ruby: remove periodic event from Profiler
[gem5.git] / src / mem / ruby / profiler / Profiler.cc
1 /*
2 * Copyright (c) 1999-2008 Mark D. Hill and David A. Wood
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions are
7 * met: redistributions of source code must retain the above copyright
8 * notice, this list of conditions and the following disclaimer;
9 * redistributions in binary form must reproduce the above copyright
10 * notice, this list of conditions and the following disclaimer in the
11 * documentation and/or other materials provided with the distribution;
12 * neither the name of the copyright holders nor the names of its
13 * contributors may be used to endorse or promote products derived from
14 * this software without specific prior written permission.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
17 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
18 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
19 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
20 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
21 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
22 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
26 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 */
28
29 /*
30 This file has been modified by Kevin Moore and Dan Nussbaum of the
31 Scalable Systems Research Group at Sun Microsystems Laboratories
32 (http://research.sun.com/scalable/) to support the Adaptive
33 Transactional Memory Test Platform (ATMTP).
34
35 Please send email to atmtp-interest@sun.com with feedback, questions, or
36 to request future announcements about ATMTP.
37
38 ----------------------------------------------------------------------
39
40 File modification date: 2008-02-23
41
42 ----------------------------------------------------------------------
43 */
44
45 // Allows use of times() library call, which determines virtual runtime
46 #include <sys/resource.h>
47 #include <sys/times.h>
48 #include <sys/types.h>
49 #include <unistd.h>
50
51 #include <algorithm>
52 #include <fstream>
53
54 #include "base/stl_helpers.hh"
55 #include "base/str.hh"
56 #include "mem/protocol/MachineType.hh"
57 #include "mem/protocol/RubyRequest.hh"
58 #include "mem/ruby/network/Network.hh"
59 #include "mem/ruby/profiler/AddressProfiler.hh"
60 #include "mem/ruby/profiler/Profiler.hh"
61 #include "mem/ruby/system/Sequencer.hh"
62 #include "mem/ruby/system/System.hh"
63
64 using namespace std;
65 using m5::stl_helpers::operator<<;
66
67 static double process_memory_total();
68 static double process_memory_resident();
69
70 Profiler::Profiler(const Params *p)
71 : SimObject(p)
72 {
73 m_inst_profiler_ptr = NULL;
74 m_address_profiler_ptr = NULL;
75 m_real_time_start_time = time(NULL); // Not reset in clearStats()
76
77 m_hot_lines = p->hot_lines;
78 m_all_instructions = p->all_instructions;
79
80 m_num_of_sequencers = p->num_of_sequencers;
81
82 m_hot_lines = false;
83 m_all_instructions = false;
84
85 m_address_profiler_ptr = new AddressProfiler(m_num_of_sequencers);
86 m_address_profiler_ptr->setHotLines(m_hot_lines);
87 m_address_profiler_ptr->setAllInstructions(m_all_instructions);
88
89 if (m_all_instructions) {
90 m_inst_profiler_ptr = new AddressProfiler(m_num_of_sequencers);
91 m_inst_profiler_ptr->setHotLines(m_hot_lines);
92 m_inst_profiler_ptr->setAllInstructions(m_all_instructions);
93 }
94
95 p->ruby_system->registerProfiler(this);
96 }
97
98 Profiler::~Profiler()
99 {
100 }
101
102 void
103 Profiler::print(ostream& out) const
104 {
105 out << "[Profiler]";
106 }
107
108 void
109 Profiler::printRequestProfile(ostream &out) const
110 {
111 out << "Request vs. RubySystem State Profile" << endl;
