ruby: profiler: lots of inter-related changes

The patch started of with removing the global variables from the profiler for
profiling the miss latency of requests made to the cache. The corrresponding
histograms have been moved to the Sequencer. These are combined together when
the histograms are printed. Separate histograms are now maintained for
tracking latency of all requests together, of hits only and of misses only.

A particular set of histograms used to use the type GenericMachineType defined
in one of the protocol files. This patch removes this type. Now, everything
that relied on this type would use MachineType instead. To do this, SLICC has
been changed so that multiple machine types can be declared by a controller
in its preamble.

5 files changed, 387 insertions, 309 deletions

diff --git a/src/mem/ruby/common/Histogram.cc b/src/mem/ruby/common/Histogram.cc
index 0558e5198..e377bc253 100644
--- a/src/mem/ruby/common/Histogram.cc
+++ b/src/mem/ruby/common/Histogram.cc
@@ -124,7 +124,12 @@ Histogram::add(Histogram& hist)
     uint32_t t_bins = m_data.size();
 
     if (hist.getBins() != t_bins) {
-        fatal("Histograms with different number of bins cannot be combined!");
+        if (m_count == 0) {
+            m_data.resize(hist.getBins());
+        } else {
+            fatal("Histograms with different number of bins "
+                  "cannot be combined!");
+        }
     }
 
     m_max = max(m_max, hist.getMax());
diff --git a/src/mem/ruby/profiler/Profiler.cc b/src/mem/ruby/profiler/Profiler.cc
index 76c4dba74..9a963684f 100644
--- a/src/mem/ruby/profiler/Profiler.cc
+++ b/src/mem/ruby/profiler/Profiler.cc
@@ -213,6 +213,169 @@ Profiler::printOutstandingReqProfile(ostream &out) const
 }
 
