Several optimizations and speed improvements.
cstubs: Use Matrix type instead of float[4][4].
diff --git a/Include/stringobject.h b/Include/stringobject.h
index a4b4ea6..fa2e8a1 100644
--- a/Include/stringobject.h
+++ b/Include/stringobject.h
@@ -51,6 +51,9 @@
 
 typedef struct {
 	OB_VARHEAD
+#ifdef CACHE_HASH
+	long ob_shash;
+#endif
 	char ob_sval[1];
 } stringobject;