Buffer.cpp 20.9 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64
/* The copyright in this software is being made available under the BSD
* License, included below. This software may be subject to other third party
* and contributor rights, including patent rights, and no such rights are
* granted under this license.
*
* Copyright (c) 2010-2018, ITU/ISO/IEC
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
*  * Redistributions of source code must retain the above copyright notice,
*    this list of conditions and the following disclaimer.
*  * Redistributions in binary form must reproduce the above copyright notice,
*    this list of conditions and the following disclaimer in the documentation
*    and/or other materials provided with the distribution.
*  * Neither the name of the ITU/ISO/IEC nor the names of its contributors may
*    be used to endorse or promote products derived from this software without
*    specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
* THE POSSIBILITY OF SUCH DAMAGE.
*/

/** \file     Buffer.cpp
 *  \brief    Low-overhead class describing 2D memory layout
 */

#define DONT_UNDEF_SIZE_AWARE_PER_EL_OP

// unit needs to come first due to a forward declaration
#include "Unit.h"
#include "Buffer.h"
#include "InterpolationFilter.h"

#if ENABLE_SIMD_OPT_BUFFER
#ifdef TARGET_SIMD_X86

#include "CommonDefX86.h"

template< typename T >
void addAvgCore( const T* src1, int src1Stride, const T* src2, int src2Stride, T* dest, int dstStride, int width, int height, int rshift, int offset, const ClpRng& clpRng )
{
#define ADD_AVG_CORE_OP( ADDR ) dest[ADDR] = ClipPel( rightShift( ( src1[ADDR] + src2[ADDR] + offset ), rshift ), clpRng )
#define ADD_AVG_CORE_INC    \
  src1 += src1Stride;       \
  src2 += src2Stride;       \
  dest +=  dstStride;       \

  SIZE_AWARE_PER_EL_OP( ADD_AVG_CORE_OP, ADD_AVG_CORE_INC );

#undef ADD_AVG_CORE_OP
#undef ADD_AVG_CORE_INC
}

65
#if JVET_L0256_BIO
66
void addBIOAvgCore(const Pel* src0, int src0Stride, const Pel* src1, int src1Stride, Pel *dst, int dstStride, const Pel *gradX0, const Pel *gradX1, const Pel *gradY0, const Pel*gradY1, int gradStride, int width, int height, int tmpx, int tmpy, int shift, int offset, const ClpRng& clpRng)
67 68 69 70 71 72 73
{
  int b = 0;

  for (int y = 0; y < height; y++)
  {
    for (int x = 0; x < width; x += 4)
    {
74
      b = tmpx * (gradX0[x] - gradX1[x]) + tmpy * (gradY0[x] - gradY1[x]);
75 76 77
      b = ((b + 1) >> 1);
      dst[x] = ClipPel((int16_t)rightShift((src0[x] + src1[x] + b + offset), shift), clpRng);

78
      b = tmpx * (gradX0[x + 1] - gradX1[x + 1]) + tmpy * (gradY0[x + 1] - gradY1[x + 1]);
79 80 81
      b = ((b + 1) >> 1);
      dst[x + 1] = ClipPel((int16_t)rightShift((src0[x + 1] + src1[x + 1] + b + offset), shift), clpRng);

82
      b = tmpx * (gradX0[x + 2] - gradX1[x + 2]) + tmpy * (gradY0[x + 2] - gradY1[x + 2]);
83 84 85
      b = ((b + 1) >> 1);
      dst[x + 2] = ClipPel((int16_t)rightShift((src0[x + 2] + src1[x + 2] + b + offset), shift), clpRng);

86
      b = tmpx * (gradX0[x + 3] - gradX1[x + 3]) + tmpy * (gradY0[x + 3] - gradY1[x + 3]);
87 88 89 90
      b = ((b + 1) >> 1);
      dst[x + 3] = ClipPel((int16_t)rightShift((src0[x + 3] + src1[x + 3] + b + offset), shift), clpRng);
    }
    dst += dstStride;       src0 += src0Stride;     src1 += src1Stride;
91
    gradX0 += gradStride; gradX1 += gradStride; gradY0 += gradStride; gradY1 += gradStride;
92 93 94
  }
}

95
void gradFilterCore(Pel* pSrc, int srcStride, int width, int height, int gradStride, Pel* gradX, Pel* gradY)
96
{
97 98 99
  Pel* srcTmp = pSrc + srcStride + 1;
  Pel* gradXTmp = gradX + gradStride + 1;
  Pel* gradYTmp = gradY + gradStride + 1;
100

101
  for (int y = 0; y < (height - 2 * BIO_EXTEND_SIZE); y++)
102
  {
103
    for (int x = 0; x < (width - 2 * BIO_EXTEND_SIZE); x++)
104
    {
105 106
      gradYTmp[x] = (srcTmp[x + srcStride] - srcTmp[x - srcStride]) >> 4;
      gradXTmp[x] = (srcTmp[x + 1] - srcTmp[x - 1]) >> 4;
107
    }
108 109 110
    gradXTmp += gradStride;
    gradYTmp += gradStride;
    srcTmp += srcStride;
111 112
  }

113 114 115
  gradXTmp = gradX + gradStride + 1;
  gradYTmp = gradY + gradStride + 1;
  for (int y = 0; y < (height - 2 * BIO_EXTEND_SIZE); y++)
116
  {
117 118 119
    gradXTmp[-1] = gradXTmp[0];
    gradXTmp[width - 2 * BIO_EXTEND_SIZE] = gradXTmp[width - 2 * BIO_EXTEND_SIZE - 1];
    gradXTmp += gradStride;
120

121 122 123
    gradYTmp[-1] = gradYTmp[0];
    gradYTmp[width - 2 * BIO_EXTEND_SIZE] = gradYTmp[width - 2 * BIO_EXTEND_SIZE - 1];
    gradYTmp += gradStride;
124 125
  }

126 127 128 129 130 131
  gradXTmp = gradX + gradStride;
  gradYTmp = gradY + gradStride;
  ::memcpy(gradXTmp - gradStride, gradXTmp, sizeof(Pel)*(width));
  ::memcpy(gradXTmp + (height - 2 * BIO_EXTEND_SIZE)*gradStride, gradXTmp + (height - 2 * BIO_EXTEND_SIZE - 1)*gradStride, sizeof(Pel)*(width));
  ::memcpy(gradYTmp - gradStride, gradYTmp, sizeof(Pel)*(width));
  ::memcpy(gradYTmp + (height - 2 * BIO_EXTEND_SIZE)*gradStride, gradYTmp + (height - 2 * BIO_EXTEND_SIZE - 1)*gradStride, sizeof(Pel)*(width));
132 133
}

134
void calcBIOParCore(const Pel* srcY0Temp, const Pel* srcY1Temp, const Pel* gradX0, const Pel* gradX1, const Pel* gradY0, const Pel* gradY1, int* dotProductTemp1, int* dotProductTemp2, int* dotProductTemp3, int* dotProductTemp5, int* dotProductTemp6, const int src0Stride, const int src1Stride, const int gradStride, const int widthG, const int heightG)
135
{
136
  for (int y = 0; y < heightG; y++)
137
  {
138
    for (int x = 0; x < widthG; x++)
139
    {
140 141 142 143 144 145 146 147
      int temp = (srcY0Temp[x] >> 6) - (srcY1Temp[x] >> 6);
      int tempX = (gradX0[x] + gradX1[x]) >> 3;
      int tempY = (gradY0[x] + gradY1[x]) >> 3;
      dotProductTemp1[x] = tempX * tempX;
      dotProductTemp2[x] = tempX * tempY;
      dotProductTemp3[x] = -tempX * temp;
      dotProductTemp5[x] = tempY * tempY;
      dotProductTemp6[x] = -tempY * temp;
148
    }
149 150 151 152 153 154 155 156 157 158 159
    srcY0Temp += src0Stride;
    srcY1Temp += src1Stride;
    gradX0 += gradStride;
    gradX1 += gradStride;
    gradY0 += gradStride;
    gradY1 += gradStride;
    dotProductTemp1 += widthG;
    dotProductTemp2 += widthG;
    dotProductTemp3 += widthG;
    dotProductTemp5 += widthG;
    dotProductTemp6 += widthG;
160 161 162 163 164
  }
}

