diff --git a/bench/benchmain.cpp b/bench/benchmain.cpp
index 73aa988..fa43155 100644
--- a/bench/benchmain.cpp
+++ b/bench/benchmain.cpp
@@ -439,8 +439,11 @@
 }
 
 int main (int argc, char * const argv[]) {
+#ifdef SK_ENABLE_INST_COUNT
+    gPrintInstCount = true;
+#endif
     SkAutoGraphics ag;
-    
+
     SkTDict<const char*> defineDict(1024);
     int repeatDraw = 1;
     bool logPerIter = false;
diff --git a/include/core/SkPath.h b/include/core/SkPath.h
index 2bd2d76..ae8ecb7 100644
--- a/include/core/SkPath.h
+++ b/include/core/SkPath.h
@@ -10,6 +10,7 @@
 #ifndef SkPath_DEFINED
 #define SkPath_DEFINED
 
+#include "SkInstCnt.h"
 #include "SkMatrix.h"
 #include "SkTDArray.h"
 
@@ -33,6 +34,8 @@
 */
 class SK_API SkPath {
 public:
+    SK_DECLARE_INST_COUNT_ROOT(SkPath);
+
     SkPath();
     SkPath(const SkPath&);
     ~SkPath();
diff --git a/src/core/SkPath.cpp b/src/core/SkPath.cpp
index 1127e95..a9937ab 100644
--- a/src/core/SkPath.cpp
+++ b/src/core/SkPath.cpp
@@ -11,6 +11,8 @@
 #include "SkBuffer.h"
 #include "SkMath.h"
 
+SK_DEFINE_INST_COUNT(SkPath);
+
 // This value is just made-up for now. When count is 4, calling memset was much
 // slower than just writing the loop. This seems odd, and hopefully in the
 // future this we appear to have been a fluke...