 void
+Profiler::printMissLatencyProfile(ostream &out) const
+{
+    // Collate the miss latencies histograms from all the sequencers
+    Histogram latency_hist;
+    std::vector<Histogram> type_latency_hist(RubyRequestType_NUM);
+
+    Histogram hit_latency_hist;
+    std::vector<Histogram> hit_type_latency_hist(RubyRequestType_NUM);
+
+    std::vector<Histogram> hit_mach_latency_hist(MachineType_NUM);
+    std::vector<std::vector<Histogram> >
+        hit_type_mach_latency_hist(RubyRequestType_NUM,
+                               std::vector<Histogram>(MachineType_NUM));
+
+    Histogram miss_latency_hist;
+    std::vector<Histogram> miss_type_latency_hist(RubyRequestType_NUM);
+
+    std::vector<Histogram> miss_mach_latency_hist(MachineType_NUM);
+    std::vector<std::vector<Histogram> >
+        miss_type_mach_latency_hist(RubyRequestType_NUM,
+                               std::vector<Histogram>(MachineType_NUM));
+
+    std::vector<Histogram> issue_to_initial_delay_hist(MachineType_NUM);
+    std::vector<Histogram> initial_to_forward_delay_hist(MachineType_NUM);
+    std::vector<Histogram>
+        forward_to_first_response_delay_hist(MachineType_NUM);
+    std::vector<Histogram>
+        first_response_to_completion_delay_hist(MachineType_NUM);
+    std::vector<uint64_t> incomplete_times(MachineType_NUM);
+
+    for (uint32_t i = 0; i < MachineType_NUM; i++) {
+        for (map<uint32_t, AbstractController*>::iterator it =
+                  g_abs_controls[i].begin();
+             it != g_abs_controls[i].end(); ++it) {
+
+            AbstractController *ctr = (*it).second;
+            Sequencer *seq = ctr->getSequencer();
+            if (seq != NULL) {
+                // add all the latencies
+                latency_hist.add(seq->getLatencyHist());
+                hit_latency_hist.add(seq->getHitLatencyHist());
+                miss_latency_hist.add(seq->getMissLatencyHist());
+
+                // add the per request type latencies
+                for (uint32_t j = 0; j < RubyRequestType_NUM; ++j) {
+                    type_latency_hist[j]
+                        .add(seq->getTypeLatencyHist(j));
+                    hit_type_latency_hist[j]
+                        .add(seq->getHitTypeLatencyHist(j));
+                    miss_type_latency_hist[j]
+                        .add(seq->getMissTypeLatencyHist(j));
+                }
+
+                // add the per machine type miss latencies
+                for (uint32_t j = 0; j < MachineType_NUM; ++j) {
+                    hit_mach_latency_hist[j]
+                        .add(seq->getHitMachLatencyHist(j));
+                    miss_mach_latency_hist[j]
+                        .add(seq->getMissMachLatencyHist(j));
+
+                    issue_to_initial_delay_hist[j].add(
+                        seq->getIssueToInitialDelayHist(MachineType(j)));
+
+                    initial_to_forward_delay_hist[j].add(
+                        seq->getInitialToForwardDelayHist(MachineType(j)));
+                    forward_to_first_response_delay_hist[j].add(seq->
+                        getForwardRequestToFirstResponseHist(MachineType(j)));
+
+                    first_response_to_completion_delay_hist[j].add(seq->
+                        getFirstResponseToCompletionDelayHist(MachineType(j)));
+                    incomplete_times[j] +=
+                        seq->getIncompleteTimes(MachineType(j));
+                }
+
+                // add the per (request, machine) type miss latencies
+                for (uint32_t j = 0; j < RubyRequestType_NUM; j++) {
+                    for (uint32_t k = 0; k < MachineType_NUM; k++) {
+                        hit_type_mach_latency_hist[j][k].add(
+                            seq->getHitTypeMachLatencyHist(j,k));
+                        miss_type_mach_latency_hist[j][k].add(
+                            seq->getMissTypeMachLatencyHist(j,k));
+                    }
+                }
+            }
+        }
+    }
+
+    out << "latency: " << latency_hist << endl;
+    for (int i = 0; i < RubyRequestType_NUM; i++) {
+        if (type_latency_hist[i].size() > 0) {
+            out << "latency: " << RubyRequestType(i) << ": "
+                << type_latency_hist[i] << endl;
+        }
+    }
+
+    out << "hit latency: " << hit_latency_hist << endl;
+    for (int i = 0; i < RubyRequestType_NUM; i++) {
+        if (hit_type_latency_hist[i].size() > 0) {
+            out << "hit latency: " << RubyRequestType(i) << ": "
+                << hit_type_latency_hist[i] << endl;
+        }
+    }
+
+    for (int i = 0; i < MachineType_NUM; i++) {
+        if (hit_mach_latency_hist[i].size() > 0) {
+            out << "hit latency: " << MachineType(i) << ": "
+                << hit_mach_latency_hist[i] << endl;
+        }
+    }
+
+    for (int i = 0; i < RubyRequestType_NUM; i++) {
+        for (int j = 0; j < MachineType_NUM; j++) {
+            if (hit_type_mach_latency_hist[i][j].size() > 0) {
+                out << "hit latency: " << RubyRequestType(i)
+                    << ": " << MachineType(j) << ": "
+                    << hit_type_mach_latency_hist[i][j] << endl;
+            }
+        }
+    }
+
+    out << "miss latency: " << miss_latency_hist << endl;
+    for (int i = 0; i < RubyRequestType_NUM; i++) {
+        if (miss_type_latency_hist[i].size() > 0) {
+            out << "miss latency: " << RubyRequestType(i) << ": "
+                << miss_type_latency_hist[i] << endl;
+        }
+    }
+
+    for (int i = 0; i < MachineType_NUM; i++) {
+        if (miss_mach_latency_hist[i].size() > 0) {
+            out << "miss latency: " << MachineType(i) << ": "
+                << miss_mach_latency_hist[i] << endl;
+
+            out << "miss latency: " << MachineType(i)
+                << "::issue_to_initial_request: "
+                << issue_to_initial_delay_hist[i] << endl;
+            out << "miss latency: " << MachineType(i)
+                << "::initial_to_forward_request: "
+                << initial_to_forward_delay_hist[i] << endl;
+            out << "miss latency: " << MachineType(i)
+                << "::forward_to_first_response: "
+                << forward_to_first_response_delay_hist[i] << endl;
+            out << "miss latency: " << MachineType(i)
+                << "::first_response_to_completion: "
+                << first_response_to_completion_delay_hist[i] << endl;
+            out << "incomplete times: " << incomplete_times[i] << endl;
+        }
+    }
+
+    for (int i = 0; i < RubyRequestType_NUM; i++) {
+        for (int j = 0; j < MachineType_NUM; j++) {
+            if (miss_type_mach_latency_hist[i][j].size() > 0) {
+                out << "miss latency: " << RubyRequestType(i)
+                    << ": " << MachineType(j) << ": "
+                    << miss_type_mach_latency_hist[i][j] << endl;
+            }
+        }
+    }
+
+    out << endl;
+}
+
+void
 Profiler::printStats(ostream& out, bool short_stats)
 {
     out << endl;
@@ -296,68 +459,7 @@ Profiler::printStats(ostream& out, bool short_stats)
     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 << "miss_latency_wCC_issue_to_initial_request: " 
-            << m_wCCIssueToInitialRequestHistogram << endl;
-        out << "miss_latency_wCC_initial_forward_request: " 
-            << m_wCCInitialRequestToForwardRequestHistogram << endl;
-        out << "miss_latency_wCC_forward_to_first_response: " 
-            << m_wCCForwardRequestToFirstResponseHistogram << endl;
-        out << "miss_latency_wCC_first_response_to_completion: " 
-            << m_wCCFirstResponseToCompleteHistogram << endl;
-        out << "imcomplete_wCC_Times: " << m_wCCIncompleteTimes << endl;
-        out << "miss_latency_dir_issue_to_initial_request: " 
-            << m_dirIssueToInitialRequestHistogram << endl;
-        out << "miss_latency_dir_initial_forward_request: " 
-            << m_dirInitialRequestToForwardRequestHistogram << endl;
-        out << "miss_latency_dir_forward_to_first_response: " 
-            << m_dirForwardRequestToFirstResponseHistogram << endl;
-        out << "miss_latency_dir_first_response_to_completion: " 
-            << m_dirFirstResponseToCompleteHistogram << endl;
-        out << "imcomplete_dir_Times: " << m_dirIncompleteTimes << endl;
-
-        for (int i = 0; i < m_missMachLatencyHistograms.size(); i++) {
-            for (int j = 0; j < m_missMachLatencyHistograms[i].size(); j++) {
-                if (m_missMachLatencyHistograms[i][j].size() > 0) {
-                    out << "miss_latency_" << RubyRequestType(i) 
-                        << "_" << GenericMachineType(j) << ": "
-                        << m_missMachLatencyHistograms[i][j] << 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_" << RubyRequestType(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;
+        printMissLatencyProfile(out);
 
