shithub: openh264

ref: 608b3ef520fda300208a70490bd171be84e97eb8
dir: /codec/encoder/core/src/decode_mb_aux.cpp/

View raw version
/*!
 * \copy
 *     Copyright (c)  2013, Cisco Systems
 *     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.
 *
 *     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.
 *
 */

#include "decode_mb_aux.h"
#include "cpu_core.h"

namespace WelsEnc {
/****************************************************************************
 * Dequant and Ihdm functions
 ****************************************************************************/
void WelsIHadamard4x4Dc (int16_t* pRes) { //pBuffer size : 4x4
  int16_t iTemp[4];
  int32_t i = 4;

  while (--i >= 0) {
    const int32_t kiIdx  = i << 2;
    const int32_t kiIdx1 = 1 + kiIdx;
    const int32_t kiIdx2 = 1 + kiIdx1;
    const int32_t kiIdx3 = 1 + kiIdx2;

    iTemp[0] = pRes[kiIdx ] + pRes[kiIdx2];
    iTemp[1] = pRes[kiIdx ] - pRes[kiIdx2];
    iTemp[2] = pRes[kiIdx1] - pRes[kiIdx3];
    iTemp[3] = pRes[kiIdx1] + pRes[kiIdx3];

    pRes[kiIdx ] = iTemp[0] + iTemp[3];
    pRes[kiIdx1] = iTemp[1] + iTemp[2];
    pRes[kiIdx2] = iTemp[1] - iTemp[2];
    pRes[kiIdx3] = iTemp[0] - iTemp[3];
  }

  i = 4;
  while (--i >= 0) {
    const int32_t kiI4  = 4 + i;
    const int32_t kiI8  = 4 + kiI4;
    const int32_t kiI12 = 4 + kiI8;

    iTemp[0] = pRes[i  ] + pRes[kiI8 ];
    iTemp[1] = pRes[i  ] - pRes[kiI8 ];
    iTemp[2] = pRes[kiI4 ] - pRes[kiI12];
    iTemp[3] = pRes[kiI4 ] + pRes[kiI12];

    pRes[i  ] = iTemp[0] + iTemp[3];
    pRes[kiI4 ] = iTemp[1] + iTemp[2];
    pRes[kiI8 ] = iTemp[1] - iTemp[2];
    pRes[kiI12] = iTemp[0] - iTemp[3];
  }
}

/* for qp < 12 */
void WelsDequantLumaDc4x4 (int16_t* pRes, const int32_t kiQp) {
  int32_t i = 15;
  const uint16_t kuiDequantValue = g_kuiDequantCoeff[kiQp % 6][0];
  const int16_t kiQF0   = kiQp / 6;
  const int16_t kiQF1   = 2 - kiQF0;
  const int16_t kiQF0S  = 1 << (1 - kiQF0);

  while (i >= 0) {
    pRes[i  ]   = (pRes[i  ]   * kuiDequantValue + kiQF0S) >> kiQF1;
    pRes[i - 1] = (pRes[i - 1] * kuiDequantValue + kiQF0S) >> kiQF1;
    pRes[i - 2] = (pRes[i - 2] * kuiDequantValue + kiQF0S) >> kiQF1;
    pRes[i - 3] = (pRes[i - 3] * kuiDequantValue + kiQF0S) >> kiQF1;

    i -= 4;
  }
}

/* for qp >= 12 */
void WelsDequantIHadamard4x4_c (int16_t* pRes, const uint16_t kuiMF) {
  int16_t iTemp[4];
  int32_t i;

  for (i = 0; i < 16; i += 4) {
    iTemp[0] = pRes[i  ]   + pRes[i + 2];
    iTemp[1] = pRes[i  ]   - pRes[i + 2];
    iTemp[2] = pRes[i + 1] - pRes[i + 3];
    iTemp[3] = pRes[i + 1] + pRes[i + 3];

