#ifndef SORT_HEADER #define SORT_HEADER /*************************************************************************** * sort.h * * Fri Jan 17 11:26:42 2003 * Copyright 2003 Roman Dementiev * dementiev@mpi-sb.mpg.de ****************************************************************************/ #include #include "../mng/mng.h" #include "../common/rand.h" #include "../mng/adaptor.h" #include "../common/simple_vector.h" #include "../common/switch.h" #include "interleaved_alloc.h" #include "intksort.h" #include "adaptor.h" #include "async_schedule.h" #include "../mng/block_prefetcher.h" #include "../mng/buf_writer.h" #include "run_cursor.h" #include "loosertree.h" #include "inmemsort.h" //#define SORT_OPT_PREFETCHING //#define INTERLEAVED_ALLOC __STXXL_BEGIN_NAMESPACE //! \addtogroup stlalgo //! \{ /*! \internal */ namespace sort_local { template struct trigger_entry { typedef BIDTp_ bid_type; typedef ValTp_ value_type; bid_type bid; value_type value; operator bid_type() { return bid; }; }; template struct trigger_entry_cmp { typedef trigger_entry trigger_entry_type; ValueCmp_ cmp; trigger_entry_cmp(ValueCmp_ c): cmp(c) { } trigger_entry_cmp(const trigger_entry_cmp & a): cmp(a.cmp) { } bool operator ()(const trigger_entry_type & a,const trigger_entry_type & b) const { return cmp(a.value, b.value); }; }; template struct run_cursor2_cmp { typedef run_cursor2 cursor_type; value_cmp cmp; run_cursor2_cmp(value_cmp c):cmp(c) { } run_cursor2_cmp(const run_cursor2_cmp & a):cmp(a.cmp) { } inline bool operator () (const cursor_type & a, const cursor_type & b) { if (UNLIKELY (b.empty ())) return true; // sentinel emulation if (UNLIKELY (a.empty ())) return false; //sentinel emulation return (cmp(a.current (),b.current())); }; }; template < typename block_type, typename run_type, typename input_bid_iterator, typename value_cmp> void create_runs( input_bid_iterator it, run_type ** runs, int nruns, int _m, value_cmp cmp, unsigned _first_element_offset, unsigned _last_element_offset) { typedef typename block_type::value_type type; typedef typename block_type::bid_type bid_type; int m2 = _m / 2; block_manager *bm = block_manager::get_instance(); block_type *Blocks1 = new block_type[m2]; block_type *Blocks2 = new block_type[m2]; request_ptr * read_reqs1 = new request_ptr[m2]; request_ptr * read_reqs2 = new request_ptr[m2]; request_ptr * write_reqs = new request_ptr[m2]; bid_type * bids = new bid_type[m2]; run_type * run; disk_queues::get_instance ()->set_priority_op(disk_queue::WRITE); int i; int k = 0; int run_size = 0,next_run_size=0; assert(nruns >= 2); run = runs[0]; run_size = run->size (); for(i = 0; i < run_size; i++) { bids[i] = *(it++); read_reqs1[i] = Blocks1[i].read(bids[i]); } for (i = _first_element_offset?1:0; i < run_size; i++) bm->delete_block(bids[i]); for(k=0; k < nruns-1; k++) { run = runs[k]; run_size = run->size (); next_run_size = runs[k+1]->size(); if(_last_element_offset && k == nruns - 2) { for(i = 1; i < next_run_size; i++) { bids[i] = *(it++); read_reqs2[i] = Blocks2[i].read(bids[i]); } read_reqs2[0] = Blocks2[0].read(*(it++)); for (i = 1; i < next_run_size ; i++) bm->delete_block(bids[i]); } else { for(i = 0; i < next_run_size; i++) { bids[i] = *(it++); read_reqs2[i] = Blocks2[i].read(bids[i]); } for (i = 0; i < next_run_size; i++) bm->delete_block(bids[i]); } wait_all(read_reqs1, run_size); if(block_type::has_filler) std::sort( TwoToOneDimArrayRowAdaptor< block_type, typename block_type::value_type, block_type::size > (Blocks1,k?0:_first_element_offset ), TwoToOneDimArrayRowAdaptor< block_type, typename