112 out << "--------------------------------" << endl;
113 out << endl;
114
115 map<string, uint64_t> m_requestProfileMap;
116 uint64_t m_requests = 0;
117
118 for (uint32_t i = 0; i < MachineType_NUM; i++) {
119 for (map<uint32_t, AbstractController*>::iterator it =
120 g_abs_controls[i].begin();
121 it != g_abs_controls[i].end(); ++it) {
122
123 AbstractController *ctr = (*it).second;
124 map<string, uint64_t> mp = ctr->getRequestProfileMap();
125
126 for (map<string, uint64_t>::iterator jt = mp.begin();
127 jt != mp.end(); ++jt) {
128
129 map<string, uint64_t>::iterator kt =
130 m_requestProfileMap.find((*jt).first);
131 if (kt != m_requestProfileMap.end()) {
132 (*kt).second += (*jt).second;
133 } else {
134 m_requestProfileMap[(*jt).first] = (*jt).second;
135 }
136 }
137
138 m_requests += ctr->getRequestCount();
139 }
140 }
141
142 map<string, uint64_t>::const_iterator i = m_requestProfileMap.begin();
143 map<string, uint64_t>::const_iterator end = m_requestProfileMap.end();
144 for (; i != end; ++i) {
145 const string &key = i->first;
146 uint64_t count = i->second;
147
148 double percent = (100.0 * double(count)) / double(m_requests);
149 vector<string> items;
150 tokenize(items, key, ':');
151 vector<string>::iterator j = items.begin();
152 vector<string>::iterator end = items.end();
153 for (; j != end; ++i)
154 out << setw(10) << *j;
155 out << setw(11) << count;
156 out << setw(14) << percent << endl;
157 }
158 out << endl;
159 }
160
161 void
162 Profiler::printDelayProfile(ostream &out) const
163 {
164 out << "Message Delayed Cycles" << endl;
165 out << "----------------------" << endl;
166
167 uint32_t numVNets = Network::getNumberOfVirtualNetworks();
168 Histogram delayHistogram;
169 std::vector<Histogram> delayVCHistogram(numVNets);
170
171 for (uint32_t i = 0; i < MachineType_NUM; i++) {
172 for (map<uint32_t, AbstractController*>::iterator it =
173 g_abs_controls[i].begin();
174 it != g_abs_controls[i].end(); ++it) {
175
176 AbstractController *ctr = (*it).second;
177 delayHistogram.add(ctr->getDelayHist());
178
179 for (uint32_t i = 0; i < numVNets; i++) {
180 delayVCHistogram[i].add(ctr->getDelayVCHist(i));
181 }
182 }
183 }
184
185 out << "Total_delay_cycles: " << delayHistogram << endl;
186
187 for (int i = 0; i < numVNets; i++) {
188 out << " virtual_network_" << i << "_delay_cycles: "
189 << delayVCHistogram[i] << endl;
190 }
191 }
192
193 void
194 Profiler::printOutstandingReqProfile(ostream &out) const
195 {
196 Histogram sequencerRequests;
197
198 for (uint32_t i = 0; i < MachineType_NUM; i++) {
199 for (map<uint32_t, AbstractController*>::iterator it =
200 g_abs_controls[i].begin();
201 it != g_abs_controls[i].end(); ++it) {
202
203 AbstractController *ctr = (*it).second;
204 Sequencer *seq = ctr->getSequencer();
205 if (seq != NULL) {
206 sequencerRequests.add(seq->getOutstandReqHist());
207 }
208 }
209 }
210
211 out << "sequencer_requests_outstanding: "
212 << sequencerRequests << endl;
213 }
214
215 void
216 Profiler::printStats(ostream& out, bool short_stats)
217 {
218 out << endl;
219 if (short_stats) {
220 out << "SHORT ";
221 }
222 out << "Profiler Stats" << endl;
223 out << "--------------" << endl;
224
225 time_t real_time_current = time(NULL);
226 double seconds = difftime(real_time_current, m_real_time_start_time);
227 double minutes = seconds / 60.0;
228 double hours = minutes / 60.0;
229 double days = hours / 24.0;
230 Cycles ruby_cycles = g_system_ptr->curCycle()-m_ruby_start;
231
232 if (!short_stats) {
233 out << "Elapsed_time_in_seconds: " << seconds << endl;
234 out << "Elapsed_time_in_minutes: " << minutes << endl;
235 out << "Elapsed_time_in_hours: " << hours << endl;
