--- parser3/INSTALL	2005/12/29 09:11:11	1.48.2.3
+++ parser3/INSTALL	2009/05/24 12:36:03	1.51
@@ -1,5 +1,5 @@
 Alexander Petrosian <paf@design.ru> (http://paf.design.ru)
-$Id: INSTALL,v 1.48.2.3 2005/12/29 09:11:11 paf Exp $
+$Id: INSTALL,v 1.51 2009/05/24 12:36:03 misha Exp $
 
 1.What is the process to compile Parser3?
 
@@ -126,3 +126,39 @@ $Id: INSTALL,v 1.48.2.3 2005/12/29 09:11
 
   WARNING: if you want to use apache configure option --enable-shared=max, do this:
   ./p3runConfigure --enable-shared=max --disable-shared=mod_parser3.c
+
+6.
+  Since Parser 3.4.0 the several optimisations were implemented.
+  If you experience problems you can try to disable them by commenting corresponding defines and recompiling parser3.
+
+  src/include/pa_opcode.h
+    #define OPTIMIZE_BYTECODE_GET_CLASS               -- $a:
+    #define OPTIMIZE_BYTECODE_GET_ELEMENT             -- $a ^a
+    #define OPTIMIZE_BYTECODE_CUT_REM_OPERATOR        -- cut rem operator with any number of params during compilation
+    //#define OPTIMIZE_BYTECODE_STRING_POOL           -- simplifying string's bytecode into expression
+    #define OPTIMIZE_BYTECODE_GET_OBJECT_ELEMENT      -- $a.b ^a.b
+    #define OPTIMIZE_BYTECODE_GET_OBJECT_VAR_ELEMENT  -- $a.$b ^a.$b
+    #define OPTIMIZE_BYTECODE_CONSTRUCT               -- $a(1),       $.a(1),       $a[b],       $.a[b]
+                                                         $a($b),      $.a($b),      $a[$b],      $.a[$b]
+                                                         $a($b.c),    $.a($b.c),    $a[$b.c],    $.a[$b.c]
+                                                         $a($b.$c),   $.a($b.$c),   $a[$b.$c],   $.a[$b.$c]
+
+  src/include/pa_string.h
+    #define HASH_CODE_CACHING  -- calculated hash codes are cached and used for sequential hash lookups
+
+  src/include/pa_memory.h
+    #define USE_DESTRUCTORS    -- destructors are used to decrease memory consumption during code
+                                  processing and make ^memory:compact[] calls non-essential.
+
+  src/types/pa_method.h
+    #define OPTIMIZE_CALL      -- allows faster operators execution by eliminating method frame
+                                  with local variables creation and extra write context switches.
+    #define OPTIMIZE_RESULT    -- parser methods are marked when the $result variable was used.
+                                  This allows write operations and context switching optimization.
+
+  src/lib/cord/include/cord.h
+    #define CORD_CAT_OPTIMIZATION -- CORD library never modifies source concatenations. But in parser
+                                     write operations it is safe to modify them and save some memory.
+    #define CORD_CHARS_CACHE      -- language cords with same language and length are cached and reused.
+
+