Fork me on GitHub

source: git/external/fastjet/contribs/SoftKiller/SoftKiller.cc@ c9803f7

ImprovedOutputFile Timing dual_readout llp
Last change on this file since c9803f7 was 35cdc46, checked in by Pavel Demin <demin@…>, 10 years ago

upgrade FastJet to version 3.1.0-beta.1, upgrade Nsubjettiness to version 2.1.0, add SoftKiller version 1.0.0

  • Property mode set to 100644
File size: 12.1 KB
Line 
1// $Id$
2//
3// Copyright (c) 2014-, Matteo Cacciari, Gavin. P. Salam and Gregory Soyez
4//
5//----------------------------------------------------------------------
6// This file is part of FastJet contrib.
7//
8// It is free software; you can redistribute it and/or modify it under
9// the terms of the GNU General Public License as published by the
10// Free Software Foundation; either version 2 of the License, or (at
11// your option) any later version.
12//
13// It is distributed in the hope that it will be useful, but WITHOUT
14// ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
15// or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public
16// License for more details.
17//
18// You should have received a copy of the GNU General Public License
19// along with this code. If not, see <http://www.gnu.org/licenses/>.
20//----------------------------------------------------------------------
21
22#include "SoftKiller.hh"
23#include <sstream>
24
25using namespace std;
26
27FASTJET_BEGIN_NAMESPACE // defined in fastjet/internal/base.hh
28
29namespace contrib{
30
31/// ctor with simple initialisation
32/// \param rapmax the maximal absolute rapidity extent of the grid
33/// \param cell_size the grid spacing (equivalently, cell size)
34/// \param sifter when provided, the soft killer is applied
35/// only to particles that pass the sifter (the
36/// others are kept untouched)
37SoftKiller::SoftKiller(double rapmax, double cell_size,
38 Selector sifter) :
39#ifdef FJCONTRIB_SOFTKILLER_USEFJGRID
40 RectangularGrid(rapmax, cell_size), _sifter(sifter) {}
41#else // not FJCONTRIB_SOFTKILLER_USEFJGRID
42 _ymax(rapmax), _ymin(-rapmax),
43 _requested_drap(cell_size), _requested_dphi(cell_size),
44 _sifter(sifter) {
45 _setup_grid();
46}
47#endif
48
49/// ctor with more control over initialisation
50/// \param rapmin the minimum rapidity extent of the grid
51/// \param rapmax the maximum rapidity extent of the grid
52/// \param drap the grid spacing in rapidity
53/// \param dphi the grid spacing in azimuth
54/// \param sifter when provided, the soft killer is applied
55/// onyl to particles that pass the sifter (the
56/// others are kept untouched)
57SoftKiller::SoftKiller(double rapmin, double rapmax, double drap, double dphi,
58 Selector sifter) :
59#ifdef FJCONTRIB_SOFTKILLER_USEFJGRID
60 RectangularGrid(rapmin, rapmax, drap, dphi), _sifter(sifter) {}
61#else
62 _ymax(rapmax), _ymin(rapmin),
63 _requested_drap(drap), _requested_dphi(dphi),
64 _sifter(sifter) {
65 _setup_grid();
66}
67#endif
68
69#ifdef FJCONTRIB_SOFTKILLER_USEFJGRID
70SoftKiller::SoftKiller(const RectangularGrid & grid, Selector sifter) :
71 RectangularGrid(grid), _sifter(sifter) {}
72#endif
73
74/// dummy ctor (will give an unusable SoftKiller)
75SoftKiller::SoftKiller()
76#ifdef FJCONTRIB_SOFTKILLER_USEFJGRID
77 {}
78#else
79 : _ymax(-1.0), _ymin(1.0), _requested_drap(-1.0), _requested_dphi(-1.0) {
80 _ntotal = 0;
81}
82#endif // FJCONTRIB_SOFTKILLER_USEFJGRID
83
84
85//------------------------------------------------------------------------
86// description of the soft killer
87std::string SoftKiller::description() const{
88 ostringstream oss;
89
90
91#ifdef FJCONTRIB_SOFTKILLER_USEFJGRID
92 oss << "SoftKiller with " << RectangularGrid::description();
93#else
94 if (_requested_drap < 0 || _requested_dphi < 0)
95 return "Uninitialised SoftKiller";
96
97 oss << "SoftKiller with rapidity extent " << _ymin << " < rap < " << _ymax
98 << ", cell size drap x dphi = " << _dy << " x " << _dphi;
99#endif
100 if (_sifter.worker()) {
101 oss << " and applied to particles passing the selection ("
102 << _sifter.description() << ")";
103 }
104 return oss.str();
105}
106
107//------------------------------------------------------------------------
108// similarly to Transformers in FastJet, introduce a 'result'
109// method equivalent to the () operator, i.e. returns the event
110// after the soft killer has been applied
111//vector<PseudoJet> SoftKiller::result(const vector<PseudoJet> & event) const {
112void SoftKiller::apply(const vector<PseudoJet> & event,
113 vector<PseudoJet> & reduced_event,
114 double & pt_threshold) const {
115 // a safety check: we impose at least 2 cells (otherwise, this is
116 // equivalent to asking an empty event and there are more
117 // efficient ways to do that)
118 if (n_tiles()<2){
119 throw Error("SoftKiller not properly initialised.");
120 }
121
122 // we're not set up to handle the case where the event and reduced
123 // event are the same vector; so crash in that case.