236 out << "Elapsed_time_in_days: " << days << endl;
237 out << endl;
238 }
239
240 // print the virtual runtimes as well
241 struct tms vtime;
242 times(&vtime);
243 seconds = (vtime.tms_utime + vtime.tms_stime) / 100.0;
244 minutes = seconds / 60.0;
245 hours = minutes / 60.0;
246 days = hours / 24.0;
247 out << "Virtual_time_in_seconds: " << seconds << endl;
248 out << "Virtual_time_in_minutes: " << minutes << endl;
249 out << "Virtual_time_in_hours: " << hours << endl;
250 out << "Virtual_time_in_days: " << days << endl;
251 out << endl;
252
253 out << "Ruby_current_time: " << g_system_ptr->curCycle() << endl;
254 out << "Ruby_start_time: " << m_ruby_start << endl;
255 out << "Ruby_cycles: " << ruby_cycles << endl;
256 out << endl;
257
258 if (!short_stats) {
259 out << "mbytes_resident: " << process_memory_resident() << endl;
260 out << "mbytes_total: " << process_memory_total() << endl;
261 if (process_memory_total() > 0) {
262 out << "resident_ratio: "
263 << process_memory_resident()/process_memory_total() << endl;
264 }
265 out << endl;
266 }
267
268 vector<int64_t> perProcCycleCount(m_num_of_sequencers);
269
270 for (int i = 0; i < m_num_of_sequencers; i++) {
271 perProcCycleCount[i] =
272 g_system_ptr->curCycle() - m_cycles_executed_at_start[i] + 1;
273 // The +1 allows us to avoid division by zero
274 }
275
276 out << "ruby_cycles_executed: " << perProcCycleCount << endl;
277
278 out << endl;
279
280 if (!short_stats) {
281 out << "Busy Controller Counts:" << endl;
282 for (uint32_t i = 0; i < MachineType_NUM; i++) {
283 uint32_t size = MachineType_base_count((MachineType)i);
284
285 for (uint32_t j = 0; j < size; j++) {
286 MachineID machID;
287 machID.type = (MachineType)i;
288 machID.num = j;
289
290 AbstractController *ctr =
291 (*(g_abs_controls[i].find(j))).second;
292 out << machID << ":" << ctr->getFullyBusyCycles() << " ";
293 if ((j + 1) % 8 == 0) {
294 out << endl;
295 }
296 }
297 out << endl;
298 }
299 out << endl;
300
301 out << "Busy Bank Count:" << m_busyBankCount << endl;
302 out << endl;
303
304 printOutstandingReqProfile(out);
305 out << endl;
306 }
307
308 if (!short_stats) {
309 out << "All Non-Zero Cycle Demand Cache Accesses" << endl;
310 out << "----------------------------------------" << endl;
311 out << "miss_latency: " << m_allMissLatencyHistogram << endl;
312 for (int i = 0; i < m_missLatencyHistograms.size(); i++) {
313 if (m_missLatencyHistograms[i].size() > 0) {
314 out << "miss_latency_" << RubyRequestType(i) << ": "
315 << m_missLatencyHistograms[i] << endl;
316 }
317 }
318 for (int i = 0; i < m_machLatencyHistograms.size(); i++) {
319 if (m_machLatencyHistograms[i].size() > 0) {
320 out << "miss_latency_" << GenericMachineType(i) << ": "
321 << m_machLatencyHistograms[i] << endl;
322 }
323 }
324
325 out << "miss_latency_wCC_issue_to_initial_request: "
326 << m_wCCIssueToInitialRequestHistogram << endl;
327 out << "miss_latency_wCC_initial_forward_request: "
328 << m_wCCInitialRequestToForwardRequestHistogram << endl;
329 out << "miss_latency_wCC_forward_to_first_response: "
330 << m_wCCForwardRequestToFirstResponseHistogram << endl;
331 out << "miss_latency_wCC_first_response_to_completion: "
332 << m_wCCFirstResponseToCompleteHistogram << endl;
333 out << "imcomplete_wCC_Times: " << m_wCCIncompleteTimes << endl;
334 out << "miss_latency_dir_issue_to_initial_request: "
335 << m_dirIssueToInitialRequestHistogram << endl;
336 out << "miss_latency_dir_initial_forward_request: "
337 << m_dirInitialRequestToForwardRequestHistogram << endl;
338 out << "miss_latency_dir_forward_to_first_response: "
339 << m_dirForwardRequestToFirstResponseHistogram << endl;
340 out << "miss_latency_dir_first_response_to_completion: "
