summaryrefslogtreecommitdiff
path: root/src/mem/ruby/profiler/Profiler.cc
blob: c6fbd1aa4c64aa46cb8013f5488ad5ec66449935 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
/*
 * Copyright (c) 1999-2008 Mark D. Hill and David A. Wood
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are
 * met: redistributions of source code must retain the above copyright
 * notice, this list of conditions and the following disclaimer;
 * redistributions in binary form must reproduce the above copyright
 * notice, this list of conditions and the following disclaimer in the
 * documentation and/or other materials provided with the distribution;
 * neither the name of the copyright holders nor the names of its
 * contributors may be used to endorse or promote products derived from
 * this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */

/*
   This file has been modified by Kevin Moore and Dan Nussbaum of the
   Scalable Systems Research Group at Sun Microsystems Laboratories
   (http://research.sun.com/scalable/) to support the Adaptive
   Transactional Memory Test Platform (ATMTP).

   Please send email to atmtp-interest@sun.com with feedback, questions, or
   to request future announcements about ATMTP.

   ----------------------------------------------------------------------

   File modification date: 2008-02-23

   ----------------------------------------------------------------------
*/

/*
 * Profiler.cc
 *
 * Description: See Profiler.hh
 *
 * $Id$
 *
 */

#include "mem/ruby/profiler/Profiler.hh"
#include "mem/ruby/profiler/AddressProfiler.hh"
#include "mem/ruby/system/System.hh"
#include "mem/ruby/network/Network.hh"
#include "mem/gems_common/PrioHeap.hh"
#include "mem/protocol/CacheMsg.hh"
#include "mem/protocol/Protocol.hh"
#include "mem/gems_common/util.hh"
#include "mem/gems_common/Map.hh"
#include "mem/ruby/common/Debug.hh"
#include "mem/protocol/MachineType.hh"

#include "mem/ruby/system/System.hh"

// Allows use of times() library call, which determines virtual runtime
#include <sys/times.h>

extern std::ostream * debug_cout_ptr;

static double process_memory_total();
static double process_memory_resident();

Profiler::Profiler(const Params *p)
    : SimObject(p)
{
  m_requestProfileMap_ptr = new Map<string, int>;

  m_inst_profiler_ptr = NULL;
  m_address_profiler_ptr = NULL;

  m_real_time_start_time = time(NULL); // Not reset in clearStats()
  m_stats_period = 1000000; // Default
  m_periodic_output_file_ptr = &cerr;

  m_hot_lines = p->hot_lines;
  m_all_instructions = p->all_instructions;

  RubySystem::m_profiler_ptr = this;
}

Profiler::~Profiler()
{
  if (m_periodic_output_file_ptr != &cerr) {
    delete m_periodic_output_file_ptr;
  }
  delete m_requestProfileMap_ptr;
}

void Profiler::init(const vector<string> & argv, vector<string> memory_control_names)
{
  // added by SS
  vector<string>::iterator it;
  memory_control_profiler* mcp;
  m_memory_control_names = memory_control_names;
//  printf ( "Here in Profiler::init \n");
  for ( it=memory_control_names.begin() ; it < memory_control_names.end(); it++ ){
//    printf ( "Here in Profiler::init memory control name %s \n", (*it).c_str());
    mcp = new memory_control_profiler;
    mcp->m_memReq = 0;
    mcp->m_memBankBusy = 0;
    mcp->m_memBusBusy = 0;
    mcp->m_memReadWriteBusy = 0;
    mcp->m_memDataBusBusy = 0;
    mcp->m_memTfawBusy = 0;
    mcp->m_memRefresh = 0;
    mcp->m_memRead = 0;
    mcp->m_memWrite = 0;
    mcp->m_memWaitCycles = 0;
    mcp->m_memInputQ = 0;
    mcp->m_memBankQ = 0;
    mcp->m_memArbWait = 0;
    mcp->m_memRandBusy = 0;
    mcp->m_memNotOld = 0;

    mcp->m_banks_per_rank = RubySystem::getMemoryControl((*it).c_str())->getBanksPerRank();
    mcp->m_ranks_per_dimm = RubySystem::getMemoryControl((*it).c_str())->getRanksPerDimm();
    mcp->m_dimms_per_channel = RubySystem::getMemoryControl((*it).c_str())->getDimmsPerChannel();

    int totalBanks = mcp->m_banks_per_rank
                 * mcp->m_ranks_per_dimm
                 * mcp->m_dimms_per_channel;

    mcp->m_memBankCount.setSize(totalBanks);

    m_memory_control_profilers [(*it).c_str()] = mcp;
  }

  clearStats();
  m_hot_lines = false;
  m_all_instructions = false;

  m_address_profiler_ptr = new AddressProfiler;
  m_address_profiler_ptr -> setHotLines(m_hot_lines);
  m_address_profiler_ptr -> setAllInstructions(m_all_instructions);