         if (m_all_sharing_histogram.size() > 0) {
             out << "all_sharing: " << m_all_sharing_histogram << endl;
@@ -434,44 +536,6 @@ Profiler::clearStats()
     m_real_time_start_time = time(NULL);
 
     m_busyBankCount = 0;
-
-    m_missLatencyHistograms.resize(RubyRequestType_NUM);
-    for (int i = 0; i < m_missLatencyHistograms.size(); i++) {
-        m_missLatencyHistograms[i].clear(200);
-    }
-    m_machLatencyHistograms.resize(GenericMachineType_NUM+1);
-    for (int i = 0; i < m_machLatencyHistograms.size(); i++) {
-        m_machLatencyHistograms[i].clear(200);
-    }
-    m_missMachLatencyHistograms.resize(RubyRequestType_NUM);
-    for (int i = 0; i < m_missLatencyHistograms.size(); i++) {
-        m_missMachLatencyHistograms[i].resize(GenericMachineType_NUM+1);
-        for (int j = 0; j < m_missMachLatencyHistograms[i].size(); j++) {
-            m_missMachLatencyHistograms[i][j].clear(200);
-        }
-    }
-    m_allMissLatencyHistogram.clear(200);
-    m_wCCIssueToInitialRequestHistogram.clear(200);
-    m_wCCInitialRequestToForwardRequestHistogram.clear(200);
-    m_wCCForwardRequestToFirstResponseHistogram.clear(200);
-    m_wCCFirstResponseToCompleteHistogram.clear(200);
-    m_wCCIncompleteTimes = 0;
-    m_dirIssueToInitialRequestHistogram.clear(200);
-    m_dirInitialRequestToForwardRequestHistogram.clear(200);
-    m_dirForwardRequestToFirstResponseHistogram.clear(200);
-    m_dirFirstResponseToCompleteHistogram.clear(200);
-    m_dirIncompleteTimes = 0;
-
-    m_SWPrefetchLatencyHistograms.resize(RubyRequestType_NUM);
-    for (int i = 0; i < m_SWPrefetchLatencyHistograms.size(); i++) {
-        m_SWPrefetchLatencyHistograms[i].clear(200);
-    }
-    m_SWPrefetchMachLatencyHistograms.resize(GenericMachineType_NUM+1);
-    for (int i = 0; i < m_SWPrefetchMachLatencyHistograms.size(); i++) {
-        m_SWPrefetchMachLatencyHistograms[i].clear(200);
-    }
-    m_allSWPrefetchLatencyHistogram.clear(200);
-
     m_read_sharing_histogram.clear();
     m_write_sharing_histogram.clear();
     m_all_sharing_histogram.clear();
@@ -530,85 +594,6 @@ Profiler::bankBusy()
     m_busyBankCount++;
 }
 