341 << m_dirFirstResponseToCompleteHistogram << endl;
342 out << "imcomplete_dir_Times: " << m_dirIncompleteTimes << endl;
343
344 for (int i = 0; i < m_missMachLatencyHistograms.size(); i++) {
345 for (int j = 0; j < m_missMachLatencyHistograms[i].size(); j++) {
346 if (m_missMachLatencyHistograms[i][j].size() > 0) {
347 out << "miss_latency_" << RubyRequestType(i)
348 << "_" << GenericMachineType(j) << ": "
349 << m_missMachLatencyHistograms[i][j] << endl;
350 }
351 }
352 }
353
354 out << endl;
355
356 out << "All Non-Zero Cycle SW Prefetch Requests" << endl;
357 out << "------------------------------------" << endl;
358 out << "prefetch_latency: " << m_allSWPrefetchLatencyHistogram << endl;
359 for (int i = 0; i < m_SWPrefetchLatencyHistograms.size(); i++) {
360 if (m_SWPrefetchLatencyHistograms[i].size() > 0) {
361 out << "prefetch_latency_" << RubyRequestType(i) << ": "
362 << m_SWPrefetchLatencyHistograms[i] << endl;
363 }
364 }
365 for (int i = 0; i < m_SWPrefetchMachLatencyHistograms.size(); i++) {
366 if (m_SWPrefetchMachLatencyHistograms[i].size() > 0) {
367 out << "prefetch_latency_" << GenericMachineType(i) << ": "
368 << m_SWPrefetchMachLatencyHistograms[i] << endl;
369 }
370 }
371 out << "prefetch_latency_L2Miss:"
372 << m_SWPrefetchL2MissLatencyHistogram << endl;
373
374 if (m_all_sharing_histogram.size() > 0) {
375 out << "all_sharing: " << m_all_sharing_histogram << endl;
376 out << "read_sharing: " << m_read_sharing_histogram << endl;
377 out << "write_sharing: " << m_write_sharing_histogram << endl;
378
379 out << "all_sharing_percent: ";
380 m_all_sharing_histogram.printPercent(out);
381 out << endl;
382
383 out << "read_sharing_percent: ";
384 m_read_sharing_histogram.printPercent(out);
385 out << endl;
386
387 out << "write_sharing_percent: ";
388 m_write_sharing_histogram.printPercent(out);
389 out << endl;
390
391 int64 total_miss = m_cache_to_cache + m_memory_to_cache;
392 out << "all_misses: " << total_miss << endl;
393 out << "cache_to_cache_misses: " << m_cache_to_cache << endl;
394 out << "memory_to_cache_misses: " << m_memory_to_cache << endl;
395 out << "cache_to_cache_percent: "
396 << 100.0 * (double(m_cache_to_cache) / double(total_miss))
397 << endl;
398 out << "memory_to_cache_percent: "
399 << 100.0 * (double(m_memory_to_cache) / double(total_miss))
400 << endl;
401 out << endl;
402 }
403
404 printRequestProfile(out);
405
406 out << "filter_action: " << m_filter_action_histogram << endl;
407
408 if (!m_all_instructions) {
409 m_address_profiler_ptr->printStats(out);
410 }
411
412 if (m_all_instructions) {
413 m_inst_profiler_ptr->printStats(out);
414 }
415
416 out << endl;
417 printDelayProfile(out);
418 printResourceUsage(out);
419 }
420 }
421
422 void
423 Profiler::printResourceUsage(ostream& out) const
424 {
425 out << endl;
426 out << "Resource Usage" << endl;
427 out << "--------------" << endl;
428
429 int64_t pagesize = getpagesize(); // page size in bytes
430 out << "page_size: " << pagesize << endl;
431
432 rusage usage;
433 getrusage (RUSAGE_SELF, &usage);
434
435 out << "user_time: " << usage.ru_utime.tv_sec << endl;
436 out << "system_time: " << usage.ru_stime.tv_sec << endl;
437 out << "page_reclaims: " << usage.ru_minflt << endl;
438 out << "page_faults: " << usage.ru_majflt << endl;
439 out << "swaps: " << usage.ru_nswap << endl;
440 out << "block_inputs: " << usage.ru_inblock << endl;
441 out << "block_outputs: " << usage.ru_oublock << endl;
442 }
443
444 void
445 Profiler::clearStats()
446 {
447 m_ruby_start = g_system_ptr->curCycle();
448 m_real_time_start_time = time(NULL);
449
450 m_cycles_executed_at_start.resize(m_num_of_sequencers);
