Commit afba52d9 authored by Jakub Klinkovský's avatar Jakub Klinkovský
Browse files

Renamed prefixSum methods to scan 

Closes #49
parent d070cc39

src/Benchmarks/BLAS/vector-operations.h

+16 −16
Original line number Diff line number Diff line
@@ -562,31 +562,31 @@ benchmarkVectorOperations( Benchmark & benchmark, 
#endif



   ////

   // Inclusive prefix sum

   auto inclusivePrefixSumHost = [&]() {

      hostVector.prefixSum();

   // Inclusive scan

   auto inclusiveScanHost = [&]() {

      hostVector.scan();

   };

   benchmark.setOperation( "inclusive prefix sum", 2 * datasetSize );

   benchmark.time< Devices::Host >( reset1, "CPU ET", inclusivePrefixSumHost );

   benchmark.setOperation( "inclusive scan", 2 * datasetSize );

   benchmark.time< Devices::Host >( reset1, "CPU ET", inclusiveScanHost );

#ifdef HAVE_CUDA

   auto inclusivePrefixSumCuda = [&]() {

      deviceVector.prefixSum();

   auto inclusiveScanCuda = [&]() {

      deviceVector.scan();

   };

   benchmark.time< Devices::Cuda >( reset1, "GPU ET", inclusivePrefixSumCuda );

   benchmark.time< Devices::Cuda >( reset1, "GPU ET", inclusiveScanCuda );

#endif



   ////

   // Exclusive prefix sum

   auto exclusivePrefixSumHost = [&]() {

      hostVector.template prefixSum< Algorithms::ScanType::Exclusive >();

   // Exclusive scan

   auto exclusiveScanHost = [&]() {

      hostVector.template scan< Algorithms::ScanType::Exclusive >();

   };

   benchmark.setOperation( "exclusive prefix sum", 2 * datasetSize );

   benchmark.time< Devices::Host >( reset1, "CPU ET", exclusivePrefixSumHost );

   benchmark.setOperation( "exclusive scan", 2 * datasetSize );

   benchmark.time< Devices::Host >( reset1, "CPU ET", exclusiveScanHost );

#ifdef HAVE_CUDA

   auto exclusivePrefixSumCuda = [&]() {

      deviceVector.template prefixSum< Algorithms::ScanType::Exclusive >();

   auto exclusiveScanCuda = [&]() {

      deviceVector.template scan< Algorithms::ScanType::Exclusive >();

   };

   benchmark.time< Devices::Cuda >( reset1, "GPU ET", exclusivePrefixSumCuda );

   benchmark.time< Devices::Cuda >( reset1, "GPU ET", exclusiveScanCuda );

#endif



#ifdef HAVE_CUDA

src/TNL/Algorithms/DistributedScan.h

+1 −1
Original line number Diff line number Diff line
@@ -54,7 +54,7 @@ struct DistributedScan 
         // NOTE: exchanging general data types does not work with MPI

         CommunicatorType::Alltoall( dataForScatter, 1, rankSums.getData(), 1, group );



         // compute prefix-sum of the per-rank sums

         // compute the scan of the per-rank sums

         Scan< Devices::Host, ScanType::Exclusive >::perform( rankSums, 0, nproc, reduction, zero );



         // perform second phase: shift by the per-block and per-rank offsets

src/TNL/Containers/DistributedVector.h

+1 −1
Original line number Diff line number Diff line
@@ -131,7 +131,7 @@ public: 
   DistributedVector& operator/=( const Vector& vector );



   template< Algorithms::ScanType Type = Algorithms::ScanType::Inclusive >

   void prefixSum( IndexType begin = 0, IndexType end = 0 );

   void scan( IndexType begin = 0, IndexType end = 0 );

};



} // namespace Containers

src/TNL/Containers/DistributedVector.hpp

+1 −1
Original line number Diff line number Diff line
@@ -273,7 +273,7 @@ template< typename Real, 
   template< Algorithms::ScanType Type >

void

DistributedVector< Real, Device, Index, Communicator >::

prefixSum( IndexType begin, IndexType end )

scan( IndexType begin, IndexType end )

{

   if( end == 0 )

      end = this->getSize();

src/TNL/Containers/DistributedVectorView.h

+1 −1
Original line number Diff line number Diff line
@@ -134,7 +134,7 @@ public: 
   DistributedVectorView& operator/=( const Vector& vector );



   template< Algorithms::ScanType Type = Algorithms::ScanType::Inclusive >

   void prefixSum( IndexType begin = 0, IndexType end = 0 );

   void scan( IndexType begin = 0, IndexType end = 0 );

};



} // namespace Containers