block_type::value_type,block_type::size > (Blocks1, run_size*block_type::size ) ,cmp); else std::sort(Blocks1[0].elem + (k?0:_first_element_offset), Blocks1[run_size].elem, cmp); if(k) wait_all(write_reqs, m2); for (i = 0; i < m2; i++) { (*run)[i].value = Blocks1[i][(k?0:_first_element_offset)]; write_reqs[i] = Blocks1[i].write ((*run)[i].bid); } std::swap (Blocks1, Blocks2); std::swap (read_reqs1, read_reqs2); } run = runs[k]; run_size = run->size (); wait_all(read_reqs1, run_size); if(_last_element_offset) { memmove(Blocks1[0].elem + block_type::size - _last_element_offset, Blocks1[0].elem, _last_element_offset); } if(block_type::has_filler) std::sort( TwoToOneDimArrayRowAdaptor< block_type, typename block_type::value_type,block_type::size > (Blocks1, _last_element_offset?(block_type::size - _last_element_offset):0), TwoToOneDimArrayRowAdaptor< block_type, typename block_type::value_type,block_type::size > (Blocks1, run_size*block_type::size ), cmp); else std::sort(Blocks1[0].elem + (_last_element_offset?(block_type::size - _last_element_offset):0), Blocks1[run_size].elem, cmp); wait_all(write_reqs, m2); for (i = 0; i < run_size; i++) { (*run)[i].value = Blocks1[i][_last_element_offset?(block_type::size - _last_element_offset):0]; write_reqs[i] = Blocks1[i].write ((*run)[i].bid); } wait_all(write_reqs, run_size); delete [] Blocks1; delete [] Blocks2; delete [] read_reqs1; delete [] read_reqs2; delete [] write_reqs; delete [] bids; } template < typename block_type,typename run_type , typename value_cmp> void merge_runs(run_type ** in_runs, int nruns, run_type * out_run,unsigned _m,value_cmp cmp, unsigned _first_element_offset,unsigned _last_element_offset) { typedef typename block_type::bid_type bid_type; typedef typename block_type::value_type value_type; typedef block_prefetcher prefetcher_type; typedef run_cursor2 run_cursor_type; typedef run_cursor2_cmp run_cursor2_cmp_type; int i; run_type consume_seq(out_run->size()); int * prefetch_seq = new int[out_run->size()]; typename run_type::iterator copy_start = consume_seq.begin (); for (i = 0; i < nruns; i++) { // TODO: try to avoid copy copy_start = std::copy( in_runs[i]->begin (), in_runs[i]->end (), copy_start ); } std::stable_sort(consume_seq.begin (), consume_seq.end (), trigger_entry_cmp(cmp)); int disks_number = config::get_instance ()->disks_number (); #ifdef PLAY_WITH_OPT_PREF const int n_write_buffers = 4 * disks_number; #else const int n_prefetch_buffers = std::max( 2 * disks_number , (3 * (int(_m) - nruns) / 4)); const int n_write_buffers = std::max( 2 * disks_number , int(_m) - nruns - n_prefetch_buffers ); // heuristic const int n_opt_prefetch_buffers = 2 * disks_number + (3*(n_prefetch_buffers - 2 * disks_number))/10; #endif #ifdef SORT_OPT_PREFETCHING compute_prefetch_schedule( consume_seq, prefetch_seq, n_opt_prefetch_buffers, disks_number ); #else for(i=0;isize();i++) prefetch_seq[i] = i; #endif prefetcher_type prefetcher( consume_seq.begin(), consume_seq.end(), prefetch_seq, nruns + n_prefetch_buffers); buffered_writer writer(n_write_buffers,n_write_buffers/2); int out_run_size = out_run->size (); looser_tree loosers (&prefetcher, nruns,run_cursor2_cmp_type(cmp) ); block_type *out_buffer = writer.get_free_block(); for (i = 0; i < out_run_size; i++) { loosers.multi_merge(out_buffer->elem); (*out_run)[i].value = *(out_buffer->elem); out_buffer = writer.write(out_buffer,(*out_run)[i].bid); } delete [] prefetch_seq; block_manager *bm = block_manager::get_instance (); for (i = 0; i < nruns; i++) { unsigned sz = in_runs[i]->size (); for (unsigned j = 0; j < sz; j++) bm->delete_block ((*in_runs[i])[j].bid); } } template simple_vector< trigger_entry > * sort_blocks( input_bid_iterator input_bids, unsigned _n, unsigned _m, value_cmp cmp, unsigned _first_element_offset = 0, unsigned _last_element_offset = 0) { typedef typename block_type::value_type type; typedef typename block_type::bid_type bid_type; typedef trigger_entry< bid_type,type > trigger_entry_type; typedef simple_vector< trigger_entry_type > run_type; typedef typename interleaved_alloc_traits::strategy interleaved_alloc_strategy; unsigned int m2 = _m / 2; unsigned int full_runs = _n / m2; unsigned int partial_runs = ((_n % m2) ? 1 : 0); unsigned int nruns = full_runs + partial_runs; unsigned int i; config *cfg = config::get_instance(); block_manager *mng = block_manager::get_instance (); const unsigned ndisks = cfg->disks_number (); //STXXL_VERBOSE ("n=" << _n << " nruns=" << nruns << "=" << full_runs << "+" // << partial_runs) #ifdef STXXL_IO_STATS stats *iostats = stats::get_instance(); iostats->reset(); #endif double begin = stxxl_timestamp (), after_runs_creation, end; run_type **runs = new run_type *[nruns]; for (i = 0; i < full_runs; i++) runs[i] = new run_type(m2); if (partial_runs) runs[i] = new run_type (_n - full_runs * m2); for(i=0;inew_blocks( alloc_strategy(0,ndisks), trigger_entry_iterator(runs[i]->begin()), trigger_entry_iterator(runs[i]->end()) ); } create_runs< block_type, run_type, input_bid_iterator, value_cmp > (input_bids, runs, nruns,_m,cmp, _first_element_offset,_last_element_offset); after_runs_creation = stxxl_timestamp (); #ifdef COUNT_WAIT_TIME double io_wait_after_rf = stxxl::wait_time_counter; #endif disk_queues::get_instance ()->set_priority_op (disk_queue::WRITE); // Optimal merging: merge r = pow(nruns,1/ceil(log(nruns)/log(m))) at once const int merge_factor = static_cast(ceil(pow(nruns,1./ceil(log(nruns)/log(_m))))); run_type **new_runs; while(nruns > 1) { int new_nruns = div_and_round_up(nruns,merge_factor); STXXL_VERBOSE("Starting new merge phase: nruns: "< 0) { int runs2merge = STXXL_MIN(runs_left,merge_factor); blocks_in_new_run = 0; for(unsigned i = nruns - runs_left; i < (nruns - runs_left + runs2merge);i++) blocks_in_new_run += runs[i]->size(); // allocate run new_runs[cur_out_run++] = new run_type(blocks_in_new_run); runs_left -= runs2merge; } // allocate blocks for the new runs mng->new_blocks( interleaved_alloc_strategy(new_nruns, 0, ndisks), RunsToBIDArrayAdaptor2 (new_runs,0,new_nruns,blocks_in_new_run), RunsToBIDArrayAdaptor2 (new_runs,_n,new_nruns,blocks_in_new_run)); // merge all runs_left = nruns; cur_out_run = 0; while(runs_left > 0) { int runs2merge = STXXL_MIN(runs_left,merge_factor); STXXL_VERBOSE("Merging "< (runs + nruns - runs_left, runs2merge ,*(new_runs + (cur_out_run++)),_m,cmp, _first_element_offset,_last_element_offset); runs_left -= runs2merge; } nruns = new_nruns; delete [] runs; runs = new_runs; } run_type * result = *runs; delete [] runs; end = stxxl_timestamp (); (void)(begin); STXXL_VERBOSE ("Elapsed time : " << end - begin << " s. Run creation time: " << after_runs_creation - begin << " s") #ifdef STXXL_IO_STATS STXXL_VERBOSE ("reads : " << iostats->get_reads ()) STXXL_VERBOSE ("writes : " << iostats->get_writes ()) STXXL_VERBOSE ("read time : " << iostats->get_read_time () << " s") STXXL_VERBOSE ("write time : " << iostats->get_write_time () <<" s") STXXL_VERBOSE ("parallel read time : " << iostats->get_pread_time () << " s") STXXL_VERBOSE ("parallel write time : " << iostats->get_pwrite_time () << " s") STXXL_VERBOSE ("parallel io time : " << iostats->get_pio_time () << " s") #endif #ifdef COUNT_WAIT_TIME STXXL_VERBOSE ("Time in I/O wait(rf): " << io_wait_after_rf << " s") STXXL_VERBOSE ("Time in I/O wait : " << stxxl::wait_time_counter << " s") #endif return result; } }; //! \brief External sorting routine for records with defined operator < //! \param first object of model of \c ext_random_access_iterator concept //! \param last object of model of \c ext_random_access_iterator concept //! \param cmp comparison object //! \param M amount of buffers for internal use //! \remark Implements external merge sort described in [Dementiev & Sanders'03] //! \remark non-stable template void sort(ExtIterator_ first, ExtIterator_ last,StrictWeakOrdering_ cmp,unsigned M) { typedef simple_vector< sort_local::trigger_entry > run_type; typedef typename ExtIterator_::vector_type::value_type value_type; typedef typename ExtIterator_::block_type block_type; unsigned n=0; block_manager *mng = block_manager::get_instance (); first.flush(); if((last - first)*sizeof(value_type) < M) { stl_in_memory_sort(first,last,cmp); } else { if(first.block_offset()) { if(last.block_offset()) // first and last element are // not first elemetns of their block { typename ExtIterator_::block_type * first_block = new typename ExtIterator_::block_type; typename ExtIterator_::block_type * last_block = new typename ExtIterator_::block_type; typename ExtIterator_::bid_type first_bid,last_bid; request_ptr req; req = first_block->read(*first.bid()); mng->new_blocks( FR(), &first_bid,(&first_bid) + 1); // try to overlap mng->new_blocks( FR(), &last_bid,(&last_bid) + 1); req->wait(); req = last_block->read(*last.bid()); unsigned i=0; for(;ielem[i] = cmp.min_value(); } req->wait(); req = first_block->write(first_bid); for(i=last.block_offset(); i < block_type::size;i++) { last_block->elem[i] = cmp.max_value(); } req->wait(); req = last_block->write(last_bid); n=last.bid() - first.bid() + 1; std::swap(first_bid,*first.bid()); std::swap(last_bid,*last.bid()); req->wait(); delete first_block; delete last_block; run_type * out = sort_local::sort_blocks< typename ExtIterator_::block_type, typename ExtIterator_::vector_type::alloc_strategy, typename ExtIterator_::bids_container_iterator> (first.bid(),n,M/block_type::raw_size,cmp); first_block = new typename ExtIterator_::block_type; last_block = new typename ExtIterator_::block_type; typename ExtIterator_::block_type * sorted_first_block = new typename ExtIterator_::block_type; typename ExtIterator_::block_type * sorted_last_block = new typename ExtIterator_::block_type; request_ptr * reqs = new request_ptr [2]; reqs[0] = first_block->read(first_bid); reqs[1] = sorted_first_block->read((*(out->begin())).bid); wait_all(reqs,2); reqs[0] = last_block->read(last_bid); reqs[1] = sorted_last_block->read( ((*out)[out->size() - 1]).bid); for(i=first.block_offset();ielem[i] = sorted_first_block->elem[i]; } wait_all(reqs,2); req = first_block->write(first_bid); for(i=0;ielem[i] = sorted_last_block->elem[i]; } req->wait(); req = last_block->write(last_bid); mng->delete_block(out->begin()->bid); mng->delete_block((*out)[out->size() - 1].bid); *first.bid() = first_bid; *last.bid() = last_bid; typename run_type::iterator it = out->begin(); it++; typename ExtIterator_::bids_container_iterator cur_bid = first.bid(); cur_bid ++; for(;cur_bid != last.bid(); cur_bid++,it++) { *cur_bid = (*it).bid; } delete first_block; delete sorted_first_block; delete sorted_last_block; delete [] reqs; delete out; req->wait(); delete last_block; } else { // first element is // not the first element