-// non-zero cycle demand request
-void
-Profiler::missLatency(Cycles cycles,
-                      RubyRequestType type,
-                      const GenericMachineType respondingMach)
-{
-    m_allMissLatencyHistogram.add(cycles);
-    m_missLatencyHistograms[type].add(cycles);
-    m_machLatencyHistograms[respondingMach].add(cycles);
-    m_missMachLatencyHistograms[type][respondingMach].add(cycles);
-}
-
-void
-Profiler::missLatencyWcc(Cycles issuedTime,
-                         Cycles initialRequestTime,
-                         Cycles forwardRequestTime,
-                         Cycles firstResponseTime,
-                         Cycles completionTime)
-{
-    if ((issuedTime <= initialRequestTime) &&
-        (initialRequestTime <= forwardRequestTime) &&
-        (forwardRequestTime <= firstResponseTime) &&
-        (firstResponseTime <= completionTime)) {
-        m_wCCIssueToInitialRequestHistogram.add(initialRequestTime - issuedTime);
-        
-        m_wCCInitialRequestToForwardRequestHistogram.add(forwardRequestTime - 
-                                                         initialRequestTime);
-        
-        m_wCCForwardRequestToFirstResponseHistogram.add(firstResponseTime - 
-                                                        forwardRequestTime);
-        
-        m_wCCFirstResponseToCompleteHistogram.add(completionTime - 
-                                                  firstResponseTime);
-    } else {
-        m_wCCIncompleteTimes++;
-    }
-}
-
-void
-Profiler::missLatencyDir(Cycles issuedTime,
-                         Cycles initialRequestTime,
-                         Cycles forwardRequestTime,
-                         Cycles firstResponseTime,
-                         Cycles completionTime)
-{
-    if ((issuedTime <= initialRequestTime) &&
-        (initialRequestTime <= forwardRequestTime) &&
-        (forwardRequestTime <= firstResponseTime) &&
-        (firstResponseTime <= completionTime)) {
-        m_dirIssueToInitialRequestHistogram.add(initialRequestTime - issuedTime);
-        
-        m_dirInitialRequestToForwardRequestHistogram.add(forwardRequestTime - 
-                                                         initialRequestTime);
-        
-        m_dirForwardRequestToFirstResponseHistogram.add(firstResponseTime - 
-                                                        forwardRequestTime);
-        
-        m_dirFirstResponseToCompleteHistogram.add(completionTime - 
-                                                  firstResponseTime);
-    } else {
-        m_dirIncompleteTimes++;
-    }
-}
-
-// non-zero cycle prefetch request
-void
-Profiler::swPrefetchLatency(Cycles cycles, RubyRequestType type,
-                            const GenericMachineType respondingMach)
-{
-    m_allSWPrefetchLatencyHistogram.add(cycles);
-    m_SWPrefetchLatencyHistograms[type].add(cycles);
-    m_SWPrefetchMachLatencyHistograms[respondingMach].add(cycles);
-
-    if (respondingMach == GenericMachineType_Directory ||
-        respondingMach == GenericMachineType_NUM) {
-        m_SWPrefetchL2MissLatencyHistogram.add(cycles);
-    }
-}
-
 // Helper function
 static double
 process_memory_total()
diff --git a/src/mem/ruby/profiler/Profiler.hh b/src/mem/ruby/profiler/Profiler.hh
index 23efed67a..e7b3c5f8d 100644
--- a/src/mem/ruby/profiler/Profiler.hh
+++ b/src/mem/ruby/profiler/Profiler.hh
@@ -52,7 +52,6 @@
 
 #include "base/hashmap.hh"
 #include "mem/protocol/AccessType.hh"
-#include "mem/protocol/GenericMachineType.hh"
 #include "mem/protocol/PrefetchBit.hh"
 #include "mem/protocol/RubyAccessMode.hh"
 #include "mem/protocol/RubyRequestType.hh"
@@ -110,21 +109,7 @@ class Profiler : public SimObject
 
     void controllerBusy(MachineID machID);
     void bankBusy();
-
-    void missLatency(Cycles t, RubyRequestType type,
-                     const GenericMachineType respondingMach);
-
-    void missLatencyWcc(Cycles issuedTime, Cycles initialRequestTime,
-                        Cycles forwardRequestTime, Cycles firstResponseTime,
-                        Cycles completionTime);
-    
-    void missLatencyDir(Cycles issuedTime, Cycles initialRequestTime,
-                        Cycles forwardRequestTime, Cycles firstResponseTime,
-                        Cycles completionTime);
     
-    void swPrefetchLatency(Cycles t, RubyRequestType type,
-                           const GenericMachineType respondingMach);
-
     void print(std::ostream& out) const;
 
     void rubyWatch(int proc);
@@ -141,6 +126,7 @@ class Profiler : public SimObject
     void printRequestProfile(std::ostream &out) const;
     void printDelayProfile(std::ostream &out) const;
     void printOutstandingReqProfile(std::ostream &out) const;
+    void printMissLatencyProfile(std::ostream &out) const;
 
   private:
     // Private copy constructor and assignment operator
@@ -161,27 +147,6 @@ class Profiler : public SimObject
     int64 m_cache_to_cache;
     int64 m_memory_to_cache;
 