  if (m_all_instructions) {
    m_inst_profiler_ptr = new AddressProfiler;
    m_inst_profiler_ptr -> setHotLines(m_hot_lines);
    m_inst_profiler_ptr -> setAllInstructions(m_all_instructions);
  }
}

void Profiler::wakeup()
{
  // FIXME - avoid the repeated code

  Vector<integer_t> perProcCycleCount;
  perProcCycleCount.setSize(RubySystem::getNumberOfSequencers());

  for(int i=0; i < RubySystem::getNumberOfSequencers(); i++) {
    perProcCycleCount[i] = g_system_ptr->getCycleCount(i) - m_cycles_executed_at_start[i] + 1;
    // The +1 allows us to avoid division by zero
  }

  integer_t total_misses = m_perProcTotalMisses.sum();
  integer_t simics_cycles_executed = perProcCycleCount.sum();
  integer_t transactions_started = m_perProcStartTransaction.sum();
  integer_t transactions_ended = m_perProcEndTransaction.sum();

  (*m_periodic_output_file_ptr) << "ruby_cycles: " << g_eventQueue_ptr->getTime()-m_ruby_start << endl;
  (*m_periodic_output_file_ptr) << "total_misses: " << total_misses << " " << m_perProcTotalMisses << endl;
  (*m_periodic_output_file_ptr) << "simics_cycles_executed: " << simics_cycles_executed << " " << perProcCycleCount << endl;
  (*m_periodic_output_file_ptr) << "transactions_started: " << transactions_started << " " << m_perProcStartTransaction << endl;
  (*m_periodic_output_file_ptr) << "transactions_ended: " << transactions_ended << " " << m_perProcEndTransaction << endl;
  (*m_periodic_output_file_ptr) << "mbytes_resident: " << process_memory_resident() << endl;
  (*m_periodic_output_file_ptr) << "mbytes_total: " << process_memory_total() << endl;
  if (process_memory_total() > 0) {
    (*m_periodic_output_file_ptr) << "resident_ratio: " << process_memory_resident()/process_memory_total() << endl;
  }
  (*m_periodic_output_file_ptr) << "miss_latency: " << m_allMissLatencyHistogram << endl;

  *m_periodic_output_file_ptr << endl;

  if (m_all_instructions) {
    m_inst_profiler_ptr->printStats(*m_periodic_output_file_ptr);
  }

  //g_system_ptr->getNetwork()->printStats(*m_periodic_output_file_ptr);
  g_eventQueue_ptr->scheduleEvent(this, m_stats_period);
}

void Profiler::setPeriodicStatsFile(const string& filename)
{
  cout << "Recording periodic statistics to file '" << filename << "' every "
       << m_stats_period << " Ruby cycles" << endl;

  if (m_periodic_output_file_ptr != &cerr) {
    delete m_periodic_output_file_ptr;
  }

  m_periodic_output_file_ptr = new ofstream(filename.c_str());
  g_eventQueue_ptr->scheduleEvent(this, 1);
}

void Profiler::setPeriodicStatsInterval(integer_t period)
{
  cout << "Recording periodic statistics every " << m_stats_period << " Ruby cycles" << endl;
  m_stats_period = period;
  g_eventQueue_ptr->scheduleEvent(this, 1);
}

void Profiler::printConfig(ostream& out) const
{
  out << endl;
  out << "Profiler Configuration" << endl;
  out << "----------------------" << endl;
  out << "periodic_stats_period: " << m_stats_period << endl;
}

void Profiler::print(ostream& out) const
{
  out << "[Profiler]";
}

void Profiler::printStats(ostream& out, bool short_stats)
{
  out << endl;
  if (short_stats) {
    out << "SHORT ";
  }
  out << "Profiler Stats" << endl;
  out << "--------------" << endl;

  time_t real_time_current = time(NULL);
  double seconds = difftime(real_time_current, m_real_time_start_time);
  double minutes = seconds/60.0;
  double hours = minutes/60.0;
  double days = hours/24.0;
  Time ruby_cycles = g_eventQueue_ptr->getTime()-m_ruby_start;

  if (!short_stats) {
    out << "Elapsed_time_in_seconds: " << seconds << endl;
    out << "Elapsed_time_in_minutes: " << minutes << endl;
    out << "Elapsed_time_in_hours: " << hours << endl;
    out << "Elapsed_time_in_days: " << days << endl;
    out << endl;
  }