124 assert(&event != &reduced_event);
125 // currently we can only handle the case where all tiles have equal
126 // area; that is the case for Rectangular tilings, but in the future
127 // one might imagine having non-rectangular tilings.
128 assert(all_tiles_equal_area());
129
130 //profiling: CPUTimer t;
131 //profiling: t.start();
132
133 // init an array to hold the max pt in each grid cell
134 //
135 // This is better (stack) but only C99 (fails with pedantic):
136 // double max_pt2[n_tiles()];
137 // See
138 // http://binglongx.wordpress.com/2011/05/08/create-variable-length-array-on-the-stack-in-c/
139 // for a possible workaround
140 vector<double> max_pt2(n_tiles(), 0.0);
141 //double max_pt2[n_tiles()];
142 //memset(max_pt2, 0, n_tiles()*sizeof(double));
143 //double *max_pt2 = new double[n_tiles()]; // if this is reinstated, reinstate also the delete, below
144 //for (int i = 0; i < n_tiles(); i++) {max_pt2[i] = 0;}
145
146 //profiling: t.stop();
147 //profiling: cout << " copy ptrs: " << t.total() << endl;
148 //profiling: t.start();
149
150 // leave away particles that are sifted
151 // vector<PseudoJet> reduced_event, saved_particles;
152 // _sifter.sift(event, reduced_event, saved_particles);
153
154 vector<const PseudoJet *> event_ptrs(event.size());
155 for (unsigned i = 0; i < event.size(); i++) {
156 event_ptrs[i] = & event[i];
157 }
158 // only run the sifter if it serves a purpose
159 if (_sifter.worker()) _sifter.nullify_non_selected(event_ptrs);
160
161 //profiling: t.stop();
162 //profiling: cout << " sifter : " << t.total() << endl;
163 //profiling: t.start();
164
165 //vector<PseudoJet> reduced_event = event;
166
167 // browse the particles and figure which is the min pt in each cell
168 for(unsigned int i=0;i<event.size(); i++){
169 if (event_ptrs[i] == 0) continue;
170 int idx = tile_index(event[i]);
171 if (idx<0) continue;
172 max_pt2[idx] = max(max_pt2[idx], event[i].pt2());
173 }
174
175 //profiling: t.stop();
176 //profiling: cout << " browsing : " << t.total() << endl;
177 //profiling: t.start();
178
179
180 // if there are some "bad" tiles, then we need to exclude them from
181 // the calculation of the median. We'll do this by condensing the
182 // max_pt2 vector down to just the values for the tiles that are
183 // good.
184 //
185 // tested answers look right in "issue" 2014-08-08-testing-rect-grid
186 if (n_good_tiles() != n_tiles()) {
187 int newn = 0;
188 for (unsigned i = 0; i < max_pt2.size(); i++) {
189 if (tile_is_good(i)) {
190 // clang gets confused with the SharedPtr swap if we don't
191 // have std:: here
192 std::swap(max_pt2[i],max_pt2[newn]);
193 newn++;
194 }
195 }
196 max_pt2.resize(newn);
197 }
198
199 // sort the list of max values (sort works with arrays)
200 //sort(max_pt2, max_pt2+n_tiles()); // use this one if we have the allocated C-style array above
201 sort(max_pt2.begin(), max_pt2.end());
202
203 // Note that if we ask that half of the event is empty that means
204 // that we need at least half of the cells empty... so we need to
205 // round up
206 //
207 // For a potentially faster median search, see
208 //
209 // http://ndevilla.free.fr/median/median/index.html?utm_source=feedburner&utm_medium=twitter&utm_campaign=Feed%3A+hnycombinator+%28HN+-+hnycombinator%29
210 int int_median_pos = max_pt2.size()/2;
211
212 //--------------------------------------------------------------
213 // alternative to what is below:
214 //
215 // apply the pt cut manually (always on pt^2): first get a vector of
216 // indices that we'll want to keep and then when we know the size of
217 // the resulting vector we actually do the transfer of the
218 // PseudoJets.
219 //
220 // This is almost a factor of two faster than doing a
221 // push_back on the PseudoJets themselves (which is logical, since a
222 // push_back probably averages out as doing the copy twice).
223 // (2014-08-11, also tried it with pointers, which seemed
224 // marginally slower).