-    std::vector<Histogram> m_missLatencyHistograms;
-    std::vector<Histogram> m_machLatencyHistograms;
-    std::vector< std::vector<Histogram> > m_missMachLatencyHistograms;
-    Histogram m_wCCIssueToInitialRequestHistogram;
-    Histogram m_wCCInitialRequestToForwardRequestHistogram;
-    Histogram m_wCCForwardRequestToFirstResponseHistogram;
-    Histogram m_wCCFirstResponseToCompleteHistogram;
-    int64 m_wCCIncompleteTimes;
-    Histogram m_dirIssueToInitialRequestHistogram;
-    Histogram m_dirInitialRequestToForwardRequestHistogram;
-    Histogram m_dirForwardRequestToFirstResponseHistogram;
-    Histogram m_dirFirstResponseToCompleteHistogram;
-    int64 m_dirIncompleteTimes;
-
-    Histogram m_allMissLatencyHistogram;
-
-    Histogram m_allSWPrefetchLatencyHistogram;
-    Histogram m_SWPrefetchL2MissLatencyHistogram;
-    std::vector<Histogram> m_SWPrefetchLatencyHistograms;
-    std::vector<Histogram> m_SWPrefetchMachLatencyHistograms;
-
     Histogram m_average_latency_estimate;
     m5::hash_set<Address> m_watch_address_set;
 
diff --git a/src/mem/ruby/system/Sequencer.cc b/src/mem/ruby/system/Sequencer.cc
index 606dff0ab..8e61766b8 100644
--- a/src/mem/ruby/system/Sequencer.cc
+++ b/src/mem/ruby/system/Sequencer.cc
@@ -136,6 +136,61 @@ Sequencer::wakeup()
 void Sequencer::clearStats()
 {
     m_outstandReqHist.clear();
+
+    // Initialize the histograms that track latency of all requests
+    m_latencyHist.clear(20);
+    m_typeLatencyHist.resize(RubyRequestType_NUM);
+    for (int i = 0; i < RubyRequestType_NUM; i++) {
+        m_typeLatencyHist[i].clear(20);
+    }
+
+    // Initialize the histograms that track latency of requests that
+    // hit in the cache attached to the sequencer.
+    m_hitLatencyHist.clear(20);
+    m_hitTypeLatencyHist.resize(RubyRequestType_NUM);
+    m_hitTypeMachLatencyHist.resize(RubyRequestType_NUM);
+
+    for (int i = 0; i < RubyRequestType_NUM; i++) {
+        m_hitTypeLatencyHist[i].clear(20);
+        m_hitTypeMachLatencyHist[i].resize(MachineType_NUM);
+        for (int j = 0; j < MachineType_NUM; j++) {
+            m_hitTypeMachLatencyHist[i][j].clear(20);
+        }
+    }
+
+    // Initialize the histograms that track the latency of requests that
+    // missed in the cache attached to the sequencer.
+    m_missLatencyHist.clear(20);
+    m_missTypeLatencyHist.resize(RubyRequestType_NUM);
+    m_missTypeMachLatencyHist.resize(RubyRequestType_NUM);
+
+    for (int i = 0; i < RubyRequestType_NUM; i++) {
+        m_missTypeLatencyHist[i].clear(20);
+        m_missTypeMachLatencyHist[i].resize(MachineType_NUM);
+        for (int j = 0; j < MachineType_NUM; j++) {
+            m_missTypeMachLatencyHist[i][j].clear(20);
+        }
+    }
+
+    m_hitMachLatencyHist.resize(MachineType_NUM);
+    m_missMachLatencyHist.resize(MachineType_NUM);
+    m_IssueToInitialDelayHist.resize(MachineType_NUM);
+    m_InitialToForwardDelayHist.resize(MachineType_NUM);
+    m_ForwardToFirstResponseDelayHist.resize(MachineType_NUM);
+    m_FirstResponseToCompletionDelayHist.resize(MachineType_NUM);
+    m_IncompleteTimes.resize(MachineType_NUM);
+
+    for (int i = 0; i < MachineType_NUM; i++) {
+        m_missMachLatencyHist[i].clear(20);
+        m_hitMachLatencyHist[i].clear(20);
+
+        m_IssueToInitialDelayHist[i].clear(20);
+        m_InitialToForwardDelayHist[i].clear(20);
+        m_ForwardToFirstResponseDelayHist[i].clear(20);
+        m_FirstResponseToCompletionDelayHist[i].clear(20);
+
+        m_IncompleteTimes[i] = 0;
+    }
 }
 
 void
@@ -370,26 +425,58 @@ Sequencer::handleLlsc(const Address& address, SequencerRequest* request)
 }
 
