Line data    Source code

       1             : /* +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
       2             :    Copyright (c) 2014-2023 The plumed team
       3             :    (see the PEOPLE file at the root of the distribution for a list of names)
       4             : 
       5             :    See http://www.plumed.org for more information.
       6             : 
       7             :    This file is part of plumed, version 2.
       8             : 
       9             :    plumed is free software: you can redistribute it and/or modify
      10             :    it under the terms of the GNU Lesser General Public License as published by
      11             :    the Free Software Foundation, either version 3 of the License, or
      12             :    (at your option) any later version.
      13             : 
      14             :    plumed is distributed in the hope that it will be useful,
      15             :    but WITHOUT ANY WARRANTY; without even the implied warranty of
      16             :    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
      17             :    GNU Lesser General Public License for more details.
      18             : 
      19             :    You should have received a copy of the GNU Lesser General Public License
      20             :    along with plumed.  If not, see <http://www.gnu.org/licenses/>.
      21             : +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ */
      22             : #ifndef __PLUMED_tools_OpenMP_h
      23             : #define __PLUMED_tools_OpenMP_h
      24             : 
      25             : #include <vector>
      26             : 
      27             : namespace PLMD {
      28             : 
      29             : namespace OpenMP {
      30             : 
      31             : /// Keep a dynamically linked OpenMP runtime resident when the kernel is unloaded.
      32             : /// Currently needed on Linux, where runtime worker threads can outlive dlclose().
      33             : /// Set PLUMED_OPENMP_RETAIN_RUNTIME=no before constructing any PLUMED instance
      34             : /// to disable retention. Once retained, a runtime cannot be made unloadable again.
      35             : void retainRuntime();
      36             : 
      37             : /// Set number of threads that can be used by openMP
      38             : void setNumThreads(const unsigned nt);
      39             : 
      40             : /// Get number of threads that can be used by openMP
      41             : unsigned getNumThreads();
      42             : 
      43             : /// Returns a unique thread identification number within the current team
      44             : unsigned getThreadNum();
      45             : 
      46             : /// get cacheline size
      47             : unsigned getCachelineSize();
      48             : 
      49             : /// Get a reasonable number of threads so as to access to an array of size s located at x
      50             : template<typename T>
      51      917044 : unsigned getGoodNumThreads(const T* /*getTheType*/,unsigned n) {
      52             :   // this is more or less the equivalent of writing "unsigned getGoodNumThreads<T>(unsigned n)"
      53             : 
      54             :   // a factor two is necessary since there is no guarantee that x is aligned
      55             :   // to cache line boundary
      56      917044 :   unsigned m=n*sizeof(T)/(2*getCachelineSize());
      57      917044 :   unsigned numThreads=getNumThreads();
      58      917044 :   if(m>=numThreads) {
      59             :     m=numThreads;
      60             :   } else {
      61             :     //it is better to use either all the active threads or only one
      62             :     //this solves a performance problem as explained in issue #415
      63             :     m=1;
      64             :   }
      65      917044 :   return m;
      66             : }
      67             : 
      68             : /// Get a reasonable number of threads so as to access to vector v
      69             : template<typename T>
      70        1350 : unsigned getGoodNumThreads(const std::vector<T> & v) {
      71        1350 :   if(v.size()==0) {
      72             :     return 1;
      73             :   } else {
      74        1314 :     return getGoodNumThreads(&v[0],v.size());
      75             :   }
      76             : }
      77             : 
      78             : }//namespace OpenMP
      79             : }//namespace PLMD
      80             : 
      81             : #endif