GitLab

* Make reference/CPU minimizer into a kernel

* Add per-platform support for GPU minimization

* Initial implementation of GPU minimization

* Fixes

* Increase robustness when initial gradient is huge

* Handle overflow leading to non-finite values gracefully

* Handle large forces in single precision more robustly

* Optimize kernels

* Fix kernel launch size

* Update banner years

* Don't create MinimizeKernel until first minimization requested

* Make some compile-time constants into kernel arguments

* Consolidate scale calculation kernel

* Condense alpha/beta reduction kernels using atomics

* Condense line search dot kernels with reductions

* Remove a download, and download grad norm separately

* Asynchronously check lbfgs convergence condition

* Restructure line search to avoid download waiting

* Start line search preemptively in case CPU evaluation is not needed

* In rare cases, constraint error might not decrease after one optimization round

* Better handling of unsupported 64-bit atomics, use FLT_MAX

* Pick gradient mode based on GPU vs. CPU evaluation

* Rework getDiff/getScale reduction, remove reduceBuffer

* Older CUDA might not like float hex literals

* Fix error in a comment

* Replace SimTK-containing file headers

* Update file headers for new Tinker reader files added

Moved SFMT into libraries folder, and restructured it to allow multiple independent random number generators

Modified the API for lots of classes so that the number of particles/bonds/etc. need not be specified in the constructor.

Changed how kernels get initialized.  StandardMMForceField now has you set the 1-4 interactions explicitly.  CMMotionRemover allows the removal frequency to be set.

Continuing to implement ReferencePlatform (implemented kinetic energy calculation and LangevinIntegrator)

Created a test case for StandardMMForceFieldImpl's findExclusions() method.  Fixed some bugs that it found.  Created a CMake script and configured it for building and running test cases.