OpenFPM  5.2.0
Project that contain the implementation of distributed structures
vector_performance_test.hpp
1 /*
2  * vector_performance_test.hpp
3  *
4  * Created on: Jul 15, 2019
5  * Author: i-bird
6  */
7 
8 #ifndef VECTOR_PERFORMANCE_TEST_HPP_
9 #define VECTOR_PERFORMANCE_TEST_HPP_
10 
11 /*
12  * vector_performance_test.hpp
13  *
14  * Created on: Jan 11, 2016
15  * Author: i-bird
16  */
17 
18 #ifndef OPENFPM_DATA_SRC_VECTOR_VECTOR_PERFORMANCE_TEST_HPP_
19 #define OPENFPM_DATA_SRC_VECTOR_VECTOR_PERFORMANCE_TEST_HPP_
20 
21 #define NADD 128*128*128
22 
23 // Property tree
24 struct report_vector_func_tests
25 {
26  boost::property_tree::ptree graphs;
27 };
28 
29 report_vector_func_tests report_vector_funcs;
30 
31 BOOST_AUTO_TEST_SUITE( vector_performance )
32 
33 BOOST_AUTO_TEST_CASE(vector_performance)
34 {
35  report_vector_funcs.graphs.put("performance.vector(0).funcs.nele",NADD);
36  report_vector_funcs.graphs.put("performance.vector(0).funcs.name","add");
37 
38  report_vector_funcs.graphs.put("performance.vector(1).funcs.nele",NADD);
39  report_vector_funcs.graphs.put("performance.vector(1).funcs.name","get");
40 
41  std::vector<double> times(N_STAT + 1);
42  std::vector<double> times_g(N_STAT + 1);
43 
44  // get test
45  double tot_accu = 0.0;
46 
47  for (size_t i = 0 ; i < N_STAT+1 ; i++)
48  {
49  timer t;
50  t.start();
51 
52  // create a vector
53  openfpm::vector<Point_test<float>> v1;
54 
55  // Point
56  Point_test<float> p;
57  p.setx(1.0);
58  p.sety(2.0);
59  p.setz(3.0);
60  p.sets(4.0);
61 
62  p.get<P::v>()[0] = 1.0;
63  p.get<P::v>()[1] = 2.0;
64  p.get<P::v>()[2] = 7.0;
65 
66  p.get<P::t>()[0][0] = 10.0;
67  p.get<P::t>()[0][1] = 13.0;
68  p.get<P::t>()[0][2] = 8.0;
69  p.get<P::t>()[1][0] = 19.0;
70  p.get<P::t>()[1][1] = 23.0;
71  p.get<P::t>()[1][2] = 5.0;
72  p.get<P::t>()[2][0] = 4.0;
73  p.get<P::t>()[2][1] = 3.0;
74  p.get<P::t>()[2][2] = 11.0;
75 
76  // Add test
77 
78  for (size_t i = 0 ; i < NADD ; i++)
79  {
80  v1.add(p);
81  }
82 
83  t.stop();
84  times[i] = t.getwct();
85 
86  timer tg;
87  tg.start();
88 
89  for (size_t i = 0 ; i < NADD ; i++)
90  {
91  double accu1 = v1.template get<P::x>(i);
92  double accu2 = v1.template get<P::y>(i);
93  double accu3 = v1.template get<P::z>(i);
94  double accu4 = v1.template get<P::s>(i);
95 
96  double accu5 = v1.template get<P::v>(i)[0];
97  double accu6 = v1.template get<P::v>(i)[1];
98  double accu7 = v1.template get<P::v>(i)[2];
99 
100  double accu8 = v1.template get<P::t>(i)[0][0];
101  double accu9 = v1.template get<P::t>(i)[0][1];
102  double accu10 = v1.template get<P::t>(i)[0][2];
103  double accu11 = v1.template get<P::t>(i)[1][0];
104  double accu12 = v1.template get<P::t>(i)[1][1];
105  double accu13 = v1.template get<P::t>(i)[1][2];
106  double accu14 = v1.template get<P::t>(i)[2][0];
107  double accu15 = v1.template get<P::t>(i)[2][1];
108  double accu16 = v1.template get<P::t>(i)[2][2];
109 
110  tot_accu += accu1 + accu2 + accu3 + accu4 + accu5 + accu6 + accu7 + accu8 + accu9 + accu10 + accu11 + accu12 +
111  accu13 + accu14 + accu15 + accu16;
112  }
113 
114  tg.stop();
115 
116  times_g[i] = tg.getwct();
117  }
118 
119  double mean;
120  double dev;
121  standard_deviation(times,mean,dev);
122 
123  report_vector_funcs.graphs.put("performance.vector(0).y.data.mean",mean);
124  report_vector_funcs.graphs.put("performance.vector(0).y.data.dev",dev);