  // print the virtual runtimes as well
  struct tms vtime;
  times(&vtime);
  seconds = (vtime.tms_utime + vtime.tms_stime) / 100.0;
  minutes = seconds / 60.0;
  hours = minutes / 60.0;
  days = hours / 24.0;
  out << "Virtual_time_in_seconds: " << seconds << endl;
  out << "Virtual_time_in_minutes: " << minutes << endl;
  out << "Virtual_time_in_hours:   " << hours << endl;
  out << "Virtual_time_in_days:    " << days << endl;
  out << endl;

  out << "Ruby_current_time: " << g_eventQueue_ptr->getTime() << endl;
  out << "Ruby_start_time: " << m_ruby_start << endl;
  out << "Ruby_cycles: " << ruby_cycles << endl;
  out << endl;

  if (!short_stats) {
    out << "mbytes_resident: " << process_memory_resident() << endl;
    out << "mbytes_total: " << process_memory_total() << endl;
    if (process_memory_total() > 0) {
      out << "resident_ratio: " << process_memory_resident()/process_memory_total() << endl;
    }
    out << endl;

  }

  Vector<integer_t> perProcCycleCount;
  Vector<double> perProcCyclesPerTrans;
  Vector<double> perProcMissesPerTrans;


  perProcCycleCount.setSize(RubySystem::getNumberOfSequencers());
  perProcCyclesPerTrans.setSize(RubySystem::getNumberOfSequencers());
  perProcMissesPerTrans.setSize(RubySystem::getNumberOfSequencers());

  for(int i=0; i < RubySystem::getNumberOfSequencers(); i++) {
    perProcCycleCount[i] = g_system_ptr->getCycleCount(i) - m_cycles_executed_at_start[i] + 1;
    // The +1 allows us to avoid division by zero

    int trans = m_perProcEndTransaction[i];
    if (trans == 0) {
      perProcCyclesPerTrans[i] = 0;
      perProcMissesPerTrans[i] = 0;
    } else {
      perProcCyclesPerTrans[i] = ruby_cycles / double(trans);
      perProcMissesPerTrans[i] = m_perProcTotalMisses[i] / double(trans);
    }
  }

  integer_t total_misses = m_perProcTotalMisses.sum();
  integer_t user_misses = m_perProcUserMisses.sum();
  integer_t supervisor_misses = m_perProcSupervisorMisses.sum();
  integer_t simics_cycles_executed = perProcCycleCount.sum();
  integer_t transactions_started = m_perProcStartTransaction.sum();
  integer_t transactions_ended = m_perProcEndTransaction.sum();

  double cycles_per_transaction = (transactions_ended != 0) ? (RubySystem::getNumberOfSequencers() * double(ruby_cycles)) / double(transactions_ended) : 0;
  double misses_per_transaction = (transactions_ended != 0) ? double(total_misses) / double(transactions_ended) : 0;

  out << "Total_misses: " << total_misses << endl;
  out << "total_misses: " << total_misses << " " << m_perProcTotalMisses << endl;
  out << "user_misses: " << user_misses << " " << m_perProcUserMisses << endl;
  out << "supervisor_misses: " << supervisor_misses << " " << m_perProcSupervisorMisses << endl;
  out << endl;
  out << "ruby_cycles_executed: " << simics_cycles_executed << " " << perProcCycleCount << endl;
  out << endl;
  out << "transactions_started: " << transactions_started << " " << m_perProcStartTransaction << endl;
  out << "transactions_ended: " << transactions_ended << " " << m_perProcEndTransaction << endl;
  out << "cycles_per_transaction: " << cycles_per_transaction  << " " << perProcCyclesPerTrans << endl;
  out << "misses_per_transaction: " << misses_per_transaction << " " << perProcMissesPerTrans << endl;

  out << endl;

  out << endl;

  vector<string>::iterator it;

  for ( it=m_memory_control_names.begin() ; it < m_memory_control_names.end(); it++ ){
    long long int m_memReq = m_memory_control_profilers[(*it).c_str()] -> m_memReq;
    long long int m_memRefresh = m_memory_control_profilers[(*it).c_str()] -> m_memRefresh;
    long long int m_memInputQ = m_memory_control_profilers[(*it).c_str()] -> m_memInputQ;
    long long int m_memBankQ = m_memory_control_profilers[(*it).c_str()] -> m_memBankQ;
    long long int m_memWaitCycles = m_memory_control_profilers[(*it).c_str()] -> m_memWaitCycles;
    long long int m_memRead = m_memory_control_profilers[(*it).c_str()] -> m_memRead;
    long long int m_memWrite = m_memory_control_profilers[(*it).c_str()] -> m_memWrite;
    long long int m_memBankBusy = m_memory_control_profilers[(*it).c_str()] -> m_memBankBusy;
    long long int m_memRandBusy = m_memory_control_profilers[(*it).c_str()] -> m_memRandBusy;
    long long int m_memNotOld = m_memory_control_profilers[(*it).c_str()] -> m_memNotOld;
    long long int m_memArbWait = m_memory_control_profilers[(*it).c_str()] -> m_memArbWait;
    long long int m_memBusBusy = m_memory_control_profilers[(*it).c_str()] -> m_memBusBusy;
    long long int m_memTfawBusy = m_memory_control_profilers[(*it).c_str()] -> m_memTfawBusy;
    long long int m_memReadWriteBusy = m_memory_control_profilers[(*it).c_str()] -> m_memReadWriteBusy;
    long long int m_memDataBusBusy = m_memory_control_profilers[(*it).c_str()] -> m_memDataBusBusy;
    Vector<long long int> m_memBankCount = m_memory_control_profilers[(*it).c_str()] -> m_memBankCount;