 void
-Sequencer::writeCallback(const Address& address, DataBlock& data)
+Sequencer::recordMissLatency(const Cycles cycles, const RubyRequestType type,
+                             const MachineType respondingMach,
+                             bool isExternalHit, Cycles issuedTime,
+                             Cycles initialRequestTime,
+                             Cycles forwardRequestTime,
+                             Cycles firstResponseTime, Cycles completionTime)
 {
-    writeCallback(address, GenericMachineType_NULL, data);
-}
+    m_latencyHist.add(cycles);
+    m_typeLatencyHist[type].add(cycles);
+
+    if (isExternalHit) {
+        m_missLatencyHist.add(cycles);
+        m_missTypeLatencyHist[type].add(cycles);
+
+        if (respondingMach != MachineType_NUM) {
+            m_missMachLatencyHist[respondingMach].add(cycles);
+            m_missTypeMachLatencyHist[type][respondingMach].add(cycles);
+
+            if ((issuedTime <= initialRequestTime) &&
+                (initialRequestTime <= forwardRequestTime) &&
+                (forwardRequestTime <= firstResponseTime) &&
+                (firstResponseTime <= completionTime)) {
+
+                m_IssueToInitialDelayHist[respondingMach].add(
+                    initialRequestTime - issuedTime);
+                m_InitialToForwardDelayHist[respondingMach].add(
+                    forwardRequestTime - initialRequestTime);
+                m_ForwardToFirstResponseDelayHist[respondingMach].add(
+                    firstResponseTime - forwardRequestTime);
+                m_FirstResponseToCompletionDelayHist[respondingMach].add(
+                    completionTime - firstResponseTime);
+            } else {
+                m_IncompleteTimes[respondingMach]++;
+            }
+        }
+    } else {
+        m_hitLatencyHist.add(cycles);
+        m_hitTypeLatencyHist[type].add(cycles);
 
-void
-Sequencer::writeCallback(const Address& address,
-                         GenericMachineType mach,
-                         DataBlock& data)
-{
-    writeCallback(address, mach, data, Cycles(0), Cycles(0), Cycles(0));
+        if (respondingMach != MachineType_NUM) {
+            m_hitMachLatencyHist[respondingMach].add(cycles);
+            m_hitTypeMachLatencyHist[type][respondingMach].add(cycles);
+        }
+    }
 }
 
 void
-Sequencer::writeCallback(const Address& address,
-                         GenericMachineType mach,
-                         DataBlock& data,
-                         Cycles initialRequestTime,
-                         Cycles forwardRequestTime,
-                         Cycles firstResponseTime)
+Sequencer::writeCallback(const Address& address, DataBlock& data,
+                         const bool externalHit, const MachineType mach,
+                         const Cycles initialRequestTime,
+                         const Cycles forwardRequestTime,
+                         const Cycles firstResponseTime)
 {
     assert(address == line_address(address));
     assert(m_writeRequestTable.count(line_address(address)));
@@ -427,28 +514,13 @@ Sequencer::writeCallback(const Address& address,
         m_controller->unblock(address);
     }
 
-    hitCallback(request, mach, data, success,
+    hitCallback(request, data, success, mach, externalHit,
                 initialRequestTime, forwardRequestTime, firstResponseTime);
 }
 
 void
-Sequencer::readCallback(const Address& address, DataBlock& data)
-{
-    readCallback(address, GenericMachineType_NULL, data);
-}
-
-void
-Sequencer::readCallback(const Address& address,
-                        GenericMachineType mach,
-                        DataBlock& data)
-{
-    readCallback(address, mach, data, Cycles(0), Cycles(0), Cycles(0));
-}
-
-void
-Sequencer::readCallback(const Address& address,
-                        GenericMachineType mach,
-                        DataBlock& data,
+Sequencer::readCallback(const Address& address, DataBlock& data,
+                        bool externalHit, const MachineType mach,
                         Cycles initialRequestTime,
                         Cycles forwardRequestTime,
                         Cycles firstResponseTime)
@@ -466,18 +538,17 @@ Sequencer::readCallback(const Address& address,
     assert((request->m_type == RubyRequestType_LD) ||
            (request->m_type == RubyRequestType_IFETCH));
 