void calcBlkGradientCore(int sx, int sy, int     *arraysGx2, int     *arraysGxGy, int     *arraysGxdI, int     *arraysGy2, int     *arraysGydI, int     &sGx2, int     &sGy2, int     &sGxGy, int     &sGxdI, int     &sGydI, int width, int height, int unitSize)
{
165 166 167 168 169
  int     *Gx2 = arraysGx2;
  int     *Gy2 = arraysGy2;
  int     *GxGy = arraysGxGy;
  int     *GxdI = arraysGxdI;
  int     *GydI = arraysGydI;
170 171

  // set to the above row due to JVET_K0485_BIO_EXTEND_SIZE
172 173 174 175 176
  Gx2 -= (BIO_EXTEND_SIZE*width);
  Gy2 -= (BIO_EXTEND_SIZE*width);
  GxGy -= (BIO_EXTEND_SIZE*width);
  GxdI -= (BIO_EXTEND_SIZE*width);
  GydI -= (BIO_EXTEND_SIZE*width);
177

178
  for (int y = -BIO_EXTEND_SIZE; y < unitSize + BIO_EXTEND_SIZE; y++)
179
  {
180
    for (int x = -BIO_EXTEND_SIZE; x < unitSize + BIO_EXTEND_SIZE; x++)
181
    {
182 183 184 185 186
      sGx2 += Gx2[x];
      sGy2 += Gy2[x];
      sGxGy += GxGy[x];
      sGxdI += GxdI[x];
      sGydI += GydI[x];
187
    }
188 189 190 191 192
    Gx2 += width;
    Gy2 += width;
    GxGy += width;
    GxdI += width;
    GydI += width;
193 194 195 196
  }
}
#endif

197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230
#if ENABLE_SIMD_OPT_GBI && JVET_L0646_GBI
void removeWeightHighFreq(int16_t* dst, int dstStride, const int16_t* src, int srcStride, int width, int height, int shift, int gbiWeight)
{
  int normalizer = ((1 << 16) + (gbiWeight > 0 ? (gbiWeight >> 1) : -(gbiWeight >> 1))) / gbiWeight;
  int weight0 = normalizer << g_GbiLog2WeightBase;
  int weight1 = (g_GbiWeightBase - gbiWeight)*normalizer;
#define REM_HF_INC  \
  src += srcStride; \
  dst += dstStride; \

#define REM_HF_OP( ADDR )      dst[ADDR] =             (dst[ADDR]*weight0 - src[ADDR]*weight1 + (1<<15))>>16

  SIZE_AWARE_PER_EL_OP(REM_HF_OP, REM_HF_INC);

#undef REM_HF_INC
#undef REM_HF_OP
#undef REM_HF_OP_CLIP
}

void removeHighFreq(int16_t* dst, int dstStride, const int16_t* src, int srcStride, int width, int height)
{
#define REM_HF_INC  \
  src += srcStride; \
  dst += dstStride; \

#define REM_HF_OP( ADDR )      dst[ADDR] =             2 * dst[ADDR] - src[ADDR]

  SIZE_AWARE_PER_EL_OP(REM_HF_OP, REM_HF_INC);

#undef REM_HF_INC
#undef REM_HF_OP
#undef REM_HF_OP_CLIP
}
#endif
231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271

template<typename T>
void reconstructCore( const T* src1, int src1Stride, const T* src2, int src2Stride, T* dest, int dstStride, int width, int height, const ClpRng& clpRng )
{
#define RECO_CORE_OP( ADDR ) dest[ADDR] = ClipPel( src1[ADDR] + src2[ADDR], clpRng )
#define RECO_CORE_INC     \
  src1 += src1Stride;     \
  src2 += src2Stride;     \
  dest +=  dstStride;     \

  SIZE_AWARE_PER_EL_OP( RECO_CORE_OP, RECO_CORE_INC );

#undef RECO_CORE_OP
#undef RECO_CORE_INC
}


template<typename T>
void linTfCore( const T* src, int srcStride, Pel *dst, int dstStride, int width, int height, int scale, int shift, int offset, const ClpRng& clpRng, bool bClip )
{
#define LINTF_CORE_OP( ADDR ) dst[ADDR] = ( Pel ) bClip ? ClipPel( rightShift( scale * src[ADDR], shift ) + offset, clpRng ) : ( rightShift( scale * src[ADDR], shift ) + offset )
#define LINTF_CORE_INC  \
  src += srcStride;     \
  dst += dstStride;     \

  SIZE_AWARE_PER_EL_OP( LINTF_CORE_OP, LINTF_CORE_INC );

#undef LINTF_CORE_OP
#undef LINTF_CORE_INC
}

PelBufferOps::PelBufferOps()
{
  addAvg4 = addAvgCore<Pel>;
  addAvg8 = addAvgCore<Pel>;

  reco4 = reconstructCore<Pel>;
  reco8 = reconstructCore<Pel>;

  linTf4 = linTfCore<Pel>;
  linTf8 = linTfCore<Pel>;
272

273 274 275 276 277 278 279
#if JVET_L0256_BIO
  addBIOAvg4      = addBIOAvgCore;
  bioGradFilter   = gradFilterCore;
  calcBIOPar      = calcBIOParCore;
  calcBlkGradient = calcBlkGradientCore;
#endif

280 281 282 283 284 285 286
#if ENABLE_SIMD_OPT_GBI
  removeWeightHighFreq8 = removeWeightHighFreq;
  removeWeightHighFreq4 = removeWeightHighFreq;
  removeHighFreq8 = removeHighFreq;
  removeHighFreq4 = removeHighFreq;
#endif

287 288 289 290 291 292 293
}

PelBufferOps g_pelBufOP = PelBufferOps();

#endif
#endif

294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324
#if JVET_L0646_GBI
template<>
void AreaBuf<Pel>::addWeightedAvg(const AreaBuf<const Pel> &other1, const AreaBuf<const Pel> &other2, const ClpRng& clpRng, const int8_t gbiIdx)
{
  const int8_t w0 = getGbiWeight(gbiIdx, REF_PIC_LIST_0);
  const int8_t w1 = getGbiWeight(gbiIdx, REF_PIC_LIST_1);
  const int8_t log2WeightBase = g_GbiLog2WeightBase;

  const Pel* src0 = other1.buf;
  const Pel* src2 = other2.buf;
  Pel* dest = buf;

  const unsigned src1Stride = other1.stride;
  const unsigned src2Stride = other2.stride;
  const unsigned destStride = stride;
  const int clipbd = clpRng.bd;
  const int shiftNum = std::max<int>(2, (IF_INTERNAL_PREC - clipbd)) + log2WeightBase;
  const int offset = (1 << (shiftNum - 1)) + (IF_INTERNAL_OFFS << log2WeightBase);

#define ADD_AVG_OP( ADDR ) dest[ADDR] = ClipPel( rightShift( ( src0[ADDR]*w0 + src2[ADDR]*w1 + offset ), shiftNum ), clpRng )
#define ADD_AVG_INC     \
    src0 += src1Stride; \
    src2 += src2Stride; \
    dest += destStride; \

  SIZE_AWARE_PER_EL_OP(ADD_AVG_OP, ADD_AVG_INC);

#undef ADD_AVG_OP
#undef ADD_AVG_INC
}
#endif
325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660

template<>
void AreaBuf<Pel>::addAvg( const AreaBuf<const Pel> &other1, const AreaBuf<const Pel> &other2, const ClpRng& clpRng)
{
  const Pel* src0 = other1.buf;
  const Pel* src2 = other2.buf;
        Pel* dest =        buf;

  const unsigned src1Stride = other1.stride;
  const unsigned src2Stride = other2.stride;
  const unsigned destStride =        stride;
  const int     clipbd      = clpRng.bd;
  const int     shiftNum    = std::max<int>(2, (IF_INTERNAL_PREC - clipbd)) + 1;
  const int     offset      = (1 << (shiftNum - 1)) + 2 * IF_INTERNAL_OFFS;

#if ENABLE_SIMD_OPT_BUFFER && defined(TARGET_SIMD_X86)
  if( ( width & 7 ) == 0 )
  {
    g_pelBufOP.addAvg8( src0, src1Stride, src2, src2Stride, dest, destStride, width, height, shiftNum, offset, clpRng );
  }
  else if( ( width & 3 ) == 0 )
  {
    g_pelBufOP.addAvg4( src0, src1Stride, src2, src2Stride, dest, destStride, width, height, shiftNum, offset, clpRng );
  }
  else
#endif
  {
#define ADD_AVG_OP( ADDR ) dest[ADDR] = ClipPel( rightShift( ( src0[ADDR] + src2[ADDR] + offset ), shiftNum ), clpRng )
#define ADD_AVG_INC     \
    src0 += src1Stride; \
    src2 += src2Stride; \
    dest += destStride; \

    SIZE_AWARE_PER_EL_OP( ADD_AVG_OP, ADD_AVG_INC );

#undef ADD_AVG_OP
#undef ADD_AVG_INC
  }
}

template<>
void AreaBuf<Pel>::toLast( const ClpRng& clpRng )
{
        Pel* src       = buf;
  const uint32_t srcStride = stride;

  const int  clipbd    = clpRng.bd;
  const int  shiftNum  = std::max<int>(2, (IF_INTERNAL_PREC - clipbd));
  const int  offset    = ( 1 << ( shiftNum - 1 ) ) + IF_INTERNAL_OFFS;

  if (width == 1)
  {
    THROW( "Blocks of width = 1 not supported" );
  }
  else if (width&2)
  {
    for ( int y = 0; y < height; y++ )
    {
      for (int x=0 ; x < width; x+=2 )
      {
        src[x + 0] = ClipPel( rightShift( ( src[x + 0] + offset ), shiftNum ), clpRng );
        src[x + 1] = ClipPel( rightShift( ( src[x + 1] + offset ), shiftNum ), clpRng );
      }
      src += srcStride;
    }
  }
  else
  {
    for ( int y = 0; y < height; y++ )
    {
      for (int x=0 ; x < width; x+=4 )
      {
        src[x + 0] = ClipPel( rightShift( ( src[x + 0] + offset ), shiftNum ), clpRng );
        src[x + 1] = ClipPel( rightShift( ( src[x + 1] + offset ), shiftNum ), clpRng );
        src[x + 2] = ClipPel( rightShift( ( src[x + 2] + offset ), shiftNum ), clpRng );
        src[x + 3] = ClipPel( rightShift( ( src[x + 3] + offset ), shiftNum ), clpRng );

      }
      src += srcStride;
    }
  }
}


template<>
void AreaBuf<Pel>::copyClip( const AreaBuf<const Pel> &src, const ClpRng& clpRng )
{
  const Pel* srcp = src.buf;
        Pel* dest =     buf;

  const unsigned srcStride  = src.stride;
  const unsigned destStride = stride;

  if( width == 1 )
  {
    THROW( "Blocks of width = 1 not supported" );
  }
  else
  {
#define RECO_OP( ADDR ) dest[ADDR] = ClipPel( srcp[ADDR], clpRng )
#define RECO_INC        \
    srcp += srcStride;  \
    dest += destStride; \

    SIZE_AWARE_PER_EL_OP( RECO_OP, RECO_INC );

#undef RECO_OP
#undef RECO_INC
  }
}


template<>
void AreaBuf<Pel>::reconstruct( const AreaBuf<const Pel> &pred, const AreaBuf<const Pel> &resi, const ClpRng& clpRng )
{
  const Pel* src1 = pred.buf;
  const Pel* src2 = resi.buf;
        Pel* dest =      buf;

  const unsigned src1Stride = pred.stride;
  const unsigned src2Stride = resi.stride;
  const unsigned destStride =      stride;

#if ENABLE_SIMD_OPT_BUFFER && defined(TARGET_SIMD_X86)
  if( ( width & 7 ) == 0 )
  {
    g_pelBufOP.reco8( src1, src1Stride, src2, src2Stride, dest, destStride, width, height, clpRng );
  }
  else if( ( width & 3 ) == 0 )
  {
    g_pelBufOP.reco4( src1, src1Stride, src2, src2Stride, dest, destStride, width, height, clpRng );
  }
  else
#endif
  {
#define RECO_OP( ADDR ) dest[ADDR] = ClipPel( src1[ADDR] + src2[ADDR], clpRng )
#define RECO_INC        \
    src1 += src1Stride; \
    src2 += src2Stride; \
    dest += destStride; \

    SIZE_AWARE_PER_EL_OP( RECO_OP, RECO_INC );

#undef RECO_OP
#undef RECO_INC
  }
}

template<>
void AreaBuf<Pel>::linearTransform( const int scale, const int shift, const int offset, bool bClip, const ClpRng& clpRng )
{
  const Pel* src = buf;
        Pel* dst = buf;

  if( width == 1 )
  {
    THROW( "Blocks of width = 1 not supported" );
  }
#if ENABLE_SIMD_OPT_BUFFER && defined(TARGET_SIMD_X86)
  else if( ( width & 7 ) == 0 )
  {
    g_pelBufOP.linTf8( src, stride, dst, stride, width, height, scale, shift, offset, clpRng, bClip );
  }
  else if( ( width & 3 ) == 0 )
  {
    g_pelBufOP.linTf4( src, stride, dst, stride, width, height, scale, shift, offset, clpRng, bClip );
  }
#endif
  else
  {
#define LINTF_OP( ADDR ) dst[ADDR] = ( Pel ) bClip ? ClipPel( rightShift( scale * src[ADDR], shift ) + offset, clpRng ) : ( rightShift( scale * src[ADDR], shift ) + offset )
#define LINTF_INC        \
    src += stride;       \
    dst += stride;       \

    SIZE_AWARE_PER_EL_OP( LINTF_OP, LINTF_INC );

#undef RECO_OP
#undef RECO_INC
  }
}

#if ENABLE_SIMD_OPT_BUFFER && defined(TARGET_SIMD_X86)
template<>
void AreaBuf<Pel>::subtract( const Pel val )
{
  ClpRng clpRngDummy;
  linearTransform( 1, 0, -val, false, clpRngDummy );
}
#endif


PelStorage::PelStorage()
{
  for( uint32_t i = 0; i < MAX_NUM_COMPONENT; i++ )
  {
    m_origin[i] = nullptr;
  }
}

PelStorage::~PelStorage()
{
  destroy();
}

void PelStorage::create( const UnitArea &_UnitArea )
{
  create( _UnitArea.chromaFormat, _UnitArea.blocks[0] );
}

void PelStorage::create( const ChromaFormat &_chromaFormat, const Area& _area, const unsigned _maxCUSize, const unsigned _margin, const unsigned _alignment, const bool _scaleChromaMargin )
{
  CHECK( !bufs.empty(), "Trying to re-create an already initialized buffer" );

  chromaFormat = _chromaFormat;

  const uint32_t numCh = getNumberValidComponents( _chromaFormat );

  unsigned extHeight = _area.height;
  unsigned extWidth  = _area.width;

  if( _maxCUSize )
  {
    extHeight = ( ( _area.height + _maxCUSize - 1 ) / _maxCUSize ) * _maxCUSize;
    extWidth  = ( ( _area.width  + _maxCUSize - 1 ) / _maxCUSize ) * _maxCUSize;
  }

  for( uint32_t i = 0; i < numCh; i++ )
  {
    const ComponentID compID = ComponentID( i );
    const unsigned scaleX = ::getComponentScaleX( compID, _chromaFormat );
    const unsigned scaleY = ::getComponentScaleY( compID, _chromaFormat );

    unsigned scaledHeight = extHeight >> scaleY;
    unsigned scaledWidth  = extWidth  >> scaleX;
    unsigned ymargin      = _margin >> (_scaleChromaMargin?scaleY:0);
    unsigned xmargin      = _margin >> (_scaleChromaMargin?scaleX:0);
    unsigned totalWidth   = scaledWidth + 2*xmargin;
    unsigned totalHeight  = scaledHeight +2*ymargin;

    if( _alignment )
    {
      // make sure buffer lines are align
      CHECK( _alignment != MEMORY_ALIGN_DEF_SIZE, "Unsupported alignment" );
      totalWidth = ( ( totalWidth + _alignment - 1 ) / _alignment ) * _alignment;
    }
    uint32_t area = totalWidth * totalHeight;
    CHECK( !area, "Trying to create a buffer with zero area" );

    m_origin[i] = ( Pel* ) xMalloc( Pel, area );
    Pel* topLeft = m_origin[i] + totalWidth * ymargin + xmargin;
    bufs.push_back( PelBuf( topLeft, totalWidth, _area.width >> scaleX, _area.height >> scaleY ) );
  }
}

void PelStorage::createFromBuf( PelUnitBuf buf )
{
  chromaFormat = buf.chromaFormat;

  const uint32_t numCh = ::getNumberValidComponents( chromaFormat );

  bufs.resize(numCh);

  for( uint32_t i = 0; i < numCh; i++ )
  {
    PelBuf cPelBuf = buf.get( ComponentID( i ) );
    bufs[i] = PelBuf( cPelBuf.bufAt( 0, 0 ), cPelBuf.stride, cPelBuf.width, cPelBuf.height );
  }
}

void PelStorage::swap( PelStorage& other )
{
  const uint32_t numCh = ::getNumberValidComponents( chromaFormat );

  for( uint32_t i = 0; i < numCh; i++ )
  {
    // check this otherwise it would turn out to get very weird
    CHECK( chromaFormat                   != other.chromaFormat                  , "Incompatible formats" );
    CHECK( get( ComponentID( i ) )        != other.get( ComponentID( i ) )       , "Incompatible formats" );
    CHECK( get( ComponentID( i ) ).stride != other.get( ComponentID( i ) ).stride, "Incompatible formats" );

    std::swap( bufs[i].buf,    other.bufs[i].buf );
    std::swap( bufs[i].stride, other.bufs[i].stride );
    std::swap( m_origin[i],    other.m_origin[i] );
  }
}

void PelStorage::destroy()
{
  chromaFormat = NUM_CHROMA_FORMAT;
  for( uint32_t i = 0; i < MAX_NUM_COMPONENT; i++ )
  {
    if( m_origin[i] )
    {
      xFree( m_origin[i] );
      m_origin[i] = nullptr;
    }
  }
  bufs.clear();
}

PelBuf PelStorage::getBuf( const ComponentID CompID )
{
  return bufs[CompID];
}

const CPelBuf PelStorage::getBuf( const ComponentID CompID ) const
{
  return bufs[CompID];
}

PelBuf PelStorage::getBuf( const CompArea &blk )
{
  const PelBuf& r = bufs[blk.compID];

  CHECKD( rsAddr( blk.bottomRight(), r.stride ) >= ( ( r.height - 1 ) * r.stride + r.width ), "Trying to access a buf outside of bound!" );

  return PelBuf( r.buf + rsAddr( blk, r.stride ), r.stride, blk );
}

const CPelBuf PelStorage::getBuf( const CompArea &blk ) const
{
  const PelBuf& r = bufs[blk.compID];
  return CPelBuf( r.buf + rsAddr( blk, r.stride ), r.stride, blk );
}

PelUnitBuf PelStorage::getBuf( const UnitArea &unit )
{
  return ( chromaFormat == CHROMA_400 ) ? PelUnitBuf( chromaFormat, getBuf( unit.Y() ) ) : PelUnitBuf( chromaFormat, getBuf( unit.Y() ), getBuf( unit.Cb() ), getBuf( unit.Cr() ) );
}

const CPelUnitBuf PelStorage::getBuf( const UnitArea &unit ) const
{
  return ( chromaFormat == CHROMA_400 ) ? CPelUnitBuf( chromaFormat, getBuf( unit.Y() ) ) : CPelUnitBuf( chromaFormat, getBuf( unit.Y() ), getBuf( unit.Cb() ), getBuf( unit.Cr() ) );
}