    pRes[i  ]   = iTemp[0] + iTemp[3];
    pRes[i + 1] = iTemp[1] + iTemp[2];
    pRes[i + 2] = iTemp[1] - iTemp[2];
    pRes[i + 3] = iTemp[0] - iTemp[3];
  }

  for (i = 0; i < 4; i++) {
    iTemp[0] = pRes[i   ]   + pRes[i + 8 ];
    iTemp[1] = pRes[i   ]   - pRes[i + 8 ];
    iTemp[2] = pRes[i + 4 ] - pRes[i + 12];
    iTemp[3] = pRes[i + 4 ] + pRes[i + 12];

    pRes[i  ]    = (iTemp[0] + iTemp[3]) * kuiMF;
    pRes[i + 4 ] = (iTemp[1] + iTemp[2]) * kuiMF;
    pRes[i + 8 ] = (iTemp[1] - iTemp[2]) * kuiMF;
    pRes[i + 12] = (iTemp[0] - iTemp[3]) * kuiMF;
  }
}

void WelsDequantIHadamard2x2Dc (int16_t* pDct, const uint16_t kuiMF) {
  const int16_t kiSumU = pDct[0] + pDct[2];
  const int16_t kiDelU = pDct[0] - pDct[2];
  const int16_t kiSumD = pDct[1] + pDct[3];
  const int16_t kiDelD = pDct[1] - pDct[3];

  pDct[0] = ((kiSumU + kiSumD) * kuiMF) >> 1;
  pDct[1] = ((kiSumU - kiSumD) * kuiMF) >> 1;
  pDct[2] = ((kiDelU + kiDelD) * kuiMF) >> 1;
  pDct[3] = ((kiDelU - kiDelD) * kuiMF) >> 1;
}

void WelsDequant4x4_c (int16_t* pRes, const uint16_t* kpMF) {
  int32_t i;
  for (i = 0; i < 8; i++) {
    pRes[i]     *= kpMF[i];
    pRes[i + 8] *= kpMF[i];
  }
}

void WelsDequantFour4x4_c (int16_t* pRes, const uint16_t* kpMF) {
  int32_t i;
  for (i = 0; i < 8; i++) {
    pRes[i]      *= kpMF[i];
    pRes[i + 8]  *= kpMF[i];
    pRes[i + 16] *= kpMF[i];
    pRes[i + 24] *= kpMF[i];
    pRes[i + 32] *= kpMF[i];
    pRes[i + 40] *= kpMF[i];
    pRes[i + 48] *= kpMF[i];
    pRes[i + 56] *= kpMF[i];
  }
}

/****************************************************************************
 * IDCT functions, final output = prediction(CS) + IDCT(scaled_coeff)
 ****************************************************************************/
void WelsIDctT4Rec_c (uint8_t* pRec, int32_t iStride, uint8_t* pPred, int32_t iPredStride, int16_t* pDct) {
  int32_t i;
  int16_t iTemp[16];

  int32_t iDstStridex2 = iStride << 1;
  int32_t iDstStridex3 = iStride + iDstStridex2;
  int32_t iPredStridex2 = iPredStride << 1;
  int32_t iPredStridex3 = iPredStride + iPredStridex2;

  for (i = 0; i < 4; i ++) { //horizon
    int32_t iIdx = i << 2;
    const int32_t kiHorSumU = pDct[iIdx] + pDct[iIdx + 2];      // add 0-2
    const int32_t kiHorDelU = pDct[iIdx] - pDct[iIdx + 2];      // sub 0-2
    const int32_t kiHorSumD = pDct[iIdx + 1] + (pDct[iIdx + 3] >> 1);
    const int32_t kiHorDelD = (pDct[iIdx + 1] >> 1) - pDct[iIdx + 3];

    iTemp[iIdx  ]   = kiHorSumU  + kiHorSumD;
    iTemp[iIdx + 1] = kiHorDelU   + kiHorDelD;
    iTemp[iIdx + 2] = kiHorDelU   -  kiHorDelD;
    iTemp[iIdx + 3] = kiHorSumU  -  kiHorSumD;
  }

  for (i = 0; i < 4; i ++) { //vertical
    const int32_t kiVerSumL = iTemp[i]                 + iTemp[8 + i];
    const int32_t kiVerDelL   = iTemp[i]                 - iTemp[8 + i];
    const int32_t kiVerDelR   = (iTemp[4 + i] >> 1) - iTemp[12 + i];
    const int32_t kiVerSumR = iTemp[4 + i]             + (iTemp[12 + i] >> 1);

    pRec[i               ] = WelsClip1 (pPred[i                ] + ((kiVerSumL + kiVerSumR + 32) >> 6));
    pRec[iStride + i     ] = WelsClip1 (pPred[iPredStride + i  ] + ((kiVerDelL + kiVerDelR + 32) >> 6));
    pRec[iDstStridex2 + i] = WelsClip1 (pPred[iPredStridex2 + i] + ((kiVerDelL - kiVerDelR + 32) >> 6));
    pRec[iDstStridex3 + i] = WelsClip1 (pPred[iPredStridex3 + i] + ((kiVerSumL - kiVerSumR + 32) >> 6));
  }
}

void WelsIDctFourT4Rec_c (uint8_t* pRec, int32_t iStride, uint8_t* pPred, int32_t iPredStride, int16_t* pDct) {
  int32_t iDstStridex4  = iStride << 2;
  int32_t iPredStridex4 = iPredStride << 2;
  WelsIDctT4Rec_c (pRec,                    iStride, pPred,                     iPredStride, pDct);
  WelsIDctT4Rec_c (&pRec[4],                iStride, &pPred[4],                 iPredStride, pDct + 16);
  WelsIDctT4Rec_c (&pRec[iDstStridex4    ], iStride, &pPred[iPredStridex4  ],   iPredStride, pDct + 32);
  WelsIDctT4Rec_c (&pRec[iDstStridex4 + 4], iStride, &pPred[iPredStridex4 + 4], iPredStride, pDct + 48);

}

void WelsIDctT4RecOnMb (uint8_t* pDst, int32_t iDstStride, uint8_t* pPred, int32_t iPredStride, int16_t* pDct,
                        PIDctFunc pfIDctFourT4) {
  int32_t iDstStridex8  = iDstStride << 3;
  int32_t iPredStridex8 = iPredStride << 3;

  pfIDctFourT4 (&pDst[0], iDstStride, &pPred[0], iPredStride, pDct);
  pfIDctFourT4 (&pDst[8], iDstStride, &pPred[8], iPredStride, pDct + 64);
  pfIDctFourT4 (&pDst[iDstStridex8], iDstStride, &pPred[iPredStridex8], iPredStride, pDct + 128);
  pfIDctFourT4 (&pDst[iDstStridex8 + 8], iDstStride, &pPred[iPredStridex8 + 8], iPredStride, pDct + 192);
}

/*
 * pfIDctI16x16Dc: do luma idct of an MB for I16x16 mode, when only dc value are non-zero
 */
void WelsIDctRecI16x16Dc_c (uint8_t* pRec, int32_t iStride, uint8_t* pPred, int32_t iPredStride, int16_t* pDctDc) {
  int32_t i, j;

  for (i = 0; i < 16; i ++) {
    for (j = 0; j < 16; j++) {
      pRec[j] = WelsClip1 (pPred[j] + ((pDctDc[ (i & 0x0C) + (j >> 2)] + 32) >> 6));
    }
    pRec += iStride;
    pPred += iPredStride;
  }
}

void WelsGetEncBlockStrideOffset (int32_t* pBlock, const int32_t kiStrideY, const int32_t kiStrideUV) {
  int32_t i, j, k, r;
  for (j = 0; j < 4; j++) {
    i = j << 2;
    k = (j & 0x01) << 1;
    r = j & 0x02;
    pBlock[i]           = (0 + k + (0 + r) * kiStrideY) << 2;
    pBlock[i + 1]       = (1 + k + (0 + r) * kiStrideY) << 2;
    pBlock[i + 2]       = (0 + k + (1 + r) * kiStrideY) << 2;
    pBlock[i + 3]       = (1 + k + (1 + r) * kiStrideY) << 2;

    pBlock[16 + j]      =
      pBlock[20 + j]    = ((j & 0x01) + r * kiStrideUV) << 2;
  }
}

void WelsInitReconstructionFuncs (SWelsFuncPtrList* pFuncList, uint32_t  uiCpuFlag) {
  pFuncList->pfDequantization4x4            = WelsDequant4x4_c;
  pFuncList->pfDequantizationFour4x4        = WelsDequantFour4x4_c;
  pFuncList->pfDequantizationIHadamard4x4   = WelsDequantIHadamard4x4_c;

  pFuncList->pfIDctT4           = WelsIDctT4Rec_c;
  pFuncList->pfIDctFourT4       = WelsIDctFourT4Rec_c;
  pFuncList->pfIDctI16x16Dc     = WelsIDctRecI16x16Dc_c;

#if defined(X86_ASM)
  if (uiCpuFlag & WELS_CPU_MMXEXT) {
    pFuncList->pfIDctT4         = WelsIDctT4Rec_mmx;
  }
  if (uiCpuFlag & WELS_CPU_SSE2) {
    pFuncList->pfDequantization4x4          = WelsDequant4x4_sse2;
    pFuncList->pfDequantizationFour4x4      = WelsDequantFour4x4_sse2;
    pFuncList->pfDequantizationIHadamard4x4 = WelsDequantIHadamard4x4_sse2;

    pFuncList->pfIDctT4         = WelsIDctT4Rec_sse2;
    pFuncList->pfIDctFourT4     = WelsIDctFourT4Rec_sse2;
    pFuncList->pfIDctI16x16Dc   = WelsIDctRecI16x16Dc_sse2;
  }
#if defined(HAVE_AVX2)
  if (uiCpuFlag & WELS_CPU_AVX2) {
    pFuncList->pfIDctT4     = WelsIDctT4Rec_avx2;
    pFuncList->pfIDctFourT4 = WelsIDctFourT4Rec_avx2;
  }
#endif

#endif//X86_ASM

#if defined(HAVE_NEON)
  if (uiCpuFlag & WELS_CPU_NEON) {
    pFuncList->pfDequantization4x4          = WelsDequant4x4_neon;
    pFuncList->pfDequantizationFour4x4      = WelsDequantFour4x4_neon;
    pFuncList->pfDequantizationIHadamard4x4 = WelsDequantIHadamard4x4_neon;

    pFuncList->pfIDctFourT4     = WelsIDctFourT4Rec_neon;
    pFuncList->pfIDctT4         = WelsIDctT4Rec_neon;
    pFuncList->pfIDctI16x16Dc   = WelsIDctRecI16x16Dc_neon;
  }
#endif

#if defined(HAVE_NEON_AARCH64)
  if (uiCpuFlag & WELS_CPU_NEON) {
    pFuncList->pfDequantization4x4          = WelsDequant4x4_AArch64_neon;
    pFuncList->pfDequantizationFour4x4      = WelsDequantFour4x4_AArch64_neon;
    pFuncList->pfDequantizationIHadamard4x4 = WelsDequantIHadamard4x4_AArch64_neon;

    pFuncList->pfIDctFourT4     = WelsIDctFourT4Rec_AArch64_neon;
    pFuncList->pfIDctT4         = WelsIDctT4Rec_AArch64_neon;
    pFuncList->pfIDctI16x16Dc   = WelsIDctRecI16x16Dc_AArch64_neon;
  }
#endif
}
}