-    hitCallback(request, mach, data, true,
+    hitCallback(request, data, true, mach, externalHit,
                 initialRequestTime, forwardRequestTime, firstResponseTime);
 }
 
 void
-Sequencer::hitCallback(SequencerRequest* srequest,
-                       GenericMachineType mach,
-                       DataBlock& data,
-                       bool success,
-                       Cycles initialRequestTime,
-                       Cycles forwardRequestTime,
-                       Cycles firstResponseTime)
+Sequencer::hitCallback(SequencerRequest* srequest, DataBlock& data,
+                       bool llscSuccess,
+                       const MachineType mach, const bool externalHit,
+                       const Cycles initialRequestTime,
+                       const Cycles forwardRequestTime,
+                       const Cycles firstResponseTime)
 {
     PacketPtr pkt = srequest->pkt;
     Address request_address(pkt->getAddr());
@@ -494,29 +565,17 @@ Sequencer::hitCallback(SequencerRequest* srequest,
     }
 
     assert(curCycle() >= issued_time);
-    Cycles miss_latency = curCycle() - issued_time;
-
-    // Profile the miss latency for all non-zero demand misses
-    if (miss_latency != 0) {
-        g_system_ptr->getProfiler()->missLatency(miss_latency, type, mach);
+    Cycles total_latency = curCycle() - issued_time;
 
-        if (mach == GenericMachineType_L1Cache_wCC) {
-            g_system_ptr->getProfiler()->missLatencyWcc(issued_time,
-                initialRequestTime, forwardRequestTime,
-                firstResponseTime, curCycle());
-        }
-
-        if (mach == GenericMachineType_Directory) {
-            g_system_ptr->getProfiler()->missLatencyDir(issued_time,
-                initialRequestTime, forwardRequestTime,
-                firstResponseTime, curCycle());
-        }
+    // Profile the latency for all demand accesses.
+    recordMissLatency(total_latency, type, mach, externalHit, issued_time,
+                      initialRequestTime, forwardRequestTime,
+                      firstResponseTime, curCycle());
 
-        DPRINTFR(ProtocolTrace, "%15s %3s %10s%20s %6s>%-6s %s %d cycles\n",
-                 curTick(), m_version, "Seq",
-                 success ? "Done" : "SC_Failed", "", "",
-                 request_address, miss_latency);
-    }
+    DPRINTFR(ProtocolTrace, "%15s %3s %10s%20s %6s>%-6s %s %d cycles\n",
+             curTick(), m_version, "Seq",
+             llscSuccess ? "Done" : "SC_Failed", "", "",
+             request_address, total_latency);
 
     // update the data
     if (g_system_ptr->m_warmup_enabled) {
diff --git a/src/mem/ruby/system/Sequencer.hh b/src/mem/ruby/system/Sequencer.hh
index 058edb9ce..86e6aa2a9 100644
--- a/src/mem/ruby/system/Sequencer.hh
+++ b/src/mem/ruby/system/Sequencer.hh
@@ -32,7 +32,7 @@
 #include <iostream>
 
 #include "base/hashmap.hh"
-#include "mem/protocol/GenericMachineType.hh"
+#include "mem/protocol/MachineType.hh"
 #include "mem/protocol/RubyRequestType.hh"
 #include "mem/protocol/SequencerRequestType.hh"
 #include "mem/ruby/common/Address.hh"
@@ -65,36 +65,24 @@ class Sequencer : public RubyPort
 
     // Public Methods
     void wakeup(); // Used only for deadlock detection
-
     void printProgress(std::ostream& out) const;
-
     void clearStats();
 
-    void writeCallback(const Address& address, DataBlock& data);
-
     void writeCallback(const Address& address,
-                       GenericMachineType mach,
-                       DataBlock& data);
-
-    void writeCallback(const Address& address,
-                       GenericMachineType mach,
                        DataBlock& data,
-                       Cycles initialRequestTime,
-                       Cycles forwardRequestTime,
-                       Cycles firstResponseTime);
-
-    void readCallback(const Address& address, DataBlock& data);
+                       const bool externalHit = false,
+                       const MachineType mach = MachineType_NUM,
+                       const Cycles initialRequestTime = Cycles(0),
+                       const Cycles forwardRequestTime = Cycles(0),
+                       const Cycles firstResponseTime = Cycles(0));
 
     void readCallback(const Address& address,
-                      GenericMachineType mach,
-                      DataBlock& data);
-
-    void readCallback(const Address& address,
-                      GenericMachineType mach,
                       DataBlock& data,
-                      Cycles initialRequestTime,
-                      Cycles forwardRequestTime,
-                      Cycles firstResponseTime);
+                      const bool externalHit = false,
+                      const MachineType mach = MachineType_NUM,
+                      const Cycles initialRequestTime = Cycles(0),
+                      const Cycles forwardRequestTime = Cycles(0),
+                      const Cycles firstResponseTime = Cycles(0));
 
     RequestStatus makeRequest(PacketPtr pkt);
     bool empty() const;
@@ -118,19 +106,63 @@ class Sequencer : public RubyPort
     void recordRequestType(SequencerRequestType requestType);
     Histogram& getOutstandReqHist() { return m_outstandReqHist; }
 
