/******************************************************************************
 *
 *  Copyright 2014 The Android Open Source Project
 *  Copyright 2003 - 2004 Open Interface North America, Inc. All rights
 *                        reserved.
 *
 *  Licensed under the Apache License, Version 2.0 (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.apache.org/licenses/LICENSE-2.0
 *
 *  Unless required by applicable law or agreed to in writing, software
 *  distributed under the License is distributed on an "AS IS" BASIS,
 *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 *  See the License for the specific language governing permissions and
 *  limitations under the License.
 *
 ******************************************************************************/

/*******************************************************************************
  $Revision: #1 $
 ******************************************************************************/

/**
@file

The functions in this file relate to the allocation of available bits to
subbands within the SBC/eSBC frame, along with support functions for computing
frame length and bitrate.

@ingroup codec_internal
*/

/**
@addtogroup codec_internal
@{
*/

#include <oi_codec_sbc_private.h>
#include "oi_utils.h"

uint32_t OI_SBC_MaxBitpool(OI_CODEC_SBC_FRAME_INFO* frame) {
  switch (frame->mode) {
    case SBC_MONO:
    case SBC_DUAL_CHANNEL:
      return 16 * frame->nrof_subbands;
    case SBC_STEREO:
    case SBC_JOINT_STEREO:
      return 32 * frame->nrof_subbands;
  }

  ERROR(("Invalid frame mode %d", frame->mode));
  OI_ASSERT(false);
  return 0; /* Should never be reached */
}

PRIVATE uint16_t internal_CalculateFramelen(OI_CODEC_SBC_FRAME_INFO* frame) {
  uint16_t nbits = frame->nrof_blocks * frame->bitpool;
  uint16_t nrof_subbands = frame->nrof_subbands;
  uint16_t result = nbits;

  if (frame->mode == SBC_JOINT_STEREO) {
    result += nrof_subbands + (8 * nrof_subbands);
  } else {
    if (frame->mode == SBC_DUAL_CHANNEL) {
      result += nbits;
    }
    if (frame->mode == SBC_MONO) {
      result += 4 * nrof_subbands;
    } else {
      result += 8 * nrof_subbands;
    }
  }
  return SBC_HEADER_LEN + (result + 7) / 8;
}

PRIVATE uint32_t internal_CalculateBitrate(OI_CODEC_SBC_FRAME_INFO* frame) {
  OI_UINT blocksbands;
  blocksbands = frame->nrof_subbands * frame->nrof_blocks;

  return DIVIDE(8 * internal_CalculateFramelen(frame) * frame->frequency,
                blocksbands);
}

INLINE uint16_t OI_SBC_CalculateFrameAndHeaderlen(
    OI_CODEC_SBC_FRAME_INFO* frame, OI_UINT* headerLen_) {
  OI_UINT headerLen =
      SBC_HEADER_LEN + frame->nrof_subbands * frame->nrof_channels / 2;

  if (frame->mode == SBC_JOINT_STEREO) {
    headerLen++;
  }

  *headerLen_ = headerLen;
  return internal_CalculateFramelen(frame);
}

#define MIN(x, y) ((x) < (y) ? (x) : (y))

/*
 * Computes the bit need for each sample and as also returns a counts of bit
 * needs that are greater than one. This count is used in the first phase of bit
 * allocation.
 *
 * We also compute a preferred bitpool value that this is the minimum bitpool
 * needed to guarantee lossless representation of the audio data. The preferred
 * bitpool may be larger than the bits actually required but the only input we
 * have are the scale factors. For example, it takes 2 bits to represent values
 * in the range -1 .. +1 but the scale factor is 0. To guarantee lossless
 * representation we add 2 to each scale factor and sum them to come up with the
 * preferred bitpool. This is not ideal because 0 requires 0 bits but we
 * currently have no way of knowing this.
 *
 * @param bitneed       Array to return bitneeds for each subband
 *
 * @param ch            Channel 0 or 1
 *
 * @param preferredBitpool  Returns the number of reserved bits
 *
 * @return              The SBC bit need
 *
 */
OI_UINT computeBitneed(OI_CODEC_SBC_COMMON_CONTEXT* common, uint8_t* bitneeds,
                       OI_UINT ch, OI_UINT* preferredBitpool) {
  static const int8_t offset4[4][4] = {
      {-1, 0, 0, 0}, {-2, 0, 0, 1}, {-2, 0, 0, 1}, {-2, 0, 0, 1}};