125 
126  standard_deviation(times_g,mean,dev);
127 
128  report_vector_funcs.graphs.put("performance.vector(1).y.data.mean",mean);
129  report_vector_funcs.graphs.put("performance.vector(1).y.data.dev",dev);
130 }
131 
132 template<typename vector_prop_type, typename vector_pos_type>
133 __device__ __host__ void read_write(vector_prop_type & vd_prop, vector_pos_type & vd_pos, unsigned int p)
134 {
135  vd_prop.template get<0>(p) = vd_pos.template get<0>(p)[0] + vd_pos.template get<0>(p)[1];
136 
137  vd_prop.template get<1>(p)[0] = vd_pos.template get<0>(p)[0];
138  vd_prop.template get<1>(p)[1] = vd_pos.template get<0>(p)[1];
139 
140  vd_prop.template get<2>(p)[0][0] = vd_pos.template get<0>(p)[0];
141  vd_prop.template get<2>(p)[0][1] = vd_pos.template get<0>(p)[1];
142  vd_prop.template get<2>(p)[1][0] = vd_pos.template get<0>(p)[0] +
143  vd_pos.template get<0>(p)[1];
144  vd_prop.template get<2>(p)[1][1] = vd_pos.template get<0>(p)[1] -
145  vd_pos.template get<0>(p)[0];
146 
147  vd_pos.template get<0>(p)[0] += 0.01f;
148  vd_pos.template get<0>(p)[1] += 0.01f;
149 }
150 
151 
152 struct ele
153 {
154  double s;
155  double v[2];
156  double t[2][2];
157 };
158 
159 __device__ __host__ void read_write_lin(double * pos, ele * prp, unsigned int p)
160 {
161  prp[p].s = pos[2*p] + pos[2*p+1];
162 
163  prp[p].v[0] = pos[2*p];
164  prp[p].v[1] = pos[2*p+1];
165 
166  prp[p].t[0][0] = pos[2*p];
167  prp[p].t[0][1] = pos[2*p+1];
168  prp[p].t[1][0] = pos[2*p] +
169  pos[2*p+1];
170  prp[p].t[1][1] = pos[2*p+1] -
171  pos[2*p];
172 
173  pos[2*p] += 0.01f;
174  pos[2*p+1] += 0.01f;
175 }
176 
177 __device__ __host__ void read_write_inte(double * pos, double * prp0, double * prp1, double * prp2, unsigned int p, unsigned int n_pos)
178 {
179  prp0[0*n_pos + p] = pos[0*n_pos + p] + pos[1*n_pos+p];
180 
181  prp1[0*n_pos + p] = pos[0*n_pos + p];
182  prp1[1*n_pos + p] = pos[1*n_pos + p];
183 
184  prp2[0*n_pos*2+0*n_pos + p] = pos[0*n_pos + p];
185  prp2[0*n_pos*2+1*n_pos + p] = pos[1*n_pos + p];
186  prp2[1*n_pos*2+0*n_pos + p] = pos[0*n_pos + p] +
187  pos[1*n_pos + p];
188  prp2[1*n_pos*2+1*n_pos + p] = pos[1*n_pos + p] -
189  pos[0*n_pos + p];
190 
191  pos[0*n_pos + p] += 0.01f;
192  pos[1*n_pos + p] += 0.01f;
193 }
194 
195 BOOST_AUTO_TEST_CASE(vector_performance_layout_vs_plain_array)
196 {
197  std::vector<double> times(N_STAT + 1);
198  std::vector<double> times_g(N_STAT + 1);
199 
200  std::vector<double> times2(N_STAT + 1);
201  std::vector<double> times2_g(N_STAT + 1);
202 
203  // get test
204  double tot_accu = 0.0;
205 
206  report_vector_funcs.graphs.put("performance.vector_layout(0).funcs.nele",NADD);
207  report_vector_funcs.graphs.put("performance.vector_layout(0).funcs.name","read_write_lin");
208 
209  for (size_t i = 0 ; i < N_STAT+1 ; i++)
210  {
211  // create a vector
212  openfpm::vector<aggregate<double,double[2],double[2][2]>> v1;
213  openfpm::vector<aggregate<double[2]>> v2;
214 
215  // Point
216  aggregate<double[2]> p;
217  p.get<0>()[0] = 1.0;
218  p.get<0>()[1] = 2.0;
219 
220  aggregate<double,double[2],double[2][2]> pa;
221  pa.get<0>() = 1.0;
222 
223  pa.get<1>()[0] = 1.0;
224  pa.get<1>()[1] = 1.0;
225 
226  pa.get<2>()[0][0] = 1.0;
227  pa.get<2>()[0][1] = 1.0;
228  pa.get<2>()[1][0] = 1.0;
229  pa.get<2>()[1][1] = 1.0;
230 
231  // Add test
232 
233  for (size_t i = 0 ; i < NADD ; i++)
234  {
