Hi. I whote a function for cumulative summation, which is defined by y_i = sum_0^i x_i. I whote this function in two style.
One is ParallelScanSum.hpp.
[cpp]//
// ParallelScanSum.hpp
//
#include
#include "tbb/task_scheduler_init.h"
#include "tbb/blocked_range.h"
#include "tbb/parallel_scan.h"
template
class ParallelScanSumClass{
public:
ParallelScanSumClass( T* __restrict results_, const T* __restrict inputData_ ) : sum_(0), inputData_( inputData_ ), results_( results_ ){}
ParallelScanSumClass( ParallelScanSumClass& obj, tbb::split ) : sum_(0), inputData_( obj.inputData_ ), results_( obj.results_ ){}
void reverse_join( ParallelScanSumClass& obj ) {
sum_ = obj.sum_ + sum_;
}
void assign( ParallelScanSumClass& obj ) {
sum_ = obj.sum_;
}
template< typename Tag > void operator()( const tbb::blocked_range< unsigned int >& range, Tag ) {
T tmp = sum_;
for ( int i = range.begin(); i != range.end(); ++i ) {
tmp = tmp + inputData_[ i ];
if ( Tag::is_final_scan() ) {
results_[ i ] = tmp;
}
}
sum_ = tmp;
}
T sum_;
T* results_;
const T* inputData_;
};
template void ParallelScanSum( unsigned int n, const T* __restrict inp, T* __restrict out){
ParallelScanSumClass pss( out, inp );
tbb::parallel_scan( tbb::blocked_range( 0, n ), pss, tbb::auto_partitioner() );
}
[/cpp]
Another is "concurrent_vector_ParallelScanSum.hpp".
[cpp]//
// concurrent_vector_ParallelScanSum.hpp
//
#include
#include "tbb/task_scheduler_init.h"
#include "tbb/blocked_range.h"
#include "tbb/concurrent_vector.h"
#include "tbb/parallel_scan.h"
namespace concurrent_vector{
template
class ParallelScanSumClass{
public:
ParallelScanSumClass( tbb::concurrent_vector &results_,
const tbb::concurrent_vector &inputData_ ) : sum_(0), results_( results_ ), inputData_( inputData_ ){}
ParallelScanSumClass( ParallelScanSumClass& obj, tbb::split ) : inputData_( obj.inputData_ ), results_( obj.results_ ),sum_(0){}
void reverse_join( ParallelScanSumClass& obj ) {
sum_ = obj.sum_ + sum_;
}
void assign( ParallelScanSumClass& obj ) {
sum_ = obj.sum_;
}
template< typename Tag > void operator()( const tbb::blocked_range< unsigned int >& range, Tag ) {
T tmp = sum_;
for ( int i = range.begin(); i != range.end(); ++i ) {
tmp = tmp + inputData_[ i ];
if ( Tag::is_final_scan() ) {
results_[ i ] = tmp;
}
}
sum_ = tmp;
}
T sum_;
tbb::concurrent_vector results_;
const tbb::concurrent_vector inputData_;
};
template void ParallelScanSum( unsigned int n, const tbb::concurrent_vector& inp, tbb::concurrent_vector& out){
ParallelScanSumClass pss( out, inp );
tbb::parallel_scan( tbb::blocked_range( 0, n ), pss, tbb::auto_partitioner() );
}
}
[/cpp]
And this is the main program.
[cpp]//
// main.cpp
//
#include
#include "tbb/concurrent_vector.h"
#include "tbb/task_scheduler_init.h"
#include "parallelScanSum.hpp"
#include "concurrent_vector_parallelScanSum.hpp"
int main(){
tbb::concurrent_vector inp;
tbb::concurrent_vector out(10);
unsigned int inp_array[10];
unsigned int out_array[10];
for(unsigned int i = 0; i < 10; ++i){
inp.push_back( i );
inp_array = i;
}
ParallelScanSum( 10, inp_array, out_array );
for( unsigned int i = 0; i < 10; ++i ){
std::cout << "out_array[" << i << "]:" << out_array << std::endl;
}
concurrent_vector::ParallelScanSum( 10, inp, out );
for( unsigned int i = 0; i < 10; ++i ){
std::cout << "out[" << i << "]:" << out << std::endl;
}
return 0;
}[/cpp]
The result is
out_array[0]:0
out_array[1]:1
out_array[2]:3
out_array[3]:6
out_array[4]:10
out_array[5]:15
out_array[6]:21
out_array[7]:28
out_array[8]:36
out_array[9]:45
out[0]:0
out[1]:0
out[2]:0
out[3]:0
out[4]:0
out[5]:0
out[6]:0
out[7]:0
out[8]:0
out[9]:0
As you see here, the conventional array version of this function (parallelScanSum.hpp) is O.K. but the tbb::concurrent_vector version (concurrent_vector_parallelScanSum.hpp) is not. Do you have any idea for this failure?
Thanks.