    if (m_memReq || m_memRefresh) {    // if there's a memory controller at all
      long long int total_stalls = m_memInputQ + m_memBankQ + m_memWaitCycles;
      double stallsPerReq = total_stalls * 1.0 / m_memReq;
      out << "Memory control " << (*it) << ":" << endl;
      out << "  memory_total_requests: " << m_memReq << endl;  // does not include refreshes
      out << "  memory_reads: " << m_memRead << endl;
      out << "  memory_writes: " << m_memWrite << endl;
      out << "  memory_refreshes: " << m_memRefresh << endl;
      out << "  memory_total_request_delays: " << total_stalls << endl;
      out << "  memory_delays_per_request: " << stallsPerReq << endl;
      out << "  memory_delays_in_input_queue: " << m_memInputQ << endl;
      out << "  memory_delays_behind_head_of_bank_queue: " << m_memBankQ << endl;
      out << "  memory_delays_stalled_at_head_of_bank_queue: " << m_memWaitCycles << endl;
      // Note:  The following "memory stalls" entries are a breakdown of the
      // cycles which already showed up in m_memWaitCycles.  The order is
      // significant; it is the priority of attributing the cycles.
      // For example, bank_busy is before arbitration because if the bank was
      // busy, we didn't even check arbitration.
      // Note:  "not old enough" means that since we grouped waiting heads-of-queues
      // into batches to avoid starvation, a request in a newer batch
      // didn't try to arbitrate yet because there are older requests waiting.
      out << "  memory_stalls_for_bank_busy: " << m_memBankBusy << endl;
      out << "  memory_stalls_for_random_busy: " << m_memRandBusy << endl;
      out << "  memory_stalls_for_anti_starvation: " << m_memNotOld << endl;
      out << "  memory_stalls_for_arbitration: " << m_memArbWait << endl;
      out << "  memory_stalls_for_bus: " << m_memBusBusy << endl;
      out << "  memory_stalls_for_tfaw: " << m_memTfawBusy << endl;
      out << "  memory_stalls_for_read_write_turnaround: " << m_memReadWriteBusy << endl;
      out << "  memory_stalls_for_read_read_turnaround: " << m_memDataBusBusy << endl;
      out << "  accesses_per_bank: ";
      for (int bank=0; bank < m_memBankCount.size(); bank++) {
        out << m_memBankCount[bank] << "  ";
        //if ((bank % 8) == 7) out << "                     " << endl;
      }
      out << endl;
      out << endl;
    }
  }
  if (!short_stats) {
    out << "Busy Controller Counts:" << endl;
    for(int i=0; i < MachineType_NUM; i++) {
      for(int j=0; j < MachineType_base_count((MachineType)i); j++) {
        MachineID machID;
        machID.type = (MachineType)i;
        machID.num = j;
        out << machID << ":" << m_busyControllerCount[i][j] << "  ";
        if ((j+1)%8 == 0) {
          out << endl;
        }
      }
      out << endl;
    }
    out << endl;

    out << "Busy Bank Count:" << m_busyBankCount << endl;
    out << endl;

    out << "sequencer_requests_outstanding: " << m_sequencer_requests << endl;
    out << endl;
  }

  if (!short_stats) {
    out << "All Non-Zero Cycle Demand Cache Accesses" << endl;
    out << "----------------------------------------" << endl;
    out << "miss_latency: " << m_allMissLatencyHistogram << endl;
    for(int i=0; i<m_missLatencyHistograms.size(); i++) {
      if (m_missLatencyHistograms[i].size() > 0) {
        out << "miss_latency_" << RubyRequestType(i) << ": " << m_missLatencyHistograms[i] << endl;
      }
    }
    for(int i=0; i<m_machLatencyHistograms.size(); i++) {
      if (m_machLatencyHistograms[i].size() > 0) {
        out << "miss_latency_" << GenericMachineType(i) << ": " << m_machLatencyHistograms[i] << endl;
      }
    }

    out << endl;