235  v1.add(pa);
236  v2.add(p);
237  }
238 
239  timer tg;
240  tg.start();
241 
242  for (size_t i = 0 ; i < NADD ; i++)
243  {
244  read_write(v1,v2,i);
245  }
246 
247  tg.stop();
248 
249  times_g[i] = tg.getwct();
250 
251  timer tga;
252  tga.start();
253 
254  double * prp = (double *)v1.getPointer<0>();
255  double * pos = (double *)v2.getPointer<0>();
256 
257  for (size_t i = 0 ; i < NADD ; i++)
258  {
259  read_write_lin(pos,(struct ele *)prp,i);
260  }
261 
262  tga.stop();
263 
264  times[i] = tga.getwct();
265  }
266 
267  double mean;
268  double dev;
269  standard_deviation(times_g,mean,dev);
270 
271  double mean_;
272  double dev_;
273  standard_deviation(times,mean_,dev_);
274 
275  report_vector_funcs.graphs.put("performance.vector_layout(0).y.data.mean",mean_/mean);
276 
277  // Deviation od x/y = x/y^2 dy + 1/y dx
278 
279  report_vector_funcs.graphs.put("performance.vector_layout(0).y.data.dev",mean_/(mean*mean)*dev + dev_ / mean );
280 
281  report_vector_funcs.graphs.put("performance.vector_layout(1).funcs.nele",NADD);
282  report_vector_funcs.graphs.put("performance.vector_layout(1).funcs.name","read_write_inte");
283 
284  for (size_t i = 0 ; i < N_STAT+1 ; i++)
285  {
286  // create a vector
287  openfpm::vector<aggregate<double,double[2],double[2][2]>,HeapMemory,memory_traits_inte> v1;
288  openfpm::vector<aggregate<double[2]>,HeapMemory,memory_traits_inte> v2;
289 
290  // Point
291  aggregate<double[2]> p;
292  p.get<0>()[0] = 1.0;
293  p.get<0>()[1] = 2.0;
294 
295  aggregate<double,double[2],double[2][2]> pa;
296  pa.get<0>() = 1.0;
297 
298  pa.get<1>()[0] = 1.0;
299  pa.get<1>()[1] = 1.0;
300 
301  pa.get<2>()[0][0] = 1.0;
302  pa.get<2>()[0][1] = 1.0;
303  pa.get<2>()[1][0] = 1.0;
304  pa.get<2>()[1][1] = 1.0;
305 
306  // Add test
307 
308  for (size_t i = 0 ; i < NADD ; i++)
309  {
310  v1.add(pa);
311  v2.add(p);
312  }
313 
314  timer tg;
315  tg.start();
316 
317  for (size_t i = 0 ; i < NADD ; i++)
318  {
319  read_write(v1,v2,i);
320  }
321 
322  tg.stop();
323 
324  times2_g[i] = tg.getwct();
325  int sz = v1.size();
326 
327  timer tga;
328  tga.start();
329 
330  double * prp0 = (double *)v1.getPointer<0>();
331  double * prp1 = (double *)v1.getPointer<1>();
332  double * prp2 = (double *)v1.getPointer<2>();
333 
334  double * pos = (double *)v2.getPointer<0>();
335 
336  for (size_t i = 0 ; i < NADD ; i++)
337  {
338  read_write_inte(pos,prp0,prp1,prp2,i,sz);
339  }
340 
341  tga.stop();
342 
343  times2[i] = tga.getwct();
344  }
345 
346  double mean2;
347  double dev2;
348  standard_deviation(times2_g,mean2,dev2);
349 
350  double mean2_;
351  double dev2_;
352  standard_deviation(times2,mean2_,dev2_);
353 
354  report_vector_funcs.graphs.put("performance.vector_layout(1).y.data.mean",mean2_/mean2);
355 
356  // Deviation od x/y = x/y^2 dy + 1/y dx
357 
358  report_vector_funcs.graphs.put("performance.vector_layout(1).y.data.dev",mean2_/(mean2*mean2)*dev2 + dev2_ / mean2 );
359 }
360 
362 
363 BOOST_AUTO_TEST_CASE(vector_performance_write_report)
364 {
365  // Create a graphs
366 
367  report_vector_funcs.graphs.put("graphs.graph(0).type","line");
368  report_vector_funcs.graphs.add("graphs.graph(0).title","Vector add and get");
369  report_vector_funcs.graphs.add("graphs.graph(0).x.title","Tests");
370  report_vector_funcs.graphs.add("graphs.graph(0).y.title","Time seconds");
