misched: Heuristics based on the machine model.

misched is disabled by default. With -enable-misched, these heuristics
balance the schedule to simultaneously avoid saturating processor
resources, expose ILP, and minimize register pressure. I've been
analyzing the performance of these heuristics on everything in the
llvm test suite in addition to a few other benchmarks. I would like
each heuristic check to be verified by a unit test, but I'm still
trying to figure out the best way to do that. The heuristics are still
in considerable flux, but as they are refined we should be rigorous
about unit testing the improvements.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167527 91177308-0d34-0410-b5e6-96231b3b80d8

include/llvm/CodeGen/MachineScheduler.h

@@ -154,6 +154,8 @@ public:
 
   bool empty() const { return Queue.empty(); }
 
+  void clear() { Queue.clear(); }
+
   unsigned size() const { return Queue.size(); }
 
   typedef std::vector<SUnit*>::iterator iterator;
@@ -171,10 +173,12 @@ public:
     SU->NodeQueueId |= ID;
   }
 
-  void remove(iterator I) {
+  iterator remove(iterator I) {
     (*I)->NodeQueueId &= ~ID;
     *I = Queue.back();
+    unsigned idx = I - Queue.begin();
     Queue.pop_back();
+    return Queue.begin() + idx;
   }
 
 #ifndef NDEBUG
@@ -306,6 +310,9 @@ protected:
   /// Reinsert debug_values recorded in ScheduleDAGInstrs::DbgValues.
   void placeDebugValues();
 
+  /// \brief dump the scheduled Sequence.
+  void dumpSchedule() const;
+
   // Lesser helpers...
 
   void initRegPressure();