+    Histogram& getLatencyHist() { return m_latencyHist; }
+    Histogram& getTypeLatencyHist(uint32_t t)
+    { return m_typeLatencyHist[t]; }
+
+    Histogram& getHitLatencyHist() { return m_hitLatencyHist; }
+    Histogram& getHitTypeLatencyHist(uint32_t t)
+    { return m_hitTypeLatencyHist[t]; }
+
+    Histogram& getHitMachLatencyHist(uint32_t t)
+    { return m_hitMachLatencyHist[t]; }
+
+    Histogram& getHitTypeMachLatencyHist(uint32_t r, uint32_t t)
+    { return m_hitTypeMachLatencyHist[r][t]; }
+
+    Histogram& getMissLatencyHist() { return m_missLatencyHist; }
+    Histogram& getMissTypeLatencyHist(uint32_t t)
+    { return m_missTypeLatencyHist[t]; }
+
+    Histogram& getMissMachLatencyHist(uint32_t t)
+    { return m_missMachLatencyHist[t]; }
+
+    Histogram& getMissTypeMachLatencyHist(uint32_t r, uint32_t t)
+    { return m_missTypeMachLatencyHist[r][t]; }
+
+    Histogram& getIssueToInitialDelayHist(uint32_t t)
+    { return m_IssueToInitialDelayHist[t]; }
+
+    Histogram& getInitialToForwardDelayHist(const MachineType t)
+    { return m_InitialToForwardDelayHist[t]; }
+
+    Histogram& getForwardRequestToFirstResponseHist(const MachineType t)
+    { return m_ForwardToFirstResponseDelayHist[t]; }
+
+    Histogram& getFirstResponseToCompletionDelayHist(const MachineType t)
+    { return m_FirstResponseToCompletionDelayHist[t]; }
+
+    const uint64_t getIncompleteTimes(const MachineType t) const
+    { return m_IncompleteTimes[t]; }
+
   private:
     void issueRequest(PacketPtr pkt, RubyRequestType type);
 
-    void hitCallback(SequencerRequest* request,
-                     GenericMachineType mach,
-                     DataBlock& data,
-                     bool success,
-                     Cycles initialRequestTime,
-                     Cycles forwardRequestTime,
-                     Cycles firstResponseTime);
+    void hitCallback(SequencerRequest* request, DataBlock& data,
+                     bool llscSuccess,
+                     const MachineType mach, const bool externalHit,
+                     const Cycles initialRequestTime,
+                     const Cycles forwardRequestTime,
+                     const Cycles firstResponseTime);
 
-    RequestStatus insertRequest(PacketPtr pkt, RubyRequestType request_type);
+    void recordMissLatency(const Cycles t, const RubyRequestType type,
+                           const MachineType respondingMach,
+                           bool isExternalHit, Cycles issuedTime,
+                           Cycles initialRequestTime,
+                           Cycles forwardRequestTime, Cycles firstResponseTime,
+                           Cycles completionTime);
 
+    RequestStatus insertRequest(PacketPtr pkt, RubyRequestType request_type);
     bool handleLlsc(const Address& address, SequencerRequest* request);
 
     // Private copy constructor and assignment operator
@@ -161,6 +193,38 @@ class Sequencer : public RubyPort
     //! Histogram for number of outstanding requests per cycle.
     Histogram m_outstandReqHist;
 
+    //! Histogram for holding latency profile of all requests.
+    Histogram m_latencyHist;
+    std::vector<Histogram> m_typeLatencyHist;
+
+    //! Histogram for holding latency profile of all requests that
+    //! hit in the controller connected to this sequencer.
+    Histogram m_hitLatencyHist;
+    std::vector<Histogram> m_hitTypeLatencyHist;
+
+    //! Histograms for profiling the latencies for requests that
+    //! did not required external messages.
+    std::vector<Histogram> m_hitMachLatencyHist;
+    std::vector< std::vector<Histogram> > m_hitTypeMachLatencyHist;
+
+    //! Histogram for holding latency profile of all requests that
+    //! miss in the controller connected to this sequencer.
+    Histogram m_missLatencyHist;
+    std::vector<Histogram> m_missTypeLatencyHist;
+
+    //! Histograms for profiling the latencies for requests that
+    //! required external messages.
+    std::vector<Histogram> m_missMachLatencyHist;
+    std::vector< std::vector<Histogram> > m_missTypeMachLatencyHist;
+
+    //! Histograms for recording the breakdown of miss latency
+    std::vector<Histogram> m_IssueToInitialDelayHist;
+    std::vector<Histogram> m_InitialToForwardDelayHist;
+    std::vector<Histogram> m_ForwardToFirstResponseDelayHist;
+    std::vector<Histogram> m_FirstResponseToCompletionDelayHist;
+    std::vector<uint64_t> m_IncompleteTimes;
+
+
     class SequencerWakeupEvent : public Event
     {
       private: