/*
 * Copyright (c) 2017, Alliance for Open Media. All rights reserved
 *
 * This source code is subject to the terms of the BSD 2 Clause License and
 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
 * was not distributed with this source code in the LICENSE file, you can
 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
 * Media Patent License 1.0 was not distributed with this source code in the
 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
 */

#include <emmintrin.h>
#include <smmintrin.h>

#include "./av1_rtcd.h"
#include "av1/common/warped_motion.h"

/* This is a modified version of 'warped_filter' from warped_motion.c:
   * Each coefficient is stored in 8 bits instead of 16 bits
   * The coefficients are rearranged in the column order 0, 2, 4, 6, 1, 3, 5, 7

     This is done in order to avoid overflow: Since the tap with the largest
     coefficient could be any of taps 2, 3, 4 or 5, we can't use the summation
     order ((0 + 1) + (4 + 5)) + ((2 + 3) + (6 + 7)) used in the regular
     convolve functions.

     Instead, we use the summation order
     ((0 + 2) + (4 + 6)) + ((1 + 3) + (5 + 7)).
     The rearrangement of coefficients in this table is so that we can get the
     coefficients into the correct order more quickly.
*/
/* clang-format off */
DECLARE_ALIGNED(8, static const int8_t,
                filter_8bit[WARPEDPIXEL_PREC_SHIFTS * 3 + 1][8]) = {
#if WARPEDPIXEL_PREC_BITS == 6
  // [-1, 0)
  { 0, 127,   0, 0,   0,   1, 0, 0}, { 0, 127,   0, 0,  -1,   2, 0, 0},
  { 1, 127,  -1, 0,  -3,   4, 0, 0}, { 1, 126,  -2, 0,  -4,   6, 1, 0},
  { 1, 126,  -3, 0,  -5,   8, 1, 0}, { 1, 125,  -4, 0,  -6,  11, 1, 0},
  { 1, 124,  -4, 0,  -7,  13, 1, 0}, { 2, 123,  -5, 0,  -8,  15, 1, 0},
  { 2, 122,  -6, 0,  -9,  18, 1, 0}, { 2, 121,  -6, 0, -10,  20, 1, 0},
  { 2, 120,  -7, 0, -11,  22, 2, 0}, { 2, 119,  -8, 0, -12,  25, 2, 0},
  { 3, 117,  -8, 0, -13,  27, 2, 0}, { 3, 116,  -9, 0, -13,  29, 2, 0},
  { 3, 114, -10, 0, -14,  32, 3, 0}, { 3, 113, -10, 0, -15,  35, 2, 0},
  { 3, 111, -11, 0, -15,  37, 3, 0}, { 3, 109, -11, 0, -16,  40, 3, 0},
  { 3, 108, -12, 0, -16,  42, 3, 0}, { 4, 106, -13, 0, -17,  45, 3, 0},
  { 4, 104, -13, 0, -17,  47, 3, 0}, { 4, 102, -14, 0, -17,  50, 3, 0},
  { 4, 100, -14, 0, -17,  52, 3, 0}, { 4,  98, -15, 0, -18,  55, 4, 0},
  { 4,  96, -15, 0, -18,  58, 3, 0}, { 4,  94, -16, 0, -18,  60, 4, 0},
  { 4,  91, -16, 0, -18,  63, 4, 0}, { 4,  89, -16, 0, -18,  65, 4, 0},
  { 4,  87, -17, 0, -18,  68, 4, 0}, { 4,  85, -17, 0, -18,  70, 4, 0},
  { 4,  82, -17, 0, -18,  73, 4, 0}, { 4,  80, -17, 0, -18,  75, 4, 0},
  { 4,  78, -18, 0, -18,  78, 4, 0}, { 4,  75, -18, 0, -17,  80, 4, 0},
  { 4,  73, -18, 0, -17,  82, 4, 0}, { 4,  70, -18, 0, -17,  85, 4, 0},
  { 4,  68, -18, 0, -17,  87, 4, 0}, { 4,  65, -18, 0, -16,  89, 4, 0},
  { 4,  63, -18, 0, -16,  91, 4, 0}, { 4,  60, -18, 0, -16,  94, 4, 0},
  { 3,  58, -18, 0, -15,  96, 4, 0}, { 4,  55, -18, 0, -15,  98, 4, 0},
  { 3,  52, -17, 0, -14, 100, 4, 0}, { 3,  50, -17, 0, -14, 102, 4, 0},
  { 3,  47, -17, 0, -13, 104, 4, 0}, { 3,  45, -17, 0, -13, 106, 4, 0},