    out << "All Non-Zero Cycle SW Prefetch Requests" << endl;
    out << "------------------------------------" << endl;
    out << "prefetch_latency: " << m_allSWPrefetchLatencyHistogram << endl;
    for(int i=0; i<m_SWPrefetchLatencyHistograms.size(); i++) {
      if (m_SWPrefetchLatencyHistograms[i].size() > 0) {
        out << "prefetch_latency_" << CacheRequestType(i) << ": " << m_SWPrefetchLatencyHistograms[i] << endl;
      }
    }
    for(int i=0; i<m_SWPrefetchMachLatencyHistograms.size(); i++) {
      if (m_SWPrefetchMachLatencyHistograms[i].size() > 0) {
        out << "prefetch_latency_" << GenericMachineType(i) << ": " << m_SWPrefetchMachLatencyHistograms[i] << endl;
      }
    }
    out << "prefetch_latency_L2Miss:" << m_SWPrefetchL2MissLatencyHistogram << endl;

    if (m_all_sharing_histogram.size() > 0) {
      out << "all_sharing: " << m_all_sharing_histogram << endl;
      out << "read_sharing: " << m_read_sharing_histogram << endl;
      out << "write_sharing: " << m_write_sharing_histogram << endl;

      out << "all_sharing_percent: "; m_all_sharing_histogram.printPercent(out); out << endl;
      out << "read_sharing_percent: "; m_read_sharing_histogram.printPercent(out); out << endl;
      out << "write_sharing_percent: "; m_write_sharing_histogram.printPercent(out); out << endl;

      int64 total_miss = m_cache_to_cache +  m_memory_to_cache;
      out << "all_misses: " << total_miss << endl;
      out << "cache_to_cache_misses: " << m_cache_to_cache << endl;
      out << "memory_to_cache_misses: " << m_memory_to_cache << endl;
      out << "cache_to_cache_percent: " << 100.0 * (double(m_cache_to_cache) / double(total_miss)) << endl;
      out << "memory_to_cache_percent: " << 100.0 * (double(m_memory_to_cache) / double(total_miss)) << endl;
      out << endl;
    }

    if (m_outstanding_requests.size() > 0) {
      out << "outstanding_requests: "; m_outstanding_requests.printPercent(out); out << endl;
      out << endl;
    }
  }

  if (!short_stats) {
    out << "Request vs. RubySystem State Profile" << endl;
    out << "--------------------------------" << endl;
    out << endl;

    Vector<string> requestProfileKeys = m_requestProfileMap_ptr->keys();
    requestProfileKeys.sortVector();

    for(int i=0; i<requestProfileKeys.size(); i++) {
      int temp_int = m_requestProfileMap_ptr->lookup(requestProfileKeys[i]);
      double percent = (100.0*double(temp_int))/double(m_requests);
      while (requestProfileKeys[i] != "") {
        out << setw(10) << string_split(requestProfileKeys[i], ':');
      }
      out << setw(11) << temp_int;
      out << setw(14) << percent << endl;
    }
    out << endl;

    out << "filter_action: " << m_filter_action_histogram << endl;

    if (!m_all_instructions) {
      m_address_profiler_ptr->printStats(out);
    }

    if (m_all_instructions) {
      m_inst_profiler_ptr->printStats(out);
    }

    out << endl;
    out << "Message Delayed Cycles" << endl;
    out << "----------------------" << endl;
    out << "Total_delay_cycles: " <<   m_delayedCyclesHistogram << endl;
    out << "Total_nonPF_delay_cycles: " << m_delayedCyclesNonPFHistogram << endl;
    for (int i = 0; i < m_delayedCyclesVCHistograms.size(); i++) {
      out << "  virtual_network_" << i << "_delay_cycles: " << m_delayedCyclesVCHistograms[i] << endl;
    }

    printResourceUsage(out);
  }

}

void Profiler::printResourceUsage(ostream& out) const
{
  out << endl;
  out << "Resource Usage" << endl;
  out << "--------------" << endl;

  integer_t pagesize = getpagesize(); // page size in bytes
  out << "page_size: " << pagesize << endl;

  rusage usage;
  getrusage (RUSAGE_SELF, &usage);

  out << "user_time: " << usage.ru_utime.tv_sec << endl;
  out << "system_time: " << usage.ru_stime.tv_sec << endl;
  out << "page_reclaims: " << usage.ru_minflt << endl;
  out << "page_faults: " << usage.ru_majflt << endl;
  out << "swaps: " << usage.ru_nswap << endl;
  out << "block_inputs: " << usage.ru_inblock << endl;
  out << "block_outputs: " << usage.ru_oublock << endl;
}

void Profiler::clearStats()
{
  m_ruby_start = g_eventQueue_ptr->getTime();

  m_cycles_executed_at_start.setSize(RubySystem::getNumberOfSequencers());
  for (int i=0; i < RubySystem::getNumberOfSequencers(); i++) {
    if (g_system_ptr == NULL) {
      m_cycles_executed_at_start[i] = 0;
    } else {
      m_cycles_executed_at_start[i] = g_system_ptr->getCycleCount(i);
    }
  }

