Commit 59a4d824 authored by Tomáš Oberhuber's avatar Tomáš Oberhuber
Browse files

Added sequential device to atomic operations.

parent d82fd0bd

src/TNL/Algorithms/AtomicOperations.h

+15 −0
Original line number Diff line number Diff line
@@ -41,6 +41,21 @@ struct AtomicOperations< Devices::Host > 
   }

};



template<>

struct AtomicOperations< Devices::Sequential >

{

   // this is __cuda_callable__ only to silence nvcc warnings (all methods inside class

   // template specializations must have the same execution space specifier, otherwise

   // nvcc complains)

   TNL_NVCC_HD_WARNING_DISABLE

   template< typename Value >

   __cuda_callable__

   static void add( Value& v, const Value& a )

   {

      v += a;

   }

};



template<>

struct AtomicOperations< Devices::Cuda >

{

src/TNL/Atomic.h

+12 −0
Original line number Diff line number Diff line
@@ -15,6 +15,7 @@ 
#include <atomic>  // std::atomic



#include <TNL/Devices/Host.h>

#include <TNL/Devices/Sequential.h>

#include <TNL/Devices/Cuda.h>



// double-precision atomicAdd function for Maxwell and older GPUs
@@ -96,6 +97,17 @@ public: 
   }

};



template< typename T >

class Atomic< T, Devices::Sequential > : public Atomic< T, Devices::Host >

{

   public:



   using Atomic;

   using operator=;

   using fetch_max;

   using fetch_min;

};



template< typename T >

class Atomic< T, Devices::Cuda >

{