of its block typename ExtIterator_::block_type * first_block = new typename ExtIterator_::block_type; typename ExtIterator_::bid_type first_bid; request_ptr req; req = first_block->read(*first.bid()); mng->new_blocks( FR(), &first_bid,(&first_bid) + 1); // try to overlap req->wait(); unsigned i=0; for(;ielem[i] = cmp.min_value(); } req = first_block->write(first_bid); n=last.bid() - first.bid(); std::swap(first_bid,*first.bid()); req->wait(); delete first_block; run_type * out = sort_local::sort_blocks< typename ExtIterator_::block_type, typename ExtIterator_::vector_type::alloc_strategy, typename ExtIterator_::bids_container_iterator > (first.bid(),n,M/block_type::raw_size,cmp); first_block = new typename ExtIterator_::block_type; typename ExtIterator_::block_type * sorted_first_block = new typename ExtIterator_::block_type; request_ptr * reqs = new request_ptr[2]; reqs[0] = first_block->read(first_bid); reqs[1] = sorted_first_block->read((*(out->begin())).bid); wait_all(reqs,2); for(i=first.block_offset();ielem[i] = sorted_first_block->elem[i]; } req = first_block->write(first_bid); mng->delete_block(out->begin()->bid); *first.bid() = first_bid; typename run_type::iterator it = out->begin(); it++; typename ExtIterator_::bids_container_iterator cur_bid = first.bid(); cur_bid ++; for(;cur_bid != last.bid(); cur_bid++,it++) { *cur_bid = (*it).bid; } *cur_bid = (*it).bid; delete sorted_first_block; delete [] reqs; delete out; req->wait(); delete first_block; } } else { if(last.block_offset()) // last is // not the first element of its block { typename ExtIterator_::block_type * last_block = new typename ExtIterator_::block_type; typename ExtIterator_::bid_type last_bid; request_ptr req; unsigned i; req = last_block->read(*last.bid()); mng->new_blocks( FR(), &last_bid,(&last_bid) + 1); req->wait(); for(unsigned i=last.block_offset(); i < block_type::size;i++) { last_block->elem[i] = cmp.max_value(); } req = last_block->write(last_bid); n=last.bid() - first.bid() + 1; std::swap(last_bid,*last.bid()); req->wait(); delete last_block; run_type * out = sort_local::sort_blocks< typename ExtIterator_::block_type, typename ExtIterator_::vector_type::alloc_strategy, typename ExtIterator_::bids_container_iterator> (first.bid(),n,M/block_type::raw_size,cmp); last_block = new typename ExtIterator_::block_type; typename ExtIterator_::block_type * sorted_last_block = new typename ExtIterator_::block_type; request_ptr * reqs = new request_ptr [2]; reqs[0] = last_block->read(last_bid); reqs[1] = sorted_last_block->read( ((*out)[out->size() - 1]).bid); wait_all(reqs,2); for(i=0;ielem[i] = sorted_last_block->elem[i]; } req = last_block->write(last_bid); mng->delete_block((*out)[out->size() - 1].bid); *last.bid() = last_bid; typename run_type::iterator it = out->begin(); typename ExtIterator_::bids_container_iterator cur_bid = first.bid(); for(;cur_bid != last.bid(); cur_bid++,it++) { *cur_bid = (*it).bid; } delete sorted_last_block; delete [] reqs; delete out; req->wait(); delete last_block; } else { // first and last element are first elements of their of blocks n = last.bid() - first.bid(); run_type * out = sort_local::sort_blocks< typename ExtIterator_::block_type, typename ExtIterator_::vector_type::alloc_strategy, typename ExtIterator_::bids_container_iterator > (first.bid(),n,M/block_type::raw_size,cmp); typename run_type::iterator it = out->begin(); typename ExtIterator_::bids_container_iterator cur_bid = first.bid(); for(;cur_bid != last.bid(); cur_bid++,it++) { *cur_bid = (*it).bid; } } } } }; //! \} __STXXL_END_NAMESPACE #endif