[692] | 1 | /*
|
---|
| 2 | -----------------------------------------------------------------------------
|
---|
| 3 | This source file is part of OGRE
|
---|
| 4 | (Object-oriented Graphics Rendering Engine)
|
---|
| 5 | For the latest info, see http://www.ogre3d.org/
|
---|
| 6 |
|
---|
| 7 | Copyright (c) 2000-2005 The OGRE Team
|
---|
| 8 | Also see acknowledgements in Readme.html
|
---|
| 9 |
|
---|
| 10 | This program is free software; you can redistribute it and/or modify it under
|
---|
| 11 | the terms of the GNU Lesser General Public License as published by the Free Software
|
---|
| 12 | Foundation; either version 2 of the License, or (at your option) any later
|
---|
| 13 | version.
|
---|
| 14 |
|
---|
| 15 | This program is distributed in the hope that it will be useful, but WITHOUT
|
---|
| 16 | ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
|
---|
| 17 | FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.
|
---|
| 18 |
|
---|
| 19 | You should have received a copy of the GNU Lesser General Public License along with
|
---|
| 20 | this program; if not, write to the Free Software Foundation, Inc., 59 Temple
|
---|
| 21 | Place - Suite 330, Boston, MA 02111-1307, USA, or go to
|
---|
| 22 | http://www.gnu.org/copyleft/lesser.txt.
|
---|
| 23 | -----------------------------------------------------------------------------
|
---|
| 24 | */
|
---|
| 25 | #ifndef __RadixSort_H__
|
---|
| 26 | #define __RadixSort_H__
|
---|
| 27 |
|
---|
| 28 | #include "OgrePrerequisites.h"
|
---|
| 29 |
|
---|
| 30 | namespace Ogre {
|
---|
| 31 |
|
---|
| 32 | /** Class for performing a radix sort (fast comparison-less sort based on
|
---|
| 33 | byte value) on various standard STL containers.
|
---|
| 34 | @remarks
|
---|
| 35 | A radix sort is a very fast sort algorithm. It doesn't use comparisons
|
---|
| 36 | and thus is able to break the theoretical minimum O(N*logN) complexity.
|
---|
| 37 | Radix sort is complexity O(k*N), where k is a constant. Note that radix
|
---|
| 38 | sorting is not in-place, it requires additional storage, so it trades
|
---|
| 39 | memory for speed. The overhead of copying means that it is only faster
|
---|
| 40 | for fairly large datasets, so you are advised to only use it for collections
|
---|
| 41 | of at least a few hundred items.
|
---|
| 42 | @par
|
---|
| 43 | This is a template class to allow it to deal with a variety of containers,
|
---|
| 44 | and a variety of value types to sort on. In addition to providing the
|
---|
| 45 | container and value type on construction, you also need to supply a
|
---|
| 46 | functor object which will retrieve the value to compare on for each item
|
---|
| 47 | in the list. For example, if you had an std::vector of by-value instances
|
---|
| 48 | of an object of class 'Bibble', and you wanted to sort on
|
---|
| 49 | Bibble::getDoobrie(), you'd have to firstly create a functor
|
---|
| 50 | like this:
|
---|
| 51 | @code
|
---|
| 52 | struct BibbleSortFunctor
|
---|
| 53 | {
|
---|
| 54 | float operator()(const Bibble& val) const
|
---|
| 55 | {
|
---|
| 56 | return val.getDoobrie();
|
---|
| 57 | }
|
---|
| 58 | }
|
---|
| 59 | @endcode
|
---|
| 60 | Then, you need to declare a RadixSort class which names the container type,
|
---|
| 61 | the value type in the container, and the type of the value you want to
|
---|
| 62 | sort by. You can then call the sort function. E.g.
|
---|
| 63 | @code
|
---|
| 64 | RadixSort<BibbleList, Bibble, float> radixSorter;
|
---|
| 65 | BibbleSortFunctor functor;
|
---|
| 66 |
|
---|
| 67 | radixSorter.sort(myBibbleList, functor);
|
---|
| 68 | @endcode
|
---|
| 69 | You should try to reuse RadixSort instances, since repeated allocation of the
|
---|
| 70 | internal storage is then avoided.
|
---|
| 71 | @note
|
---|
| 72 | Radix sorting is often associated with just unsigned integer values. Our
|
---|
| 73 | implementation can handle both unsigned and signed integers, as well as
|
---|
| 74 | floats (which are often not supported by other radix sorters). doubles
|
---|
| 75 | are not supported; you will need to implement your functor object to convert
|
---|
| 76 | to float if you wish to use this sort routine.
|
---|
| 77 | */
|
---|
| 78 | template <class TContainer, class TContainerValueType, typename TCompValueType>
|
---|
| 79 | class RadixSort
|
---|
| 80 | {
|
---|
| 81 | public:
|
---|
| 82 | typedef typename TContainer::iterator ContainerIter;
|
---|
| 83 | protected:
|
---|
| 84 | /// Alpha-pass counters of values (histogram)
|
---|
| 85 | /// 4 of them so we can radix sort a maximum of a 32bit value
|
---|
| 86 | int mCounters[4][256];
|
---|
| 87 | /// Beta-pass offsets
|
---|
| 88 | int mOffsets[256];
|
---|
| 89 | /// Sort area size
|
---|
| 90 | int mSortSize;
|
---|
| 91 | /// Number of passes for this type
|
---|
| 92 | int mNumPasses;
|
---|
| 93 |
|
---|
| 94 | struct SortEntry
|
---|
| 95 | {
|
---|
| 96 | TCompValueType key;
|
---|
| 97 | ContainerIter iter;
|
---|
| 98 | SortEntry() {}
|
---|
| 99 | SortEntry(TCompValueType k, ContainerIter it)
|
---|
| 100 | : key(k), iter(it) {}
|
---|
| 101 |
|
---|
| 102 | };
|
---|
| 103 | /// Temp sort storage
|
---|
| 104 | std::vector<SortEntry> mSortArea1;
|
---|
| 105 | std::vector<SortEntry> mSortArea2;
|
---|
| 106 | std::vector<SortEntry>* mSrc;
|
---|
| 107 | std::vector<SortEntry>* mDest;
|
---|
| 108 | TContainer mTmpContainer; // initial copy
|
---|
| 109 |
|
---|
| 110 |
|
---|
| 111 | void sortPass(int byteIndex)
|
---|
| 112 | {
|
---|
| 113 | // Calculate offsets
|
---|
| 114 | // Basically this just leaves gaps for duplicate entries to fill
|
---|
| 115 | mOffsets[0] = 0;
|
---|
| 116 | for (int i = 1; i < 256; ++i)
|
---|
| 117 | {
|
---|
| 118 | mOffsets[i] = mOffsets[i-1] + mCounters[byteIndex][i-1];
|
---|
| 119 | }
|
---|
| 120 |
|
---|
| 121 | // Sort pass
|
---|
| 122 | for (int i = 0; i < mSortSize; ++i)
|
---|
| 123 | {
|
---|
| 124 | unsigned char byteVal = getByte(byteIndex, (*mSrc)[i].key);
|
---|
| 125 | (*mDest)[mOffsets[byteVal]++] = (*mSrc)[i];
|
---|
| 126 | }
|
---|
| 127 |
|
---|
| 128 | }
|
---|
| 129 | template <typename T>
|
---|
| 130 | void finalPass(int byteIndex, T val)
|
---|
| 131 | {
|
---|
| 132 | // default is to do normal pass
|
---|
| 133 | sortPass(byteIndex);
|
---|
| 134 | }
|
---|
| 135 |
|
---|
| 136 | // special case signed int
|
---|
| 137 | void finalPass(int byteIndex, int val)
|
---|
| 138 | {
|
---|
| 139 | int numNeg = 0;
|
---|
| 140 | // all negative values are in entries 128+ in most significant byte
|
---|
| 141 | for (int i = 128; i < 256; ++i)
|
---|
| 142 | {
|
---|
| 143 | numNeg += mCounters[byteIndex][i];
|
---|
| 144 | }
|
---|
| 145 | // Calculate offsets - positive ones start at the number of negatives
|
---|
| 146 | // do positive numbers
|
---|
| 147 | mOffsets[0] = numNeg;
|
---|
| 148 | for (int i = 1; i < 128; ++i)
|
---|
| 149 | {
|
---|
| 150 | mOffsets[i] = mOffsets[i-1] + mCounters[byteIndex][i-1];
|
---|
| 151 | }
|
---|
| 152 | // Do negative numbers (must start at zero)
|
---|
| 153 | // No need to invert ordering, already correct (-1 is highest number)
|
---|
| 154 | mOffsets[128] = 0;
|
---|
| 155 | for (int i = 129; i < 256; ++i)
|
---|
| 156 | {
|
---|
| 157 | mOffsets[i] = mOffsets[i-1] + mCounters[byteIndex][i-1];
|
---|
| 158 | }
|
---|
| 159 |
|
---|
| 160 | // Sort pass
|
---|
| 161 | for (int i = 0; i < mSortSize; ++i)
|
---|
| 162 | {
|
---|
| 163 | unsigned char byteVal = getByte(byteIndex, (*mSrc)[i].key);
|
---|
| 164 | (*mDest)[mOffsets[byteVal]++] = (*mSrc)[i];
|
---|
| 165 | }
|
---|
| 166 | }
|
---|
| 167 |
|
---|
| 168 |
|
---|
| 169 | // special case float
|
---|
| 170 | void finalPass(int byteIndex, float val)
|
---|
| 171 | {
|
---|
| 172 | // floats need to be special cased since negative numbers will come
|
---|
| 173 | // after positives (high bit = sign) and will be in reverse order
|
---|
| 174 | // (no ones-complement of the +ve value)
|
---|
| 175 | int numNeg = 0;
|
---|
| 176 | // all negative values are in entries 128+ in most significant byte
|
---|
| 177 | for (int i = 128; i < 256; ++i)
|
---|
| 178 | {
|
---|
| 179 | numNeg += mCounters[byteIndex][i];
|
---|
| 180 | }
|
---|
| 181 | // Calculate offsets - positive ones start at the number of negatives
|
---|
| 182 | // do positive numbers normally
|
---|
| 183 | mOffsets[0] = numNeg;
|
---|
| 184 | for (int i = 1; i < 128; ++i)
|
---|
| 185 | {
|
---|
| 186 | mOffsets[i] = mOffsets[i-1] + mCounters[byteIndex][i-1];
|
---|
| 187 | }
|
---|
| 188 | // Do negative numbers (must start at zero)
|
---|
| 189 | // Also need to invert ordering
|
---|
| 190 | // In order to preserve the stability of the sort (essential since
|
---|
| 191 | // we rely on previous bytes already being sorted) we have to count
|
---|
| 192 | // backwards in our offsets from
|
---|
| 193 | mOffsets[255] = mCounters[byteIndex][255];
|
---|
| 194 | for (int i = 254; i > 127; --i)
|
---|
| 195 | {
|
---|
| 196 | mOffsets[i] = mOffsets[i+1] + mCounters[byteIndex][i];
|
---|
| 197 | }
|
---|
| 198 |
|
---|
| 199 | // Sort pass
|
---|
| 200 | for (int i = 0; i < mSortSize; ++i)
|
---|
| 201 | {
|
---|
| 202 | unsigned char byteVal = getByte(byteIndex, (*mSrc)[i].key);
|
---|
| 203 | if (byteVal > 127)
|
---|
| 204 | {
|
---|
| 205 | // -ve; pre-decrement since offsets set to count
|
---|
| 206 | (*mDest)[--mOffsets[byteVal]] = (*mSrc)[i];
|
---|
| 207 | }
|
---|
| 208 | else
|
---|
| 209 | {
|
---|
| 210 | // +ve
|
---|
| 211 | (*mDest)[mOffsets[byteVal]++] = (*mSrc)[i];
|
---|
| 212 | }
|
---|
| 213 | }
|
---|
| 214 | }
|
---|
| 215 |
|
---|
| 216 | inline unsigned char getByte(int byteIndex, TCompValueType val)
|
---|
| 217 | {
|
---|
| 218 | #if OGRE_ENDIAN == OGRE_ENDIAN_LITTLE
|
---|
| 219 | return ((unsigned char*)(&val))[byteIndex];
|
---|
| 220 | #else
|
---|
| 221 | return ((unsigned char*)(&val))[mNumPasses - byteIndex - 1];
|
---|
| 222 | #endif
|
---|
| 223 | }
|
---|
| 224 |
|
---|
| 225 | public:
|
---|
| 226 |
|
---|
| 227 | RadixSort() {}
|
---|
| 228 | ~RadixSort() {}
|
---|
| 229 |
|
---|
| 230 | /** Main sort function
|
---|
| 231 | @param container A container of the type you declared when declaring
|
---|
| 232 | @param func A functor which returns the value for comparison when given
|
---|
| 233 | a container value
|
---|
| 234 | */
|
---|
| 235 | template <class TFunction>
|
---|
| 236 | void sort(TContainer& container, TFunction func)
|
---|
| 237 | {
|
---|
| 238 | if (container.empty())
|
---|
| 239 | return;
|
---|
| 240 |
|
---|
| 241 | // Set up the sort areas
|
---|
| 242 | mSortSize = static_cast<int>(container.size());
|
---|
| 243 | mSortArea1.resize(container.size());
|
---|
| 244 | mSortArea2.resize(container.size());
|
---|
| 245 |
|
---|
| 246 | // Copy data now (we need constant iterators for sorting)
|
---|
| 247 | mTmpContainer = container;
|
---|
| 248 |
|
---|
| 249 | mNumPasses = sizeof(TCompValueType);
|
---|
| 250 |
|
---|
| 251 | // Counter pass
|
---|
| 252 | // Initialise the counts
|
---|
| 253 | int p;
|
---|
| 254 | for (p = 0; p < mNumPasses; ++p)
|
---|
| 255 | memset(mCounters[p], 0, sizeof(int) * 256);
|
---|
| 256 |
|
---|
| 257 | // Perform alpha pass to count
|
---|
| 258 | ContainerIter i = mTmpContainer.begin();
|
---|
| 259 | TCompValueType prevValue = func.operator()(*i);
|
---|
| 260 | bool needsSorting = false;
|
---|
| 261 | for (int u = 0; i != mTmpContainer.end(); ++i, ++u)
|
---|
| 262 | {
|
---|
| 263 | // get sort value
|
---|
| 264 | TCompValueType val = func.operator()(*i);
|
---|
| 265 | // cheap check to see if needs sorting (temporal coherence)
|
---|
| 266 | if (!needsSorting && val < prevValue)
|
---|
| 267 | needsSorting = true;
|
---|
| 268 |
|
---|
| 269 | // Create a sort entry
|
---|
| 270 | mSortArea1[u].key = val;
|
---|
| 271 | mSortArea1[u].iter = i;
|
---|
| 272 |
|
---|
| 273 | // increase counters
|
---|
| 274 | for (p = 0; p < mNumPasses; ++p)
|
---|
| 275 | {
|
---|
| 276 | unsigned char byteVal = getByte(p, val);
|
---|
| 277 | mCounters[p][byteVal]++;
|
---|
| 278 | }
|
---|
| 279 |
|
---|
| 280 | prevValue = val;
|
---|
| 281 |
|
---|
| 282 | }
|
---|
| 283 |
|
---|
| 284 | // early exit if already sorted
|
---|
| 285 | if (!needsSorting)
|
---|
| 286 | return;
|
---|
| 287 |
|
---|
| 288 |
|
---|
| 289 | // Sort passes
|
---|
| 290 | mSrc = &mSortArea1;
|
---|
| 291 | mDest = &mSortArea2;
|
---|
| 292 |
|
---|
| 293 | for (p = 0; p < mNumPasses - 1; ++p)
|
---|
| 294 | {
|
---|
| 295 | sortPass(p);
|
---|
| 296 | // flip src/dst
|
---|
| 297 | std::vector<SortEntry>* tmp = mSrc;
|
---|
| 298 | mSrc = mDest;
|
---|
| 299 | mDest = tmp;
|
---|
| 300 | }
|
---|
| 301 | // Final pass may differ, make polymorphic
|
---|
| 302 | finalPass(p, prevValue);
|
---|
| 303 |
|
---|
| 304 | // Copy everything back
|
---|
| 305 | int c = 0;
|
---|
| 306 | for (i = container.begin();
|
---|
| 307 | i != container.end(); ++i, ++c)
|
---|
| 308 | {
|
---|
| 309 | *i = *((*mDest)[c].iter);
|
---|
| 310 | }
|
---|
| 311 | }
|
---|
| 312 |
|
---|
| 313 | };
|
---|
| 314 |
|
---|
| 315 |
|
---|
| 316 | }
|
---|
| 317 | #endif
|
---|
| 318 |
|
---|