|
|
Boost-Commit : |
Subject: [Boost-commit] svn:boost r51439 - sandbox/interthreads/libs/interthreads/example
From: vicente.botet_at_[hidden]
Date: 2009-02-24 15:37:37
Author: viboes
Date: 2009-02-24 15:37:37 EST (Tue, 24 Feb 2009)
New Revision: 51439
URL: http://svn.boost.org/trac/boost/changeset/51439
Log:
0.4.1 : adding parallel-sort example
Added:
sandbox/interthreads/libs/interthreads/example/parallel_sort2.cpp (contents, props changed)
Text files modified:
sandbox/interthreads/libs/interthreads/example/parallel_sort.cpp | 226 ++++++++-------------------------------
1 files changed, 47 insertions(+), 179 deletions(-)
Modified: sandbox/interthreads/libs/interthreads/example/parallel_sort.cpp
==============================================================================
--- sandbox/interthreads/libs/interthreads/example/parallel_sort.cpp (original)
+++ sandbox/interthreads/libs/interthreads/example/parallel_sort.cpp 2009-02-24 15:37:37 EST (Tue, 24 Feb 2009)
@@ -29,8 +29,7 @@
#include <assert.h>
-#define STAT
-#ifdef STAT
+
template <typename Range, std::size_t Parts>
class partition
{
@@ -45,25 +44,11 @@
}
parts_[Parts-1]=boost::sub_range<Range>(boost::begin(range)+(Parts-1)*(size/Parts), boost::end(range));
}
- boost::sub_range<Range>& operator[](unsigned i) {
- return parts_[i];
- }
-};
-#else
-template <typename Range>
-class partition
-{
- boost::sub_range<Range> range_;
- std::size_t parts_;
-public:
- partition(boost::sub_range<Range>& range, std::size_t parts) : range_(range), parts_() {}
boost::sub_range<Range> operator[](unsigned i) {
- std::size_t size = boost::size(range_);
- if (i==parts_ - 1) return boost::sub_range<Range>(boost::begin(range_)+i*(size/parts_), boost::end(range_));
- else return boost::sub_range<Range>(boost::begin(range_)+i*(size/parts_), boost::begin(range_)+((i+1)*(size/parts_)));
+ return parts_[i];
}
};
-#endif
+
typedef boost::tp::pool<
boost::tp::unbounded_channel< boost::tp::fifo >
> pool_type;
@@ -74,129 +59,76 @@
typedef boost::tp::task< void > task_type;
#endif
-struct sort_fct {
- template<class RandomAccessRange>
- RandomAccessRange& operator()(RandomAccessRange& rng) {
- return boost::sort(rng);
- }
-};
-
-struct inplace_merge_fct {
- template<class BidirectionalRange>
- BidirectionalRange&
- operator()( BidirectionalRange& rng,
- typename boost::range_iterator<BidirectionalRange>::type middle ) {
- return boost::inplace_merge(rng, middle);
- }
-};
-
-class x_sort
-{
-private:
- pool_type & pool_;
- unsigned cutoff_;
-
- template <typename Range>
-#ifdef STAT
- void seq_( boost::sub_range<Range>& range)
-#else
- void seq_( boost::iterator_range<typename boost::range_iterator<Range>::type> range)
-#endif
- {
- sort_fct()(range);
- }
- template <typename Range>
-#ifdef STAT
- void par_( boost::sub_range<Range>& range)
-#else
- void par_( boost::iterator_range<typename boost::range_iterator<Range>::type> range)
-#endif
- {
+template <
+ typename DirectSolver,
+ typename Composer,
+ typename AE,
+ typename Range
+>
+ void inplace_solve( AE & ae, boost::sub_range<Range> range, unsigned cutoff );
+template <
+ typename DirectSolver,
+ typename Composer,