  m_perProcTotalMisses.setSize(RubySystem::getNumberOfSequencers());
  m_perProcUserMisses.setSize(RubySystem::getNumberOfSequencers());
  m_perProcSupervisorMisses.setSize(RubySystem::getNumberOfSequencers());
  m_perProcStartTransaction.setSize(RubySystem::getNumberOfSequencers());
  m_perProcEndTransaction.setSize(RubySystem::getNumberOfSequencers());

  for(int i=0; i < RubySystem::getNumberOfSequencers(); i++) {
    m_perProcTotalMisses[i] = 0;
    m_perProcUserMisses[i] = 0;
    m_perProcSupervisorMisses[i] = 0;
    m_perProcStartTransaction[i] = 0;
    m_perProcEndTransaction[i] = 0;
  }

  m_busyControllerCount.setSize(MachineType_NUM); // all machines
  for(int i=0; i < MachineType_NUM; i++) {
    m_busyControllerCount[i].setSize(MachineType_base_count((MachineType)i));
    for(int j=0; j < MachineType_base_count((MachineType)i); j++) {
      m_busyControllerCount[i][j] = 0;
    }
  }
  m_busyBankCount = 0;

  m_delayedCyclesHistogram.clear();
  m_delayedCyclesNonPFHistogram.clear();
  m_delayedCyclesVCHistograms.setSize(RubySystem::getNetwork()->getNumberOfVirtualNetworks());
  for (int i = 0; i < RubySystem::getNetwork()->getNumberOfVirtualNetworks(); i++) {
    m_delayedCyclesVCHistograms[i].clear();
  }

  m_missLatencyHistograms.setSize(RubyRequestType_NUM);
  for(int i=0; i<m_missLatencyHistograms.size(); i++) {
    m_missLatencyHistograms[i].clear(200);
  }
  m_machLatencyHistograms.setSize(GenericMachineType_NUM+1);
  for(int i=0; i<m_machLatencyHistograms.size(); i++) {
    m_machLatencyHistograms[i].clear(200);
  }
  m_allMissLatencyHistogram.clear(200);

  m_SWPrefetchLatencyHistograms.setSize(CacheRequestType_NUM);
  for(int i=0; i<m_SWPrefetchLatencyHistograms.size(); i++) {
    m_SWPrefetchLatencyHistograms[i].clear(200);
  }
  m_SWPrefetchMachLatencyHistograms.setSize(GenericMachineType_NUM+1);
  for(int i=0; i<m_SWPrefetchMachLatencyHistograms.size(); i++) {
    m_SWPrefetchMachLatencyHistograms[i].clear(200);
  }
  m_allSWPrefetchLatencyHistogram.clear(200);

  m_sequencer_requests.clear();
  m_read_sharing_histogram.clear();
  m_write_sharing_histogram.clear();
  m_all_sharing_histogram.clear();
  m_cache_to_cache = 0;
  m_memory_to_cache = 0;

  // clear HashMaps
  m_requestProfileMap_ptr->clear();

  // count requests profiled
  m_requests = 0;

  m_outstanding_requests.clear();
  m_outstanding_persistent_requests.clear();

//added by SS
  vector<string>::iterator it;

  for ( it=m_memory_control_names.begin() ; it < m_memory_control_names.end(); it++ ){
    m_memory_control_profilers[(*it).c_str()] -> m_memReq = 0;
    m_memory_control_profilers[(*it).c_str()] -> m_memBankBusy = 0;
    m_memory_control_profilers[(*it).c_str()] -> m_memBusBusy = 0;
    m_memory_control_profilers[(*it).c_str()] -> m_memTfawBusy = 0;
    m_memory_control_profilers[(*it).c_str()] -> m_memReadWriteBusy = 0;
    m_memory_control_profilers[(*it).c_str()] -> m_memDataBusBusy = 0;
    m_memory_control_profilers[(*it).c_str()] -> m_memRefresh = 0;
    m_memory_control_profilers[(*it).c_str()] -> m_memRead = 0;
    m_memory_control_profilers[(*it).c_str()] -> m_memWrite = 0;
    m_memory_control_profilers[(*it).c_str()] -> m_memWaitCycles = 0;
    m_memory_control_profilers[(*it).c_str()] -> m_memInputQ = 0;
    m_memory_control_profilers[(*it).c_str()] -> m_memBankQ = 0;
    m_memory_control_profilers[(*it).c_str()] -> m_memArbWait = 0;
    m_memory_control_profilers[(*it).c_str()] -> m_memRandBusy = 0;
    m_memory_control_profilers[(*it).c_str()] -> m_memNotOld = 0;

    for (int bank=0; bank < m_memory_control_profilers[(*it).c_str()] -> m_memBankCount.size(); bank++) {
        m_memory_control_profilers[(*it).c_str()] -> m_memBankCount[bank] = 0;
    }
  }
  // Flush the prefetches through the system - used so that there are no outstanding requests after stats are cleared
  //g_eventQueue_ptr->triggerAllEvents();

