/* -*- Mode: C++; c-basic-offset: 4; indent-tabs-mode: nil; tab-width: 4 -*- */ /* vi: set ts=4 sw=4 expandtab: (add to ~/.vimrc: set modeline modelines=5) */ /* ***** BEGIN LICENSE BLOCK ***** * Version: MPL 1.1/GPL 2.0/LGPL 2.1 * * The contents of this file are subject to the Mozilla Public License Version * 1.1 (the "License"); you may not use this file except in compliance with * the License. You may obtain a copy of the License at * http://www.mozilla.org/MPL/ * * Software distributed under the License is distributed on an "AS IS" basis, * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License * for the specific language governing rights and limitations under the * License. * * The Original Code is [Open Source Virtual Machine]. * * The Initial Developer of the Original Code is * Adobe System Incorporated. * Portions created by the Initial Developer are Copyright (C) 2004-2007 * the Initial Developer. All Rights Reserved. * * Contributor(s): * Adobe AS3 Team * Mozilla TraceMonkey Team * Asko Tontti * * Alternatively, the contents of this file may be used under the terms of * either the GNU General Public License Version 2 or later (the "GPL"), or * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"), * in which case the provisions of the GPL or the LGPL are applicable instead * of those above. If you wish to allow use of your version of this file only * under the terms of either the GPL or the LGPL, and not to allow others to * use your version of this file under the terms of the MPL, indicate your * decision by deleting the provisions above and replace them with the notice * and other provisions required by the GPL or the LGPL. If you do not delete * the provisions above, a recipient may use your version of this file under * the terms of any one of the MPL, the GPL or the LGPL. * * ***** END LICENSE BLOCK ***** */ #ifndef __nanojit_Fragmento__ #define __nanojit_Fragmento__ namespace nanojit { struct GuardRecord; /** * Fragments are linear sequences of native code that have a single entry * point at the start of the fragment and may have one or more exit points * * It may turn out that that this arrangement causes too much traffic * between d and i-caches and that we need to carve up the structure differently. */ class Fragment { public: Fragment(const void* verbose_only(, uint32_t profFragID)); NIns* code() { return _code; } void setCode(NIns* codee) { _code = codee; } int32_t& hits() { return _hits; } LirBuffer* lirbuf; LIns* lastIns; const void* ip; uint32_t recordAttempts; NIns* fragEntry; // for fragment entry and exit profiling. See detailed // how-to-use comment below. verbose_only( LIns* loopLabel; ) // where's the loop top? verbose_only( uint32_t profFragID; ) verbose_only( uint32_t profCount; ) verbose_only( uint32_t nStaticExits; ) verbose_only( size_t nCodeBytes; ) verbose_only( size_t nExitBytes; ) verbose_only( uint32_t guardNumberer; ) verbose_only( GuardRecord* guardsForFrag; ) private: NIns* _code; // ptr to start of code int32_t _hits; }; } /* * How to use fragment profiling * * Fragprofiling adds code to count how many times each fragment is * entered, and how many times each guard (exit) is taken. Using this * it's possible to easily find which fragments are hot, which ones * typically exit early, etc. The fragprofiler also gathers some * simple static info: for each fragment, the number of code bytes, * number of exit-block bytes, and number of guards (exits). * * Fragments and guards are given unique IDs (FragID, GuardID) which * are shown in debug printouts, so as to facilitate navigating from * the accumulated statistics to the associated bits of code. * GuardIDs are issued automatically, but FragIDs you must supply when * calling Fragment::Fragment. Supply values >= 1, and supply a * different value for each new fragment (doesn't matter what, they * just have to be unique and >= 1); else * js_FragProfiling_FragFinalizer will assert. * * How to use/embed: * * - use a debug build (one with NJ_VERBOSE). Without it, none of * this code is compiled in. * * - set LC_FragProfile in the lcbits of the LogControl* object handed * to Nanojit * * When enabled, Fragment::profCount is incremented every time the * fragment is entered, and GuardRecord::profCount is incremented * every time that guard exits. However, NJ has no way to know where * the fragment entry/loopback point is. So you must set * Fragment::loopLabel before running the assembler, so as to indicate * where the fragment-entry counter increment should be placed. If * the fragment does not naturally have a loop label then you will * need to artificially add one. * * It is the embedder's problem to fish out, collate and present the * accumulated stats at the end of the Fragment's lifetime. A * Fragment contains stats indicating its entry count and static code * sizes. It also has a ::guardsForFrag field, which is a linked list * of GuardRecords, and by traversing them you can get hold of the * exit counts. */ #endif // __nanojit_Fragmento__