+ typename AE,
+ typename Range
+>
+ void inplace_solve( AE & ae,
+ boost::sub_range<Range> range,
+ unsigned cutoff) {
unsigned size = boost::size(range);
//std::cout << "<<par_ " << size << std::endl;
- if ( size <= cutoff_) return seq_<Range>( range);
- else
- {
-#if 1
- #define BOOST_PARTS 2
-#ifdef STAT
+ if ( size <= cutoff) DirectSolver()(range);
+ else {
+#define BOOST_PARTS 2
partition<Range, BOOST_PARTS> parts(range);
-#else
-// partition<Range> parts(range, BOOST_PARTS);
-#endif
task_type tasks[BOOST_PARTS];
for (unsigned i=0;i < BOOST_PARTS-1; ++i) {
- task_type tmp(pool_.submit(
+ task_type tmp(ae.submit(
boost::bind(
- & x_sort::par_<Range>,
- boost::ref( * this),
-#ifdef STAT
- boost::ref(parts[i])
-#else
- boost::make_sliced_range(range, i*(size/BOOST_PARTS), ((i+1)*(size/BOOST_PARTS)))
-#endif
+ &inplace_solve<DirectSolver,Composer,AE,Range>,
+ boost::ref(ae),
+ parts[i],
+ cutoff
)));
tasks[i] = tmp;
}
-#ifdef STAT
- this->par_(parts[BOOST_PARTS-1]);
-#else
- //this->par_(parts[BOOST_PARTS-1]);
- this->par_<Range>(boost::make_sliced_range(range, (BOOST_PARTS-1)*(size/BOOST_PARTS), size));
-#endif
+ inplace_solve<DirectSolver,Composer,AE,Range>(ae, parts[BOOST_PARTS-1], cutoff);
for (unsigned i=0;i < BOOST_PARTS-1; ++i) {
tasks[i].wait();
};
-#else
- boost::sub_range<Range> left(boost::begin(range), boost::begin(range)+(size/2));
- boost::sub_range<Range> right(boost::begin(range)+(size/2), boost::end(range));
- // fork a new sub-action t1 in pool
- task_type task(
- pool_.submit(
- boost::bind(
- & x_sort::par_<Range>,
- boost::ref( * this),
- boost::ref(left))
- )
- );
-
- this->par_<Range>(right);
- task.wait();
-#endif
//std::cout << "par_inplace_merge_fct " << size << ">>"<< std::endl;
- inplace_merge_fct()(range, boost::begin(range)+(size/2));
+ Composer()(range);
//std::cout << "par_ " << size << ">>"<< std::endl;
}
}
-public:
- x_sort( pool_type & pool, unsigned cutoff)
- : pool_( pool), cutoff_( cutoff)
- {}
-
- template <typename Range>
- void execute( Range& range) {
-#ifdef STAT
- boost::sub_range<Range> rng(boost::begin(range), boost::end(range));
-#else
- boost::iterator_range<typename boost::range_iterator<Range>::type> rng(range);
-#endif
- par_<Range>( rng);
- }
+
+struct sort_fct {
+ template<class RandomAccessRange>
+ RandomAccessRange& operator()(RandomAccessRange rng) {
+ return boost::sort(rng);
+ }
};
+struct inplace_merge_fct {
+ template<class BidirectionalRange>
+ BidirectionalRange&
+ operator()( BidirectionalRange rng) {
+ return boost::inplace_merge(rng, boost::begin(rng)+(boost::size(rng)/2));
+ }
+};
template <typename Range>
void parallel_sort(Range& range, unsigned cutoff=10000) {
pool_type pool( boost::tp::poolsize( 2) );
- x_sort fct( pool, cutoff);
- fct.execute(range);
- std::cout << "parallel_sort " << ">>"<< std::endl;
+ boost::sub_range<Range> rng(boost::begin(range), boost::end(range));
+ inplace_solve<sort_fct,inplace_merge_fct,pool_type,Range>( pool, rng, cutoff);
+ //std::cout << "parallel_sort " << ">>"<< std::endl;
}
-#define NN 1000000
+#define NN 100000
int sorted[NN];
int values1[NN];
int values2[NN];
@@ -207,9 +139,7 @@
