command lj.cpp_force.setBlockSize(352) after specifying lj = pair.lj(...).
Known issues in 0.8.0
***Error! Exclusion list full for particle ... when adding more than 4 bonds on a single particle
Due to a limitation in the way non-bonded interactions exclude bonded particle pairs, there can be a maximum of up to 4 bonds on each particle. - ticket:86
***Warning! The integrator you are using doesn't report conserved_quantitiy, logging a value of 0.0 when trying to log conserved_quantity with integrate.npt
The conserved quantity calculation didn't make it into the 0.8.0 release. It should be in the next release. - ticket:113
HOOMD crashes when dumping a dcd file in windows
A "one-character" code bug slipped through testing in the format specified to strftime. This issue is fixed in the 0.8.0.1 hotfix release binaries and in branches/hoomd-0.8 r1581
***Error! unspecified launch failure after ..../BinnedNeighborListGPU.cc:628 when running on compute 1.1 devices
- Note: The line number specified may be different from 628 if you are not running with the --gpu_error_checking option enabled
- Typically, this error occurs after several hours of run time.
- A high GPU core temperature may or may not be related to the problem occurring more often.
- So far, the only reported devices this problem occurs on are all compute 1.1 (see Appendix A of the CUDA programming guide for a list of devices and their compute capability: http://www.nvidia.com/object/cuda_develop.html ).
- Again, compute 1.0 and 1.3 devices appear to be unaffected. I've got months of runtime on Tesla C870s and weeks on a GTX 280 w/o any problems.