  static const int8_t offset8[4][8] = {{-2, 0, 0, 0, 0, 0, 0, 1},
                                       {-3, 0, 0, 0, 0, 0, 1, 2},
                                       {-4, 0, 0, 0, 0, 0, 1, 2},
                                       {-4, 0, 0, 0, 0, 0, 1, 2}};

  const OI_UINT nrof_subbands = common->frameInfo.nrof_subbands;
  OI_UINT sb;
  int8_t* scale_factor = &common->scale_factor[ch ? nrof_subbands : 0];
  OI_UINT bitcount = 0;
  uint8_t maxBits = 0;
  uint8_t prefBits = 0;

  if (common->frameInfo.alloc == SBC_SNR) {
    for (sb = 0; sb < nrof_subbands; sb++) {
      OI_INT bits = scale_factor[sb];
      if (bits > maxBits) {
        maxBits = bits;
      }
      bitneeds[sb] = bits;
      if (bitneeds[sb] > 1) {
        bitcount += bits;
      }
      prefBits += 2 + bits;
    }
  } else {
    const int8_t* offset;
    if (nrof_subbands == 4) {
      offset = offset4[common->frameInfo.freqIndex];
    } else {
      offset = offset8[common->frameInfo.freqIndex];
    }
    for (sb = 0; sb < nrof_subbands; sb++) {
      OI_INT bits = scale_factor[sb];
      if (bits > maxBits) {
        maxBits = bits;
      }
      prefBits += 2 + bits;
      if (bits) {
        bits -= offset[sb];
        if (bits > 0) {
          bits /= 2;
        }
        bits += 5;
      }
      bitneeds[sb] = bits;
      if (bitneeds[sb] > 1) {
        bitcount += bits;
      }
    }
  }
  common->maxBitneed = OI_MAX(maxBits, common->maxBitneed);
  *preferredBitpool += prefBits;
  return bitcount;
}

/*
 * Explanation of the adjustToFitBitpool inner loop.
 *
 * The inner loop computes the effect of adjusting the bit allocation up or
 * down. Allocations must be 0 or in the range 2..16. This is accomplished by
 * the following code:
 *
 *           for (s = bands - 1; s >= 0; --s) {
 *              OI_INT bits = bitadjust + bitneeds[s];
 *              bits = bits < 2 ? 0 : bits;
 *              bits = bits > 16 ? 16 : bits;
 *              count += bits;
 *          }
 *
 * This loop can be optimized to perform 4 operations at a time as follows:
 *
 * Adjustment is computed as a 7 bit signed value and added to the bitneed.
 *
 * Negative allocations are zeroed by masking. (n & 0x40) >> 6 puts the
 * sign bit into bit 0, adding this to 0x7F give us a mask of 0x80
 * for -ve values and 0x7F for +ve values.
 *
 * n &= 0x7F + (n & 0x40) >> 6)
 *
 * Allocations greater than 16 are truncated to 16. Adjusted allocations are in
 * the range 0..31 so we know that bit 4 indicates values >= 16. We use this bit
 * to create a mask that zeroes bits 0 .. 3 if bit 4 is set.
 *
 * n &= (15 + (n >> 4))
 *
 * Allocations of 1 are disallowed. Add and shift creates a mask that
 * eliminates the illegal value
 *
 * n &= ((n + 14) >> 4) | 0x1E
 *
 * These operations can be performed in 8 bits without overflowing so we can
 * operate on 4 values at once.
 */

/*
 * Encoder/Decoder
 *
 * Computes adjustment +/- of bitneeds to fill bitpool and returns overall
 * adjustment and excess bits.
 *
 * @param bitpool   The bitpool we have to work within
 *
 * @param bitneeds  An array of bit needs (more acturately allocation
 *                  prioritities) for each subband across all blocks in the SBC
 *                  frame
 *
 * @param subbands  The number of subbands over which the adkustment is
 *                  calculated. For mono and dual mode this is 4 or 8, for
 *                  stereo or joint stereo this is 8 or 16.
 *
 * @param bitcount  A starting point for the adjustment
 *
 * @param excess    Returns the excess bits after the adjustment
 *
 * @return   The adjustment.
 */
OI_INT adjustToFitBitpool(const OI_UINT bitpool, uint32_t* bitneeds,
                          const OI_UINT subbands, OI_UINT bitcount,
                          OI_UINT* excess) {
  OI_INT maxBitadjust = 0;
  OI_INT bitadjust = (bitcount > bitpool) ? -8 : 8;
  OI_INT chop = 8;

  /*
   * This is essentially a binary search for the optimal adjustment value.
   */
  while ((bitcount != bitpool) && chop) {
    uint32_t total = 0;
    OI_UINT count;
    uint32_t adjust4;
    OI_INT i;

    adjust4 = bitadjust & 0x7F;
    adjust4 |= (adjust4 << 8);
    adjust4 |= (adjust4 << 16);

    for (i = (subbands / 4 - 1); i >= 0; --i) {
      uint32_t mask;
      uint32_t n = bitneeds[i] + adjust4;
      mask = 0x7F7F7F7F + ((n & 0x40404040) >> 6);
      n &= mask;
      mask = 0x0F0F0F0F + ((n & 0x10101010) >> 4);
      n &= mask;
      mask = (((n + 0x0E0E0E0E) >> 4) | 0x1E1E1E1E);
      n &= mask;
      total += n;
    }

    count = (total & 0xFFFF) + (total >> 16);
    count = (count & 0xFF) + (count >> 8);

    chop >>= 1;
    if (count > bitpool) {
      bitadjust -= chop;
    } else {
      maxBitadjust = bitadjust;
      bitcount = count;
      bitadjust += chop;
    }
  }

  *excess = bitpool - bitcount;

  return maxBitadjust;
}

/*
 * The bit allocator trys to avoid single bit allocations except as a last
 * resort. So in the case where a bitneed of 1 was passed over during the
 * adsjustment phase 2 bits are now allocated.
 */
INLINE OI_INT allocAdjustedBits(uint8_t* dest, OI_INT bits, OI_INT excess) {
  if (bits < 16) {
    if (bits > 1) {
      if (excess) {
        ++bits;
        --excess;
      }
    } else if ((bits == 1) && (excess > 1)) {
      bits = 2;
      excess -= 2;
    } else {
      bits = 0;
    }
  } else {
    bits = 16;
  }
  *dest = (uint8_t)bits;
  return excess;
}

/*
 * Excess bits not allocated by allocaAdjustedBits are allocated round-robin.
 */
INLINE OI_INT allocExcessBits(uint8_t* dest, OI_INT excess) {
  if (*dest < 16) {
    *dest += 1;
    return excess - 1;
  } else {
    return excess;
  }
}

void oneChannelBitAllocation(OI_CODEC_SBC_COMMON_CONTEXT* common,
                             BITNEED_UNION1* bitneeds, OI_UINT ch,
                             OI_UINT bitcount) {
  const uint8_t nrof_subbands = common->frameInfo.nrof_subbands;
  OI_UINT excess;
  OI_UINT sb;
  OI_INT bitadjust;
  uint8_t RESTRICT* allocBits;

  {
    OI_UINT ex;
    bitadjust = adjustToFitBitpool(common->frameInfo.bitpool, bitneeds->uint32,
                                   nrof_subbands, bitcount, &ex);
    /* We want the compiler to put excess into a register */
    excess = ex;
  }

  /*
   * Allocate adjusted bits
   */
  allocBits = &common->bits.uint8[ch ? nrof_subbands : 0];

  sb = 0;
  while (sb < nrof_subbands) {
    excess = allocAdjustedBits(&allocBits[sb], bitneeds->uint8[sb] + bitadjust,
                               excess);
    ++sb;
  }
  sb = 0;
  while (excess) {
    excess = allocExcessBits(&allocBits[sb], excess);
    ++sb;
  }
}

void monoBitAllocation(OI_CODEC_SBC_COMMON_CONTEXT* common) {
  BITNEED_UNION1 bitneeds;
  OI_UINT bitcount;
  OI_UINT bitpoolPreference = 0;

  bitcount = computeBitneed(common, bitneeds.uint8, 0, &bitpoolPreference);

  oneChannelBitAllocation(common, &bitneeds, 0, bitcount);
}

/**
@}
*/