int main() {
//pool_type ae(boost::tp::poolsize(2));
- for (unsigned i=0; i<NN; ++i) sorted[i]=i;
-
-
+ for (unsigned i=0; i<NN; ++i) sorted[i]=i;
for (unsigned i=0; i<NN; ++i) values1[i]=NN-i-1;
{
@@ -241,37 +171,6 @@
//for (unsigned i=0; i<NN; ++i) std::cout << sorted[i] << " " <<values3[i] << std::endl;
assert(boost::equal(values3, sorted));
-#if 0
- for (unsigned i=0; i<NN; ++i) values3[i]=NN-i-1;
- std::cout << "parallel_sort 8000: reverse 0.."<<NN << std::endl;
- {
- boost::progress_timer tmr; // start timing
- parallel_sort(values3);
- }
- for (unsigned i=0; i<NN; ++i) values[i]=NN-i-1;
- std::cout << "parallel_sort 4000: reverse 0.."<<NN << std::endl;
- {
- boost::progress_timer tmr; // start timing
- parallel_sort(values, 4000);
- }
-
- std::cout << "std::sort: 0.." << NN << std::endl;
- {
- boost::progress_timer t; // start timing
- std::sort(values, values+NN);
- }
- std::cout << "boost::sort: 0.."<<NN << std::endl;
- {
- boost::progress_timer t; // start timing
- boost::sort(values);
- }
- std::cout << "parallel_sort: 0.."<<NN << std::endl;
- {
- boost::progress_timer tmr; // start timing
- parallel_sort(values);
- }
-#endif
-
for (unsigned i=0; i<NN; ++i) values4[i]=NN-i-1;
{
std::cout << "std::sort: reverse 0.." << NN << std::endl;
@@ -300,37 +199,6 @@
//for (unsigned i=0; i<NN; ++i) std::cout << sorted[i] << " " <<values3[i] << std::endl;
assert(boost::equal(values6, sorted));
- for (unsigned i=0; i<NN; ++i) values1[i]=NN-i-1;
- std::cout << "parallel_sort "<<NN/16<<": reverse 0.."<<NN << std::endl;
- {
- boost::progress_timer tmr; // start timing
- parallel_sort(values1, NN/16);
- }
-
- //for (unsigned i=0; i<NN; ++i) std::cout << sorted[i] << " " <<values3[i] << std::endl;
- assert(boost::equal(values1, sorted));
-
-#if 0
- for (unsigned i=0; i<NN; ++i) values[i]=NN-i-1;
- {
- std::cout << "boost::sort: reverse 0.."<<NN << std::endl;
- boost::progress_timer t; // start timing
- boost::sort(values);
- }
- assert(boost::equal(values, sorted));
-
- for (unsigned i=0; i<NN; ++i) values[i]=NN-i-1;
- std::cout << "parallel_sort "<<NN/32<<": reverse 0.."<<NN << std::endl;
- {
- boost::progress_timer tmr; // start timing
- parallel_sort(values, NN/32);
- }
-
- //for (unsigned i=0; i<NN; ++i) std::cout << sorted[i] << " " <<values[i] << std::endl;
- assert(boost::equal(values, sorted));
- //std::cout << "sleep"<<std::endl;
-#endif
-
//boost::this_thread::sleep(boost::posix_time::milliseconds(5000));
std::cout << "shutdown"<< std::endl;
pool.shutdown();
Added: sandbox/interthreads/libs/interthreads/example/parallel_sort2.cpp
==============================================================================
--- (empty file)
+++ sandbox/interthreads/libs/interthreads/example/parallel_sort2.cpp 2009-02-24 15:37:37 EST (Tue, 24 Feb 2009)
@@ -0,0 +1,213 @@
+//////////////////////////////////////////////////////////////////////////////
+//
+// (C) Copyright Vicente J. Botet Escriba 2008-2009.
+// Distributed under the Boost Software License, Version 1.0.
+// (See accompanying file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
+//
+// See http://www.boost.org/libs/interthreads for documentation.
+//
+// Based on the shared.cpp example from the threadalert library of Roland Schwarz
+//////////////////////////////////////////////////////////////////////////////
+
+#include <iostream>
+#include <algorithm>
+
+#include <boost/progress.hpp>
+#include <boost/thread/thread.hpp>
+#include <boost/bind.hpp>
+#include <boost/tp/pool.hpp>
+#include <boost/tp/unbounded_channel.hpp>
+#include <boost/tp/fifo.hpp>
+#include <boost/range/algorithm/sort.hpp>
+#include <boost/range/algorithm/inplace_merge.hpp>
+#include <boost/range/sub_range.hpp>
+#include <boost/range/begin.hpp>
+#include <boost/range/end.hpp>
+#include <boost/range/algorithm/equal.hpp>
+#include <boost/range/algorithm/for_each.hpp>
+#include <boost/range/algorithm/transform.hpp>
+#include <boost/range/adaptor/sliced.hpp>
+#include <boost/interthreads/fork.hpp>
+#include <boost/interthreads/algorithm/wait.hpp>
+#include <boost/interthreads/scheduler.hpp>
+#include <boost/array.hpp>
+
+#include <assert.h>
+
+#define BOOST_PARTS 2
+
+template <typename Range>
+class partition
+{
+public:
+ boost::iterator_range<typename boost::range_iterator<Range>::type> range_;
+ std::size_t parts_;
+ partition(boost::iterator_range<typename boost::range_iterator<Range>::type>& range, std::size_t parts):
+ range_(range),
+ parts_(parts)
+ {}
+ boost::iterator_range<typename boost::range_iterator<Range>::type> operator[](unsigned i) {
+ unsigned size = boost::size(range_);
+ if (i<(parts_-1))
+ return boost::make_sliced_range(range_, i*(size/parts_), ((i+1)*(size/parts_)));
+ else
+ return boost::make_sliced_range(range_, (parts_-1)*(size/parts_), size);
+ }
+};
+
+typedef boost::tp::pool<
+ boost::tp::unbounded_channel< boost::tp::fifo >
+> pool_type;
+
+#ifdef TASK_POOL
+ typedef boost::tp::task< pool_type, void > task_type;
+#else
+ typedef boost::tp::task< void > task_type;
+#endif
+
+
+template <
+ typename DirectSolver,
+ typename Composer,
+ typename AE,
+ typename Range
+>
+ void inplace_solve( AE & ae,
+ boost::iterator_range<typename boost::range_iterator<Range>::type> range,
+ unsigned cutoff );
+
+template <
+ typename DirectSolver,
+ typename Composer,
+ typename AE,
+ typename Range
+>
+ void inplace_solve( AE & ae,
+ boost::iterator_range<typename boost::range_iterator<Range>::type> range,
+ unsigned cutoff )
+ {
+ unsigned size = boost::size(range);
+ //std::cout << "<<par_ " << size << std::endl;
+ if ( size <= cutoff) DirectSolver()(range);
+ else {
+ partition<Range> parts(range, BOOST_PARTS);
+ std::list<task_type> tasks;
+ #if 1 // this code do not compiles with gcc 3.4.4 cygwin
+ boost::transform(parts, boost::begin(tasks),
+ boost::bind(AE::submit, boost::ref(ae),
+ //boost::bind(&boost::interthreads::fork<AE>, boost::ref(ae),
+ boost::bind(&inplace_solve<DirectSolver,Composer,AE,Range>, boost::ref(ae),_1,cutoff)));
+ #else
+ for (unsigned i=0;i < BOOST_PARTS-1; ++i) {
+ task_type tmp(ae.submit(
+ boost::bind(
+ &inplace_solve<DirectSolver,Composer,AE,Range>,
+ boost::ref(ae),
+ parts[i],
+ cutoff
+ )));
+ tasks.push_back(tmp);
+ }
+ #endif
+ inplace_solve<DirectSolver,Composer,AE,Range>(ae, parts[BOOST_PARTS-1], cutoff);
+ boost::for_each(tasks, &boost::interthreads::wait_act<task_type>);
+
+ //std::cout << "par_inplace_merge_fct " << size << ">>"<< std::endl;
+ Composer()(range);
+ //std::cout << "par_ " << size << ">>"<< std::endl;
+
+ }
+ }
+
+struct sort_fct {
+ template<class RandomAccessRange>
+ RandomAccessRange& operator()(RandomAccessRange rng) {
+ return boost::sort(rng);
+ }
+};
+
+struct inplace_merge_fct {
+ template<class BidirectionalRange>
+ BidirectionalRange&
+ operator()( BidirectionalRange rng) {
+ return boost::inplace_merge(rng, boost::begin(rng)+(boost::size(rng)/2));
+ }
+};
+template <typename AE, typename Range>
+void parallel_sort(AE& ae, Range& range, unsigned cutoff=10000) {
+ boost::iterator_range<typename boost::range_iterator<Range>::type> rng(range);
+ inplace_solve<sort_fct,inplace_merge_fct,pool_type,Range>( ae, rng, cutoff);
+ //std::cout << "parallel_sort " << ">>"<< std::endl;
+}
+
+#define NN 100000
+int sorted[NN];
+int values1[NN];
+int values2[NN];
+int values3[NN];
+int values4[NN];
+int values5[NN];
+int values6[NN];
+
+int main() {
+ for (unsigned i=0; i<NN; ++i) sorted[i]=i;
+
+ for (unsigned i=0; i<NN; ++i) values1[i]=NN-i-1;
+ {
+ std::cout << "std::sort: reverse 0.." << NN << std::endl;
+ boost::progress_timer t; // start timing
+ std::sort(boost::begin(values1), boost::end(values1));
+ }
+ assert(boost::equal(values1, sorted));
+
+ for (unsigned i=0; i<NN; ++i) values2[i]=NN-i-1;
+ {
+ std::cout << "boost::sort: reverse 0.."<<NN << std::endl;
+ boost::progress_timer t; // start timing
+ boost::sort(values2);
+ }
+ assert(boost::equal(values2, sorted));
+
+ // creates a threadpool with two worker-threads
+ pool_type pool( boost::tp::poolsize( 2) );
+
+
+ for (unsigned i=0; i<NN; ++i) values3[i]=NN-i-1;
+ std::cout << "parallel_sort "<<NN/16<<": reverse 0.."<<NN << std::endl;
+ {
+ boost::progress_timer tmr; // start timing
+ parallel_sort(pool, values3, NN/16);
+ }
+ assert(boost::equal(values3, sorted));
+
+ for (unsigned i=0; i<NN; ++i) values4[i]=NN-i-1;
+ {
+ std::cout << "std::sort: reverse 0.." << NN << std::endl;
+ boost::progress_timer t; // start timing
+ std::sort(boost::begin(values4), boost::end(values4));
+ }
+ assert(boost::equal(values4, sorted));
+
+ for (unsigned i=0; i<NN; ++i) values5[i]=NN-i-1;
+ std::cout << "parallel_sort "<<NN/16<<": reverse 0.."<<NN << std::endl;
+ {
+ boost::progress_timer tmr; // start timing
+ parallel_sort(pool, values5, NN/16);
+ }
+ //for (unsigned i=0; i<NN; ++i) std::cout << sorted[i] << " " <<values3[i] << std::endl;
+ assert(boost::equal(values5, sorted));
+
+#if 0
+ for (unsigned i=0; i<NN; ++i) values6[i]=NN-i-1;
+ std::cout << "parallel_sort "<<NN/16<<": reverse 0.."<<NN << std::endl;
+ {
+ boost::progress_timer tmr; // start timing
+ parallel_sort(pool, values6, NN/16);
+ }
+ assert(boost::equal(values6, sorted));
+#endif
+ std::cout << "shutdown"<< std::endl;
+ pool.shutdown();
+ std::cout << "end"<< std::endl;
+ return 0;
+}
Boost-Commit list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk