TNL::Algorithms::SequentialFor< Device > Struct Template Reference

Wrapper to ParallelFor which makes it run sequentially. More...

#include <TNL/Algorithms/SequentialFor.h>

Static Public Member Functions
template<typename Index , typename Function >
static void	exec (Index start, Index end, Function f)
	Static method for execution of the loop.

Detailed Description

template<typename Device = Devices::Sequential>
struct TNL::Algorithms::SequentialFor< Device >

Wrapper to ParallelFor which makes it run sequentially.

It is helpfull for debuging or just sequential for loops on GPUs.

Member Function Documentation

exec()

template<typename Device = Devices::Sequential>

template<typename Index , typename Function >

static void TNL::Algorithms::SequentialFor< Device >::exec ( Index start, Index end, Function f )

inlinestatic

Static method for execution of the loop.

Template Parameters

Index	defines the type of indexes over which the loop iterates.
Function	is the type of function to be called in each iteration.

Parameters

start	the for-loop iterates over index interval [start, end).
end	the for-loop iterates over index interval [start, end).
f	is the function to be called in each iteration

Example

#include <iostream>
#include <cstdlib>
#include <TNL/Containers/Vector.h>
#include <TNL/Algorithms/parallelFor.h>
#include <TNL/Algorithms/SequentialFor.h>
 
using namespace TNL;
using namespace TNL::Containers;
 
template< typename Device >
void
printVector()
{
   const int size( 60 );
   TNL::Containers::Vector< float, Device > v( size, 1.0 );
   auto view = v.getView();
   auto print = [ = ] __cuda_callable__( int i ) mutable
   {
      if( i % 5 == 0 )
         printf( "v[ %d ] = %f \n", i, view[ i ] );  // we use printf because of compatibility with GPU kernels
   };
   std::cout << "Printing vector using parallel for: " << std::endl;
   Algorithms::parallelFor< Device >( 0, v.getSize(), print );
 
   std::cout << "Printing vector using sequential for: " << std::endl;
   Algorithms::SequentialFor< Device >::exec( 0, v.getSize(), print );
}
 
int
main( int argc, char* argv[] )
{
   std::cout << "Example on the host:" << std::endl;
   printVector< TNL::Devices::Host >();
 
#ifdef __CUDACC__
   std::cout << "Example on CUDA GPU:" << std::endl;
   printVector< TNL::Devices::Cuda >();
#endif
   return EXIT_SUCCESS;
}

Output

Example on the host:
Printing vector using parallel for: 
v[ 0 ] = 1.000000 
v[ 5 ] = 1.000000 
v[ 10 ] = 1.000000 
v[ 15 ] = 1.000000 
v[ 20 ] = 1.000000 
v[ 25 ] = 1.000000 
v[ 30 ] = 1.000000 
v[ 35 ] = 1.000000 
v[ 40 ] = 1.000000 
v[ 45 ] = 1.000000 
v[ 50 ] = 1.000000 
v[ 55 ] = 1.000000 
Printing vector using sequential for: 
v[ 0 ] = 1.000000 
v[ 5 ] = 1.000000 
v[ 10 ] = 1.000000 
v[ 15 ] = 1.000000 
v[ 20 ] = 1.000000 
v[ 25 ] = 1.000000 
v[ 30 ] = 1.000000 
v[ 35 ] = 1.000000 
v[ 40 ] = 1.000000 
v[ 45 ] = 1.000000 
v[ 50 ] = 1.000000 
v[ 55 ] = 1.000000 
Example on CUDA GPU:
Printing vector using parallel for: 
v[ 35 ] = 1.000000 
v[ 40 ] = 1.000000 
v[ 45 ] = 1.000000 
v[ 50 ] = 1.000000 
v[ 55 ] = 1.000000 
v[ 0 ] = 1.000000 
v[ 5 ] = 1.000000 
v[ 10 ] = 1.000000 
v[ 15 ] = 1.000000 
v[ 20 ] = 1.000000 
v[ 25 ] = 1.000000 
v[ 30 ] = 1.000000 
Printing vector using sequential for: 
v[ 0 ] = 1.000000 
v[ 5 ] = 1.000000 
v[ 10 ] = 1.000000 
v[ 15 ] = 1.000000 
v[ 20 ] = 1.000000 
v[ 25 ] = 1.000000 
v[ 30 ] = 1.000000 
v[ 35 ] = 1.000000 
v[ 40 ] = 1.000000 
v[ 45 ] = 1.000000 
v[ 50 ] = 1.000000 
v[ 55 ] = 1.000000 

---

The documentation for this struct was generated from the following file:

• src/TNL/Algorithms/SequentialFor.h