Merge branch release-2018 into release-2019

Resolved Conflicts:
	src/gromacs/ewald/pme-solve.cu
	src/gromacs/mdlib/clincs.cpp
	src/programs/mdrun/md.cpp
	src/programs/mdrun/tests/CMakeLists.txt

Change-Id: I813b4ed7ecf8056db09ab749c367a3b5e204e701

diff --git a/docs/release-notes/2018/2018.5.rst b/docs/release-notes/2018/2018.5.rst
index d1649a948a76c16e115e0b32edfeb8569000d4ab..b73325f74bad4f4b5bbc9b5008c442431fe4389a 100644 (file)
--- a/docs/release-notes/2018/2018.5.rst
+++ b/docs/release-notes/2018/2018.5.rst
@@ -19,6 +19,25 @@ was reduced to near machine precision. The change does not affect
 the results for non-polarizable systems, such as proteins or small
 molecules.
 
+Make large PME grids work on GPU
+"""""""""""""""""""""""""""""""""""""""""""
+
+PME grids with size along Z larger than 511 would make mdrun exit
+with a cryptic CUDA error.
+
+:issue: `2779`
+
+Fix LINCS accuracy with OpenMP when constraint triangles are present
+""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
+
+Constraint triangles, which usually only occur when replacing hydrogens
+by virtual interaction sites in CH3 and NH3 groups, need double the number
+of iterations as normal constraints. With OpenMP this would only happen
+when the last OpenMP thread has at least one such triangle. This would
+cause a slight loss of accuracy in inhomogeneous systems.
+
+:issue: `2808`
+
 Fixes for ``gmx`` tools
 ^^^^^^^^^^^^^^^^^^^^^^^