Merge branch release-2018

Minor conflicts documenting gpuregiontimer_ocl.h, resolved
with a suitable mix from both branches.

Change-Id: I5bb417acaff941f18730f0417c5e46c8f4c2fc1e

diff --cc docs/user-guide/mdp-options.rst
Simple merge

diff --cc docs/user-guide/terminology.rst
Simple merge

diff --cc src/gromacs/gpu_utils/gpuregiontimer_ocl.h
index 91e9e8de23bdf8faeebc28217caea500c4aa5e9f,81444cafdcee5a10b73927ba191bf2d516b52e67..00f28c633fd032235896eac981d872bfd7f28511
--- 1/src/gromacs/gpu_utils/gpuregiontimer_ocl.h
--- 2/src/gromacs/gpu_utils/gpuregiontimer_ocl.h
+++ b/src/gromacs/gpu_utils/gpuregiontimer_ocl.h
@@@ -51,14 -50,32 +51,18 @@@
  
  #include "gpuregiontimer.h"
  
 -template <> struct GpuTraits<GpuFramework::OpenCL>
 -{
 -    using CommandStream = cl_command_queue;
 -    using CommandEvent  = cl_event;
 -};
 -
 -//! Short-hand for external use
 -using GpuRegionTimer = GpuRegionTimerWrapper<GpuFramework::OpenCL>;
 -
 -// cppcheck-suppress noConstructor
  /*! \libinternal \brief
-  * This is a GPU region timing implementation for OpenCL.
-  * It provides methods for measuring the last timespan.
-  * Copying/assignment is disabled since the underlying timing events are owned by this.
+  * The OpenCL implementation of the GPU code region timing.
+  * With OpenCL, one has to use cl_event handle for each API call that has to be timed, and
+  * accumulate the timing afterwards. As we would like to avoid overhead on API calls,
+  * we only query and accumulate cl_event timing at the end of time steps, not after the API calls.
+  * Thus, this implementation does not reuse a single cl_event for multiple calls, but instead
+  * maintains an array of cl_events to be used within any single code region.
+  * The array size is fixed at a small but sufficiently large value for the number of cl_events
+  * that might contribute to a timer region, currently 10.
   */
- // cppcheck-suppress noConstructor
 -template <> class GpuRegionTimerImpl<GpuFramework::OpenCL>
 +class GpuRegionTimerImpl
  {
 -    //! Short-hands
 -    using CommandStream = typename GpuTraits<GpuFramework::OpenCL>::CommandStream;
 -    using CommandEvent  = typename GpuTraits<GpuFramework::OpenCL>::CommandEvent;
 -
      /*! \brief The underlying individual timing events array.
       * The maximum size is chosen arbitrarily to work with current code, and can be changed.
       * There is simply no need for run-time resizing, and it's unlikely we'll ever need more than 10.
@@@ -69,22 -86,11 +73,20 @@@
  
      public:
  
-         /*! \brief Will be called before the region start. */
 +        GpuRegionTimerImpl()  = default;
 +        ~GpuRegionTimerImpl() = default;
 +        //! No copying
 +        GpuRegionTimerImpl(const GpuRegionTimerImpl &)       = delete;
 +        //! No assignment
 +        GpuRegionTimerImpl &operator=(GpuRegionTimerImpl &&) = delete;
 +        //! Moving is disabled but can be considered in the future if needed
 +        GpuRegionTimerImpl(GpuRegionTimerImpl &&)            = delete;
 +
+         /*! \brief Should be called before the region start. */
          inline void openTimingRegion(CommandStream){}
- 
-         /*! \brief Will be called after the region end. */
+         /*! \brief Should be called after the region end. */
          inline void closeTimingRegion(CommandStream){}
- 
-         /*! \brief Returns the last measured region timespan (in milliseconds) and calls reset() */
+         /*! \brief Returns the last measured region timespan (in milliseconds) and calls reset(). */
          inline double getLastRangeTime()
          {
              double milliseconds = 0.0;
@@@ -122,11 -128,7 +124,10 @@@
              // As long as we're doing nullptr checks, we might want to be extra cautious.
              events_.fill(nullptr);
          }
- 
 -        /*! \brief Provides next unused cl_event for OpenCL API consumption. */
 +        /*! \brief Returns a new raw timing event
 +         * for passing into individual GPU API calls
 +         * within the region if the API requires it (e.g. on OpenCL).
 +         */
          inline CommandEvent *fetchNextEvent()
          {
              GMX_ASSERT(currentEvent_ < events_.size(), "Increase c_maxEventNumber_ if needed");