  // update the start time
  m_ruby_start = g_eventQueue_ptr->getTime();
}

void Profiler::addAddressTraceSample(const CacheMsg& msg, NodeID id)
{
  if (msg.getType() != CacheRequestType_IFETCH) {

    // Note: The following line should be commented out if you want to
    // use the special profiling that is part of the GS320 protocol

    // NOTE: Unless PROFILE_HOT_LINES is enabled, nothing will be profiled by the AddressProfiler
    m_address_profiler_ptr->addTraceSample(msg.getLineAddress(), msg.getProgramCounter(), msg.getType(), msg.getAccessMode(), id, false);
  }
}

void Profiler::profileSharing(const Address& addr, AccessType type, NodeID requestor, const Set& sharers, const Set& owner)
{
  Set set_contacted(owner);
  if (type == AccessType_Write) {
    set_contacted.addSet(sharers);
  }
  set_contacted.remove(requestor);
  int number_contacted = set_contacted.count();

  if (type == AccessType_Write) {
    m_write_sharing_histogram.add(number_contacted);
  } else {
    m_read_sharing_histogram.add(number_contacted);
  }
  m_all_sharing_histogram.add(number_contacted);

  if (number_contacted == 0) {
    m_memory_to_cache++;
  } else {
    m_cache_to_cache++;
  }

}

void Profiler::profileMsgDelay(int virtualNetwork, int delayCycles) {
  assert(virtualNetwork < m_delayedCyclesVCHistograms.size());
  m_delayedCyclesHistogram.add(delayCycles);
  m_delayedCyclesVCHistograms[virtualNetwork].add(delayCycles);
  if (virtualNetwork != 0) {
    m_delayedCyclesNonPFHistogram.add(delayCycles);
  }
}

// profiles original cache requests including PUTs
void Profiler::profileRequest(const string& requestStr)
{
  m_requests++;

  if (m_requestProfileMap_ptr->exist(requestStr)) {
    (m_requestProfileMap_ptr->lookup(requestStr))++;
  } else {
    m_requestProfileMap_ptr->add(requestStr, 1);
  }
}

void Profiler::startTransaction(int cpu)
{
  m_perProcStartTransaction[cpu]++;
}

void Profiler::endTransaction(int cpu)
{
  m_perProcEndTransaction[cpu]++;
}

void Profiler::controllerBusy(MachineID machID)
{
  m_busyControllerCount[(int)machID.type][(int)machID.num]++;
}

void Profiler::profilePFWait(Time waitTime)
{
  m_prefetchWaitHistogram.add(waitTime);
}

void Profiler::bankBusy()
{
  m_busyBankCount++;
}

// non-zero cycle demand request
void Profiler::missLatency(Time t, RubyRequestType type)
{
  m_allMissLatencyHistogram.add(t);
  m_missLatencyHistograms[type].add(t);
}

// non-zero cycle prefetch request
void Profiler::swPrefetchLatency(Time t, CacheRequestType type, GenericMachineType respondingMach)
{
  m_allSWPrefetchLatencyHistogram.add(t);
  m_SWPrefetchLatencyHistograms[type].add(t);
  m_SWPrefetchMachLatencyHistograms[respondingMach].add(t);
  if(respondingMach == GenericMachineType_Directory || respondingMach == GenericMachineType_NUM) {
    m_SWPrefetchL2MissLatencyHistogram.add(t);
  }
}

void Profiler::profileTransition(const string& component, NodeID version, Address addr,
                                 const string& state, const string& event,
                                 const string& next_state, const string& note)
{
  const int EVENT_SPACES = 20;
  const int ID_SPACES = 3;
  const int TIME_SPACES = 7;
  const int COMP_SPACES = 10;
  const int STATE_SPACES = 6;

  if ((g_debug_ptr->getDebugTime() > 0) &&
      (g_eventQueue_ptr->getTime() >= g_debug_ptr->getDebugTime())) {
    (* debug_cout_ptr).flags(ios::right);
    (* debug_cout_ptr) << setw(TIME_SPACES) << g_eventQueue_ptr->getTime() << " ";
    (* debug_cout_ptr) << setw(ID_SPACES) << version << " ";
    (* debug_cout_ptr) << setw(COMP_SPACES) << component;
    (* debug_cout_ptr) << setw(EVENT_SPACES) << event << " ";

    (* debug_cout_ptr).flags(ios::right);
    (* debug_cout_ptr) << setw(STATE_SPACES) << state;
    (* debug_cout_ptr) << ">";
    (* debug_cout_ptr).flags(ios::left);
    (* debug_cout_ptr) << setw(STATE_SPACES) << next_state;

    (* debug_cout_ptr) << " " << addr << " " << note;