371  report_vector_funcs.graphs.add("graphs.graph(0).y.data(0).source","performance.vector(#).y.data.mean");
372  report_vector_funcs.graphs.add("graphs.graph(0).x.data(0).source","performance.vector(#).funcs.name");
373  report_vector_funcs.graphs.add("graphs.graph(0).y.data(0).title","Actual");
374  report_vector_funcs.graphs.add("graphs.graph(0).interpolation","lines");
375 
376  report_vector_funcs.graphs.put("graphs.graph(1).type","line");
377  report_vector_funcs.graphs.add("graphs.graph(1).title","Vector read write");
378  report_vector_funcs.graphs.add("graphs.graph(1).x.title","Layout");
379  report_vector_funcs.graphs.add("graphs.graph(1).y.title","Time seconds");
380  report_vector_funcs.graphs.add("graphs.graph(1).y.data(0).source","performance.vector_layout(#).y.data.mean");
381  report_vector_funcs.graphs.add("graphs.graph(1).x.data(0).source","performance.vector_layout(#).funcs.name");
382  report_vector_funcs.graphs.add("graphs.graph(1).y.data(0).title","Actual");
383  report_vector_funcs.graphs.add("graphs.graph(1).interpolation","lines");
384 
385  boost::property_tree::xml_writer_settings<std::string> settings(' ', 4);
386  boost::property_tree::write_xml("vector_performance_funcs.xml", report_vector_funcs.graphs,std::locale(),settings);
387 
388  GoogleChart cg;
389 
390  std::string file_xml_ref(test_dir);
391  file_xml_ref += std::string("/openfpm_data/vector_performance_funcs_ref.xml");
392 
393  StandardXMLPerformanceGraph("vector_performance_funcs.xml",file_xml_ref,cg);
394 
395  addUpdtateTime(cg,1,"data","vector_performance_funcs");
396 
397  cg.write("vector_performance_funcs.html");
398 }
399 
400 BOOST_AUTO_TEST_SUITE_END()
401 
402 #endif /* OPENFPM_DATA_SRC_VECTOR_VECTOR_PERFORMANCE_TEST_HPP_ */
403 
404 
405 
406 #endif /* VECTOR_PERFORMANCE_TEST_HPP_ */
GoogleChart
Small class to produce graph with Google chart in HTML.
Definition: GoogleChart.hpp:216
GoogleChart::write
void write(std::string file)
It write the graphs on file in html format using Google charts.
Definition: GoogleChart.hpp:959
HeapMemory
This class allocate, and destroy CPU memory.
Definition: HeapMemory.hpp:40
Point_test
Test structure used for several test.
Definition: Point_test.hpp:106
Point_test::v
static const unsigned int v
v property is at position 4 in the boost::fusion::vector
Definition: Point_test.hpp:146
Point_test::t
static const unsigned int t
t property is at position 5 in the boost::fusion::vector
Definition: Point_test.hpp:149
openfpm::vector
Implementation of 1-D std::vector like structure.
Definition: map_vector.hpp:204
openfpm::vector::size
size_t size()
Stub size.
Definition: map_vector.hpp:212
timer
Class for cpu time benchmarking.
Definition: timer.hpp:28
timer::stop
void stop()
Stop the timer.
Definition: timer.hpp:119
timer::start
void start()
Start the timer.
Definition: timer.hpp:90
timer::getwct
double getwct()
Return the elapsed real time.
Definition: timer.hpp:130
aggregate
aggregate of properties, from a list of object if create a struct that follow the OPENFPM native stru...
Definition: aggregate.hpp:221
aggregate::get
__device__ __host__ boost::mpl::at< type, boost::mpl::int_< i > >::type & get()
get the properties i
Definition: aggregate.hpp:249
memory_traits_inte
Transform the boost::fusion::vector into memory specification (memory_traits)
Definition: memory_conf.hpp:84
report_vector_func_tests
Definition: vector_performance_test.cu:24