451 for (int i = 0; i < m_num_of_sequencers; i++) {
452 m_cycles_executed_at_start[i] = g_system_ptr->curCycle();
453 }
454
455 m_busyBankCount = 0;
456
457 m_missLatencyHistograms.resize(RubyRequestType_NUM);
458 for (int i = 0; i < m_missLatencyHistograms.size(); i++) {
459 m_missLatencyHistograms[i].clear(200);
460 }
461 m_machLatencyHistograms.resize(GenericMachineType_NUM+1);
462 for (int i = 0; i < m_machLatencyHistograms.size(); i++) {
463 m_machLatencyHistograms[i].clear(200);
464 }
465 m_missMachLatencyHistograms.resize(RubyRequestType_NUM);
466 for (int i = 0; i < m_missLatencyHistograms.size(); i++) {
467 m_missMachLatencyHistograms[i].resize(GenericMachineType_NUM+1);
468 for (int j = 0; j < m_missMachLatencyHistograms[i].size(); j++) {
469 m_missMachLatencyHistograms[i][j].clear(200);
470 }
471 }
472 m_allMissLatencyHistogram.clear(200);
473 m_wCCIssueToInitialRequestHistogram.clear(200);
474 m_wCCInitialRequestToForwardRequestHistogram.clear(200);
475 m_wCCForwardRequestToFirstResponseHistogram.clear(200);
476 m_wCCFirstResponseToCompleteHistogram.clear(200);
477 m_wCCIncompleteTimes = 0;
478 m_dirIssueToInitialRequestHistogram.clear(200);
479 m_dirInitialRequestToForwardRequestHistogram.clear(200);
480 m_dirForwardRequestToFirstResponseHistogram.clear(200);
481 m_dirFirstResponseToCompleteHistogram.clear(200);
482 m_dirIncompleteTimes = 0;
483
484 m_SWPrefetchLatencyHistograms.resize(RubyRequestType_NUM);
485 for (int i = 0; i < m_SWPrefetchLatencyHistograms.size(); i++) {
486 m_SWPrefetchLatencyHistograms[i].clear(200);
487 }
488 m_SWPrefetchMachLatencyHistograms.resize(GenericMachineType_NUM+1);
489 for (int i = 0; i < m_SWPrefetchMachLatencyHistograms.size(); i++) {
490 m_SWPrefetchMachLatencyHistograms[i].clear(200);
491 }
492 m_allSWPrefetchLatencyHistogram.clear(200);
493
494 m_read_sharing_histogram.clear();
495 m_write_sharing_histogram.clear();
496 m_all_sharing_histogram.clear();
497 m_cache_to_cache = 0;
498 m_memory_to_cache = 0;
499
500 // update the start time
501 m_ruby_start = g_system_ptr->curCycle();
502 }
503
504 void
505 Profiler::addAddressTraceSample(const RubyRequest& msg, NodeID id)
506 {
507 if (msg.getType() != RubyRequestType_IFETCH) {
508 // Note: The following line should be commented out if you
509 // want to use the special profiling that is part of the GS320
510 // protocol
511
512 // NOTE: Unless PROFILE_HOT_LINES is enabled, nothing will be
513 // profiled by the AddressProfiler
514 m_address_profiler_ptr->
515 addTraceSample(msg.getLineAddress(), msg.getProgramCounter(),
516 msg.getType(), msg.getAccessMode(), id, false);
517 }
518 }
519
520 void
521 Profiler::profileSharing(const Address& addr, AccessType type,
522 NodeID requestor, const Set& sharers,
523 const Set& owner)
524 {
525 Set set_contacted(owner);
526 if (type == AccessType_Write) {
527 set_contacted.addSet(sharers);
528 }
529 set_contacted.remove(requestor);
530 int number_contacted = set_contacted.count();
531
532 if (type == AccessType_Write) {
533 m_write_sharing_histogram.add(number_contacted);
534 } else {
535 m_read_sharing_histogram.add(number_contacted);
536 }
537 m_all_sharing_histogram.add(number_contacted);
538
539 if (number_contacted == 0) {
540 m_memory_to_cache++;
541 } else {
542 m_cache_to_cache++;
543 }
544 }
545
546 void
547 Profiler::profilePFWait(Cycles waitTime)
548 {
549 m_prefetchWaitHistogram.add(waitTime);
550 }
551
552 void
553 Profiler::bankBusy()
554 {
555 m_busyBankCount++;
556 }
557
558 // non-zero cycle demand request
559 void
560 Profiler::missLatency(Cycles cycles,
561 RubyRequestType type,
562 const GenericMachineType respondingMach)
563 {
564 m_allMissLatencyHistogram.add(cycles);