225 double pt2cut = (1+1e-12)*max_pt2[int_median_pos];
226
227 vector<int> indices;
228 for(unsigned int i=0;i<event.size(); i++){
229 if ((event_ptrs[i] == 0) || (event[i].pt2() >= pt2cut))
230 indices.push_back(i);
231 }
232 reduced_event.reserve(indices.size());
233 for(unsigned int i=0;i<indices.size(); i++){
234 reduced_event.push_back(event[indices[i]]);
235 }
236
237 // //vector<PseudoJet> reduced_event;
238 // reduced_event.clear();
239 // for(unsigned int i=0;i<event.size(); i++){
240 // if ((event_ptrs[i] == 0) ||
241 // (event[i].pt2() >= pt2cut))
242 // reduced_event.push_back(event[i]);
243 // }
244
245 // free memory: reinstate this is max_pt2 becomes an allocated variable again
246 //delete max_pt2;
247
248 //return reduced_event;
249 pt_threshold = sqrt(pt2cut);
250
251 // end of alternative
252 //--------------------------------------------------------------
253
254 // double median_maxpt = sqrt(max_pt2[int_median_pos]);
255 //
256 // //profiling: t.stop();
257 // //profiling: cout << " median : " << t.total() << endl;
258 // //profiling: t.start();
259 //
260 // // apply a cut on pt using a selector
261 // //
262 // // Watch out that the Selector checks pt >= ptcut, and, since it
263 // // uses pt2 to perform the comparison, may lead to rounding
264 // // errors. By multiplying the ptcut by a small amount, we make
265 // // sure the particle with pt=ptcut is killed.
266 // //
267 // // We're actually going to set to null the ones that will be kept
268 // SelectorPtMax((1+1e-12)*median_maxpt).nullify_non_selected(event_ptrs);
269 //
270 // // basic information
271 // _ptcut = median_maxpt; // a good first start
272 //
273 // //profiling: t.stop();
274 // //profiling: cout << " cutting : " << t.total() << endl;
275 // //profiling: t.start();
276 //
277 // // then put back in the saved particles
278 // //
279 // // Note that here we may want to use the killer independently on
280 // // the 2 sets of particles but then the best woulb be to move most
281 // // of the above in a separate method and call it twice (left for
282 // // future experimentation)
283 // //copy(saved_particles.begin(), saved_particles.end(), back_inserter(reduced_event));
284 //
285 // vector<PseudoJet> reduced_event;
286 // // no sizeable speed gain: reduced_event.reserve(event.size());
287 // for (unsigned int i=0;i<event.size();i++)
288 // if (event_ptrs[i] == 0) reduced_event.push_back(event[i]);
289 //
290 // // free memory
291 // delete max_pt2;
292 //
293 // //profiling: t.stop();
294 // //profiling: cout << " finishing: " << t.total() << endl;
295 //
296 // return reduced_event;
297}
298
299#ifndef FJCONTRIB_SOFTKILLER_USEFJGRID
300
301// configure the grid
302void SoftKiller::_setup_grid() {
303 // this grid-definition code is becoming repetitive -- it should
304 // probably be moved somewhere central...
305 double ny_double = (_ymax-_ymin) / _requested_drap;
306 _ny = max(int(ny_double+0.5),1);
307 _dy = (_ymax-_ymin) / _ny;
308 _inverse_dy = _ny/(_ymax-_ymin);
309
310 _nphi = int (twopi / _requested_dphi + 0.5);
311 _dphi = twopi / _nphi;
312 _inverse_dphi = _nphi/twopi;
313
314 // some sanity checking (could throw a fastjet::Error)
315 assert(_ny >= 1 && _nphi >= 1);
316
317 _ntotal = _nphi * _ny;
318 //_max_pt.resize(_ntotal);
319 _cell_area = _dy * _dphi;
320}
321
322// retrieve the grid cell index for a given PseudoJet
323inline int SoftKiller::tile_index(const PseudoJet & p) const {
324 // writing it as below gives a huge speed gain (factor two!). Even
325 // though answers are identical and the routine here is not the
326 // speed-critical step. It's not at all clear why.
327 int iy = int(floor( (p.rap() - _ymin) * _inverse_dy ));
328 if (iy < 0 || iy >= _ny) return -1;
329
330 int iphi = int( p.phi() * _inverse_dphi );
331 //assert(iphi >= 0 && iphi <= _nphi);
332 if (iphi == _nphi) iphi = 0; // just in case of rounding errors
333
334 //int igrid_res = iy*_nphi + iphi;
335 //assert (igrid_res >= 0 && igrid_res < _ny*_nphi);
336 return iy*_nphi + iphi; //igrid_res;
337}
338
339#endif // not FJCONTRIB_SOFTKILLER_USEFJGRID
340
341
342
343} // namespace contrib
344
345FASTJET_END_NAMESPACE
Note: See TracBrowser for help on using the repository browser.