    (* debug_cout_ptr) << endl;
  }
}

// Helper function
static double process_memory_total()
{
  const double MULTIPLIER = 4096.0/(1024.0*1024.0); // 4kB page size, 1024*1024 bytes per MB,
  ifstream proc_file;
  proc_file.open("/proc/self/statm");
  int total_size_in_pages = 0;
  int res_size_in_pages = 0;
  proc_file >> total_size_in_pages;
  proc_file >> res_size_in_pages;
  return double(total_size_in_pages)*MULTIPLIER; // size in megabytes
}

static double process_memory_resident()
{
  const double MULTIPLIER = 4096.0/(1024.0*1024.0); // 4kB page size, 1024*1024 bytes per MB,
  ifstream proc_file;
  proc_file.open("/proc/self/statm");
  int total_size_in_pages = 0;
  int res_size_in_pages = 0;
  proc_file >> total_size_in_pages;
  proc_file >> res_size_in_pages;
  return double(res_size_in_pages)*MULTIPLIER; // size in megabytes
}

void Profiler::rubyWatch(int id){
    //int rn_g1 = 0;//SIMICS_get_register_number(id, "g1");
  uint64 tr = 0;//SIMICS_read_register(id, rn_g1);
    Address watch_address = Address(tr);
    const int ID_SPACES = 3;
    const int TIME_SPACES = 7;

    (* debug_cout_ptr).flags(ios::right);
    (* debug_cout_ptr) << setw(TIME_SPACES) << g_eventQueue_ptr->getTime() << " ";
    (* debug_cout_ptr) << setw(ID_SPACES) << id << " "
                       << "RUBY WATCH "
                       << watch_address
                       << endl;

    if(!m_watch_address_list_ptr->exist(watch_address)){
      m_watch_address_list_ptr->add(watch_address, 1);
    }
}

bool Profiler::watchAddress(Address addr){
    if (m_watch_address_list_ptr->exist(addr))
      return true;
    else
      return false;
}

int64 Profiler::getTotalTransactionsExecuted() const {
  return m_perProcEndTransaction.sum();
}

// For MemoryControl:
void Profiler::profileMemReq(string name, int bank) {
//  printf("name is %s", name.c_str());
  assert(m_memory_control_profilers.count(name) == 1);
  m_memory_control_profilers[name] -> m_memReq++;
  m_memory_control_profilers[name] -> m_memBankCount[bank]++;
}
void Profiler::profileMemBankBusy(string name) {   assert(m_memory_control_profilers.count(name) == 1); m_memory_control_profilers[name] -> m_memBankBusy++; }
void Profiler::profileMemBusBusy(string name) {   assert(m_memory_control_profilers.count(name) == 1); m_memory_control_profilers[name] -> m_memBusBusy++; }
void Profiler::profileMemReadWriteBusy(string name) {   assert(m_memory_control_profilers.count(name) == 1); m_memory_control_profilers[name] -> m_memReadWriteBusy++; }
void Profiler::profileMemDataBusBusy(string name) {   assert(m_memory_control_profilers.count(name) == 1); m_memory_control_profilers[name] -> m_memDataBusBusy++; }
void Profiler::profileMemTfawBusy(string name) {   assert(m_memory_control_profilers.count(name) == 1); m_memory_control_profilers[name] -> m_memTfawBusy++; }
void Profiler::profileMemRefresh(string name) {   assert(m_memory_control_profilers.count(name) == 1); m_memory_control_profilers[name] -> m_memRefresh++; }
void Profiler::profileMemRead(string name) {   assert(m_memory_control_profilers.count(name) == 1); m_memory_control_profilers[name] -> m_memRead++; }
void Profiler::profileMemWrite(string name) {   assert(m_memory_control_profilers.count(name) == 1); m_memory_control_profilers[name] -> m_memWrite++; }
void Profiler::profileMemWaitCycles(string name, int cycles) {   assert(m_memory_control_profilers.count(name) == 1); m_memory_control_profilers[name] -> m_memWaitCycles += cycles; }
void Profiler::profileMemInputQ(string name, int cycles) {   assert(m_memory_control_profilers.count(name) == 1); m_memory_control_profilers[name] -> m_memInputQ += cycles; }
void Profiler::profileMemBankQ(string name, int cycles) {   assert(m_memory_control_profilers.count(name) == 1); m_memory_control_profilers[name] -> m_memBankQ += cycles; }
void Profiler::profileMemArbWait(string name, int cycles) {   assert(m_memory_control_profilers.count(name) == 1); m_memory_control_profilers[name] -> m_memArbWait += cycles; }
void Profiler::profileMemRandBusy(string name) {   assert(m_memory_control_profilers.count(name) == 1); m_memory_control_profilers[name] -> m_memRandBusy++; }
void Profiler::profileMemNotOld(string name) {   assert(m_memory_control_profilers.count(name) == 1); m_memory_control_profilers[name] -> m_memNotOld++; }


Profiler *
RubyProfilerParams::create()
{
    return new Profiler(this);
}