/** * @file lscript_alloc.h * @brief General heap management for scripting system * * $LicenseInfo:firstyear=2002&license=viewergpl$ * * Copyright (c) 2002-2008, Linden Research, Inc. * * Second Life Viewer Source Code * The source code in this file ("Source Code") is provided by Linden Lab * to you under the terms of the GNU General Public License, version 2.0 * ("GPL"), unless you have obtained a separate licensing agreement * ("Other License"), formally executed by you and Linden Lab. Terms of * the GPL can be found in doc/GPL-license.txt in this distribution, or * online at http://secondlife.com/developers/opensource/gplv2 * * There are special exceptions to the terms and conditions of the GPL as * it is applied to this Source Code. View the full text of the exception * in the file doc/FLOSS-exception.txt in this software distribution, or * online at http://secondlife.com/developers/opensource/flossexception * * By copying, modifying or distributing this software, you acknowledge * that you have read and understood your obligations described above, * and agree to abide by those obligations. * * ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO * WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY, * COMPLETENESS OR PERFORMANCE. * $/LicenseInfo$ */ #ifndef LL_LSCRIPT_ALLOC_H #define LL_LSCRIPT_ALLOC_H // #define at top of file accelerates gcc compiles // Under gcc 2.9, the manual is unclear if comments can appear above #ifndef // Under gcc 3, the manual explicitly states comments can appear above the #ifndef #include "lscript_byteconvert.h" #include "lscript_library.h" void reset_hp_to_safe_spot(const U8 *buffer); // supported data types // basic types // integer 4 bytes of integer data // float 4 bytes of float data // string data null terminated 1 byte string // key data null terminated 1 byte string // vector data 12 bytes of 3 floats // quaternion data 16 bytes of 4 floats // list type // list data 4 bytes of number of entries followed by followed by pointer // string pointer 4 bytes of address of string data on the heap (only used in list data) // key pointer 4 bytes of address of key data on the heap (only used in list data) // heap format // // 4 byte offset to next block (in bytes) // 1 byte of type of variable or empty // 2 bytes of reference count // nn bytes of data const S32 MAX_HEAP_SIZE = TOP_OF_MEMORY; class LLScriptAllocEntry { public: LLScriptAllocEntry() : mSize(0), mType(LST_NULL), mReferenceCount(0) {} LLScriptAllocEntry(S32 offset, U8 type) : mSize(offset), mType(type), mReferenceCount(1) {} friend std::ostream& operator<<(std::ostream& s, const LLScriptAllocEntry &a) { s << "Size: " << a.mSize << " Type: " << LSCRIPTTypeNames[a.mType] << " Count: " << a.mReferenceCount; return s; } S32 mSize; U8 mType; S16 mReferenceCount; }; // this is only OK because we only load/save via accessors below const S32 SIZEOF_SCRIPT_ALLOC_ENTRY = 7; inline void alloc_entry2bytestream(U8 *buffer, S32 &offset, const LLScriptAllocEntry &entry) { if ( (offset < 0) ||(offset > MAX_HEAP_SIZE)) { set_fault(buffer, LSRF_BOUND_CHECK_ERROR); } else { integer2bytestream(buffer, offset, entry.mSize); byte2bytestream(buffer, offset, entry.mType); s162bytestream(buffer, offset, entry.mReferenceCount); } } inline void bytestream2alloc_entry(LLScriptAllocEntry &entry, U8 *buffer, S32 &offset) { if ( (offset < 0) ||(offset > MAX_HEAP_SIZE)) { set_fault(buffer, LSRF_BOUND_CHECK_ERROR); reset_hp_to_safe_spot(buffer); } else { entry.mSize = bytestream2integer(buffer, offset); entry.mType = bytestream2byte(buffer, offset); entry.mReferenceCount = bytestream2s16(buffer, offset); } } // create a heap from the HR to TM BOOL lsa_create_heap(U8 *heap_start, S32 size); void lsa_fprint_heap(U8 *buffer, FILE *fp); void lsa_print_heap(U8 *buffer); // adding to heap // if block is empty // if block is at least block size + 4 larger than data // split block // insert data into first part // return address // else // insert data into block // return address // else // if next block is >= SP // set Stack-Heap collision // return NULL // if next block is empty // merge next block with current block // go to start of algorithm // else // move to next block // go to start of algorithm S32 lsa_heap_add_data(U8 *buffer, LLScriptLibData *data, S32 heapsize, BOOL b_delete); S32 lsa_heap_top(U8 *heap_start, S32 maxsize); // split block // set offset to point to new block // set offset of new block to point to original offset - block size - data size // set new block to empty // set new block reference count to 0 void lsa_split_block(U8 *buffer, S32 &offset, S32 size, LLScriptAllocEntry &entry); // insert data // if data is non-list type // set type to basic type, set reference count to 1, copy data, return address // else // set type to list data type, set reference count to 1 // for each list entry // insert data // return address void lsa_insert_data(U8 *buffer, S32 &offset, LLScriptLibData *data, LLScriptAllocEntry &entry, S32 heapsize); S32 lsa_create_data_block(U8 **buffer, LLScriptLibData *data, S32 base_offset); // increase reference count // increase reference count by 1 void lsa_increase_ref_count(U8 *buffer, S32 offset); // decrease reference count // decrease reference count by 1 // if reference count == 0 // set type to empty void lsa_decrease_ref_count(U8 *buffer, S32 offset); inline S32 get_max_heap_size(U8 *buffer) { return get_register(buffer, LREG_SP) - get_register(buffer, LREG_HR); } LLScriptLibData *lsa_get_data(U8 *buffer, S32 &offset, BOOL b_dec_ref); LLScriptLibData *lsa_get_list_ptr(U8 *buffer, S32 &offset, BOOL b_dec_ref); S32 lsa_cat_strings(U8 *buffer, S32 offset1, S32 offset2, S32 heapsize); S32 lsa_cmp_strings(U8 *buffer, S32 offset1, S32 offset2); S32 lsa_cat_lists(U8 *buffer, S32 offset1, S32 offset2, S32 heapsize); S32 lsa_cmp_lists(U8 *buffer, S32 offset1, S32 offset2); S32 lsa_preadd_lists(U8 *buffer, LLScriptLibData *data, S32 offset2, S32 heapsize); S32 lsa_postadd_lists(U8 *buffer, S32 offset1, LLScriptLibData *data, S32 heapsize); // modifying a list // insert new list that is modified // store returned address in original list's variable // decrease reference count on old list // list l1 = [10]; // list l2 = l1; // l1 = [11]; // we want l2 == [10]; // more complicated example: // list l1 = [10, 11]; // list l2 = l1; // l1[0] = 12 // I think that we want l2 = [10, 11]; // one option would be to use syntax like: // l1 = llSetList(l1, 0, 12); // but this would require variable argument list matching // which maybe is ok, but would be work // the other option would be changes to lists that have multiple references causes a copy to occur // popl @l1, 0, integer, 12 // // would cause l1 to be copied, 12 to replace the 0th entry, and the address of the new list to be saved in l1 // inline LLScriptLibData *lsa_bubble_sort(LLScriptLibData *src, S32 stride, S32 ascending) { S32 number = src->getListLength(); if (number <= 0) { return NULL; } if (stride <= 0) { stride = 1; } S32 i = 0; if (number % stride) { LLScriptLibData *retval = src->mListp; src->mListp = NULL; return retval; } LLScriptLibData **sortarray = new LLScriptLibData*[number]; LLScriptLibData *temp = src->mListp; while (temp) { sortarray[i] = temp; i++; temp = temp->mListp; } S32 j, s; for (i = 0; i < number; i += stride) { for (j = i; j < number; j += stride) { if ( ((*sortarray[i]) <= (*sortarray[j])) != (ascending == TRUE)) { for (s = 0; s < stride; s++) { temp = sortarray[i + s]; sortarray[i + s] = sortarray[j + s]; sortarray[j + s] = temp; } } } } i = 1; temp = sortarray[0]; while (i < number) { temp->mListp = sortarray[i++]; temp = temp->mListp; } temp->mListp = NULL; src->mListp = NULL; temp = sortarray[0]; delete[] sortarray; return temp; } LLScriptLibData* lsa_randomize(LLScriptLibData* src, S32 stride); #endif