ruby: patch checkpoint restore with garnet

Due to recent changes to clocking system in Ruby and the way Ruby restores
state from a checkpoint, garnet was failing to run from a checkpointed state.
The problem is that Ruby resets the time to zero while warming up the caches.
If any component records a local copy of the time (read calls curCycle())
before the simulation has started, then that component will not operate until
that time is reached. In the context of this particular patch, the Garnet
Network class calls curCycle() at multiple places. Any non-operational
component can block in requests in the memory system, which the system
interprets as a deadlock. This patch makes changes so that Garnet can
successfully run from checkpointed state.

It adds a globally visible time at which the actual execution started. This
time is initialized in RubySystem::startup() function. This variable is only
meant for components with in Ruby. This replaces the private variable that
was maintained within Garnet since it is not possible to figure out the
correct time when the value of this variable can be set.

The patch also does away with all cases where curCycle() is called with in
some Ruby component before the system has actually started executing. This
is required due to the quirky manner in which ruby restores from a checkpoint.

1 files changed, 21 insertions, 13 deletions

diff --git a/src/mem/ruby/system/System.cc b/src/mem/ruby/system/System.cc
index 617788b99..357511127 100644
--- a/src/mem/ruby/system/System.cc
+++ b/src/mem/ruby/system/System.cc
@@ -94,13 +94,6 @@ RubySystem::RubySystem(const Params *p)
 }
 
 void
-RubySystem::init()
-{
-    m_profiler_ptr->clearStats();
-    m_network_ptr->clearStats();
-}
-
-void
 RubySystem::registerNetwork(Network* network_ptr)
 {
   m_network_ptr = network_ptr;
@@ -311,12 +304,6 @@ RubySystem::readCompressedTrace(string filename, uint8_t *&raw_data,
 void
 RubySystem::unserialize(Checkpoint *cp, const string &section)
 {
-    //
-    // The main purpose for clearing stats in the unserialize process is so
-    // that the profiler can correctly set its start time to the unserialized
-    // value of curTick()
-    //
-    resetStats();
     uint8_t *uncompressed_trace = NULL;
 
     if (m_mem_vec_ptr != NULL) {
@@ -368,6 +355,23 @@ RubySystem::unserialize(Checkpoint *cp, const string &section)
 void
 RubySystem::startup()
 {
+
+    // Ruby restores state from a checkpoint by resetting the clock to 0 and
+    // playing the requests that can possibly re-generate the cache state.
+    // The clock value is set to the actual checkpointed value once all the
+    // requests have been executed.
+    //
+    // This way of restoring state is pretty finicky. For example, if a
+    // Ruby component reads time before the state has been restored, it would
+    // cache this value and hence its clock would not be reset to 0, when
+    // Ruby resets the global clock. This can potentially result in a
+    // deadlock.
+    //
+    // The solution is that no Ruby component should read time before the
+    // simulation starts. And then one also needs to hope that the time
+    // Ruby finishes restoring the state is less than the time when the
+    // state was checkpointed.
+
     if (m_warmup_enabled) {
         // save the current tick value
         Tick curtick_original = curTick();
@@ -397,6 +401,8 @@ RubySystem::startup()
         setCurTick(curtick_original);
         resetClock();
     }
+
+    resetStats();
 }
 
 void
@@ -417,6 +423,8 @@ RubySystem::resetStats()
     for (uint32_t cntrl = 0; cntrl < m_abs_cntrl_vec.size(); cntrl++) {
         m_abs_cntrl_vec[cntrl]->clearStats();
     }
+
+    g_ruby_start = curCycle();
 }
 
 bool