565 m_missLatencyHistograms[type].add(cycles);
566 m_machLatencyHistograms[respondingMach].add(cycles);
567 m_missMachLatencyHistograms[type][respondingMach].add(cycles);
568 }
569
570 void
571 Profiler::missLatencyWcc(Cycles issuedTime,
572 Cycles initialRequestTime,
573 Cycles forwardRequestTime,
574 Cycles firstResponseTime,
575 Cycles completionTime)
576 {
577 if ((issuedTime <= initialRequestTime) &&
578 (initialRequestTime <= forwardRequestTime) &&
579 (forwardRequestTime <= firstResponseTime) &&
580 (firstResponseTime <= completionTime)) {
581 m_wCCIssueToInitialRequestHistogram.add(initialRequestTime - issuedTime);
582
583 m_wCCInitialRequestToForwardRequestHistogram.add(forwardRequestTime -
584 initialRequestTime);
585
586 m_wCCForwardRequestToFirstResponseHistogram.add(firstResponseTime -
587 forwardRequestTime);
588
589 m_wCCFirstResponseToCompleteHistogram.add(completionTime -
590 firstResponseTime);
591 } else {
592 m_wCCIncompleteTimes++;
593 }
594 }
595
596 void
597 Profiler::missLatencyDir(Cycles issuedTime,
598 Cycles initialRequestTime,
599 Cycles forwardRequestTime,
600 Cycles firstResponseTime,
601 Cycles completionTime)
602 {
603 if ((issuedTime <= initialRequestTime) &&
604 (initialRequestTime <= forwardRequestTime) &&
605 (forwardRequestTime <= firstResponseTime) &&
606 (firstResponseTime <= completionTime)) {
607 m_dirIssueToInitialRequestHistogram.add(initialRequestTime - issuedTime);
608
609 m_dirInitialRequestToForwardRequestHistogram.add(forwardRequestTime -
610 initialRequestTime);
611
612 m_dirForwardRequestToFirstResponseHistogram.add(firstResponseTime -
613 forwardRequestTime);
614
615 m_dirFirstResponseToCompleteHistogram.add(completionTime -
616 firstResponseTime);
617 } else {
618 m_dirIncompleteTimes++;
619 }
620 }
621
622 // non-zero cycle prefetch request
623 void
624 Profiler::swPrefetchLatency(Cycles cycles, RubyRequestType type,
625 const GenericMachineType respondingMach)
626 {
627 m_allSWPrefetchLatencyHistogram.add(cycles);
628 m_SWPrefetchLatencyHistograms[type].add(cycles);
629 m_SWPrefetchMachLatencyHistograms[respondingMach].add(cycles);
630
631 if (respondingMach == GenericMachineType_Directory ||
632 respondingMach == GenericMachineType_NUM) {
633 m_SWPrefetchL2MissLatencyHistogram.add(cycles);
634 }
635 }
636
637 // Helper function
638 static double
639 process_memory_total()
640 {
641 // 4kB page size, 1024*1024 bytes per MB,
642 const double MULTIPLIER = 4096.0 / (1024.0 * 1024.0);
643 ifstream proc_file;
644 proc_file.open("/proc/self/statm");
645 int total_size_in_pages = 0;
646 int res_size_in_pages = 0;
647 proc_file >> total_size_in_pages;
648 proc_file >> res_size_in_pages;
649 return double(total_size_in_pages) * MULTIPLIER; // size in megabytes
650 }
651
652 static double
653 process_memory_resident()
654 {
655 // 4kB page size, 1024*1024 bytes per MB,
656 const double MULTIPLIER = 4096.0 / (1024.0 * 1024.0);
657 ifstream proc_file;
658 proc_file.open("/proc/self/statm");
659 int total_size_in_pages = 0;
660 int res_size_in_pages = 0;
661 proc_file >> total_size_in_pages;
662 proc_file >> res_size_in_pages;
663 return double(res_size_in_pages) * MULTIPLIER; // size in megabytes
664 }
665
666 void
667 Profiler::rubyWatch(int id)
668 {
669 uint64 tr = 0;
670 Address watch_address = Address(tr);
671
672 DPRINTFN("%7s %3s RUBY WATCH %d\n", g_system_ptr->curCycle(), id,
673 watch_address);
674
675 // don't care about success or failure
676 m_watch_address_set.insert(watch_address);
677 }
678
679 bool
680 Profiler::watchAddress(Address addr)
681 {
682 return m_watch_address_set.count(addr) > 0;
683 }
684
685 Profiler *
686 RubyProfilerParams::create()
687 {
688 return new Profiler(this);
689 }