Subzero: Fix stats collection and output for multithreading.

src/IceGlobalContext.h

Index: src/IceGlobalContext.h
diff --git a/src/IceGlobalContext.h b/src/IceGlobalContext.h
index 04f08a2dbffa3191633fb39343b9e6e03f1ef1e0..0a1ecccd29badf0e17a8da5bf767c6be712c6ab4 100644
--- a/src/IceGlobalContext.h
+++ b/src/IceGlobalContext.h
@@ -73,6 +73,13 @@ class GlobalContext {
     void updateFrameBytes(uint32_t Bytes) { FrameBytes += Bytes; }
     void updateSpills() { ++Spills; }
     void updateFills() { ++Fills; }
+    void add(const CodeStats &Other) {
+      InstructionsEmitted += Other.InstructionsEmitted;
+      RegistersSaved += Other.RegistersSaved;
+      FrameBytes += Other.FrameBytes;
+      Spills += Other.Spills;
+      Fills += Other.Fills;
+    }
     void dump(const IceString &Name, Ostream &Str);
 
   private:
@@ -121,6 +128,7 @@ class GlobalContext {
   public:
     ThreadContext() {}
     CodeStats StatsFunction;
+    CodeStats StatsCumulative;
     TimerList Timers;
   };
 
@@ -209,32 +217,37 @@ public:
   void statsUpdateEmitted(uint32_t InstCount) {
     if (!ALLOW_DUMP || !getFlags().DumpStats)
       return;
-    ICE_TLS_GET_FIELD(TLS)->StatsFunction.updateEmitted(InstCount);
-    getStatsCumulative()->updateEmitted(InstCount);
+    auto TLS = ICE_TLS_GET_FIELD(TLS);

[JF, 2015/01/31 05:48:38]

auto *TLS = ...;

[Jim Stichnoth, 2015/01/31 16:42:28]

Done, took it one step further...

My original idea was that when we eventually get rid of these ICE_TLS macros,
this initialization would reduce to "auto TLS = TLS;" and ultimately be removed
entirely without any surrounding code modifications, so using the simplest
possible "auto" initialization would make it easier to come to that conclusion. 
But, that's probably overthinking it so I put back the full declaration.

+    TLS->StatsFunction.updateEmitted(InstCount);
+    TLS->StatsCumulative.updateEmitted(InstCount);
   }
   void statsUpdateRegistersSaved(uint32_t Num) {
     if (!ALLOW_DUMP || !getFlags().DumpStats)
       return;
-    ICE_TLS_GET_FIELD(TLS)->StatsFunction.updateRegistersSaved(Num);
-    getStatsCumulative()->updateRegistersSaved(Num);
+    auto TLS = ICE_TLS_GET_FIELD(TLS);
+    TLS->StatsFunction.updateRegistersSaved(Num);
+    TLS->StatsCumulative.updateRegistersSaved(Num);
   }
   void statsUpdateFrameBytes(uint32_t Bytes) {
     if (!ALLOW_DUMP || !getFlags().DumpStats)
       return;
-    ICE_TLS_GET_FIELD(TLS)->StatsFunction.updateFrameBytes(Bytes);
-    getStatsCumulative()->updateFrameBytes(Bytes);
+    auto TLS = ICE_TLS_GET_FIELD(TLS);
+    TLS->StatsFunction.updateFrameBytes(Bytes);
+    TLS->StatsCumulative.updateFrameBytes(Bytes);
   }
   void statsUpdateSpills() {
     if (!ALLOW_DUMP || !getFlags().DumpStats)
       return;
-    ICE_TLS_GET_FIELD(TLS)->StatsFunction.updateSpills();
-    getStatsCumulative()->updateSpills();
+    auto TLS = ICE_TLS_GET_FIELD(TLS);
+    TLS->StatsFunction.updateSpills();
+    TLS->StatsCumulative.updateSpills();
   }
   void statsUpdateFills() {
     if (!ALLOW_DUMP || !getFlags().DumpStats)
       return;
-    ICE_TLS_GET_FIELD(TLS)->StatsFunction.updateFills();
-    getStatsCumulative()->updateFills();
+    auto TLS = ICE_TLS_GET_FIELD(TLS);
+    TLS->StatsFunction.updateFills();
+    TLS->StatsCumulative.updateFills();
   }
 
   // These are predefined TimerStackIdT values.
@@ -296,6 +309,13 @@ public:
       for (ThreadContext *TLS : AllThreadContexts)
         Timers->mergeFrom(TLS->Timers);
     }
+    if (ALLOW_DUMP) {
+      // Do a separate loop over AllThreadContexts to avoid holding
+      // two locks at once.
+      auto Stats = getStatsCumulative();
+      for (ThreadContext *TLS : AllThreadContexts)
+        Stats->add(TLS->StatsCumulative);
+    }
   }
 
   // Translation thread startup routine.