  { 3,  42, -16, 0, -12, 108, 3, 0}, { 3,  40, -16, 0, -11, 109, 3, 0},
  { 3,  37, -15, 0, -11, 111, 3, 0}, { 2,  35, -15, 0, -10, 113, 3, 0},
  { 3,  32, -14, 0, -10, 114, 3, 0}, { 2,  29, -13, 0,  -9, 116, 3, 0},
  { 2,  27, -13, 0,  -8, 117, 3, 0}, { 2,  25, -12, 0,  -8, 119, 2, 0},
  { 2,  22, -11, 0,  -7, 120, 2, 0}, { 1,  20, -10, 0,  -6, 121, 2, 0},
  { 1,  18,  -9, 0,  -6, 122, 2, 0}, { 1,  15,  -8, 0,  -5, 123, 2, 0},
  { 1,  13,  -7, 0,  -4, 124, 1, 0}, { 1,  11,  -6, 0,  -4, 125, 1, 0},
  { 1,   8,  -5, 0,  -3, 126, 1, 0}, { 1,   6,  -4, 0,  -2, 126, 1, 0},
  { 0,   4,  -3, 0,  -1, 127, 1, 0}, { 0,   2,  -1, 0,   0, 127, 0, 0},
  // [0, 1)
  { 0,   0,   1, 0, 0, 127,   0,  0}, { 0,  -1,   2, 0, 0, 127,   0,  0},
  { 0,  -3,   4, 1, 1, 127,  -2,  0}, { 0,  -5,   6, 1, 1, 127,  -2,  0},
  { 0,  -6,   8, 1, 2, 126,  -3,  0}, {-1,  -7,  11, 2, 2, 126,  -4, -1},
  {-1,  -8,  13, 2, 3, 125,  -5, -1}, {-1, -10,  16, 3, 3, 124,  -6, -1},
  {-1, -11,  18, 3, 4, 123,  -7, -1}, {-1, -12,  20, 3, 4, 122,  -7, -1},
  {-1, -13,  23, 3, 4, 121,  -8, -1}, {-2, -14,  25, 4, 5, 120,  -9, -1},
  {-1, -15,  27, 4, 5, 119, -10, -1}, {-1, -16,  30, 4, 5, 118, -11, -1},
  {-2, -17,  33, 5, 6, 116, -12, -1}, {-2, -17,  35, 5, 6, 114, -12, -1},
  {-2, -18,  38, 5, 6, 113, -13, -1}, {-2, -19,  41, 6, 7, 111, -14, -2},
  {-2, -19,  43, 6, 7, 110, -15, -2}, {-2, -20,  46, 6, 7, 108, -15, -2},
  {-2, -20,  49, 6, 7, 106, -16, -2}, {-2, -21,  51, 7, 7, 104, -16, -2},
  {-2, -21,  54, 7, 7, 102, -17, -2}, {-2, -21,  56, 7, 8, 100, -18, -2},
  {-2, -22,  59, 7, 8,  98, -18, -2}, {-2, -22,  62, 7, 8,  96, -19, -2},
  {-2, -22,  64, 7, 8,  94, -19, -2}, {-2, -22,  67, 8, 8,  91, -20, -2},
  {-2, -22,  69, 8, 8,  89, -20, -2}, {-2, -22,  72, 8, 8,  87, -21, -2},
  {-2, -21,  74, 8, 8,  84, -21, -2}, {-2, -22,  77, 8, 8,  82, -21, -2},
  {-2, -21,  79, 8, 8,  79, -21, -2}, {-2, -21,  82, 8, 8,  77, -22, -2},
  {-2, -21,  84, 8, 8,  74, -21, -2}, {-2, -21,  87, 8, 8,  72, -22, -2},
  {-2, -20,  89, 8, 8,  69, -22, -2}, {-2, -20,  91, 8, 8,  67, -22, -2},
  {-2, -19,  94, 8, 7,  64, -22, -2}, {-2, -19,  96, 8, 7,  62, -22, -2},
  {-2, -18,  98, 8, 7,  59, -22, -2}, {-2, -18, 100, 8, 7,  56, -21, -2},
  {-2, -17, 102, 7, 7,  54, -21, -2}, {-2, -16, 104, 7, 7,  51, -21, -2},
  {-2, -16, 106, 7, 6,  49, -20, -2}, {-2, -15, 108, 7, 6,  46, -20, -2},
  {-2, -15, 110, 7, 6,  43, -19, -2}, {-2, -14, 111, 7, 6,  41, -19, -2},
  {-1, -13, 113, 6, 5,  38, -18, -2}, {-1, -12, 114, 6, 5,  35, -17, -2},
  {-1, -12, 116, 6, 5,  33, -17, -2}, {-1, -11, 118, 5, 4,  30, -16, -1},
  {-1, -10, 119, 5, 4,  27, -15, -1}, {-1,  -9, 120, 5, 4,  25, -14, -2},
  {-1,  -8, 121, 4, 3,  23, -13, -1}, {-1,  -7, 122, 4, 3,  20, -12, -1},
  {-1,  -7, 123, 4, 3,  18, -11, -1}, {-1,  -6, 124, 3, 3,  16, -10, -1},
  {-1,  -5, 125, 3, 2,  13,  -8, -1}, {-1,  -4, 126, 2, 2,  11,  -7, -1},
  { 0,  -3, 126, 2, 1,   8,  -6,  0}, { 0,  -2, 127, 1, 1,   6,  -5,  0},
  { 0,  -2, 127, 1, 1,   4,  -3,  0}, { 0,   0, 127, 0, 0,   2,  -1,  0},
  // [1, 2)
  { 0, 0, 127,   0, 0,   1,   0, 0}, { 0, 0, 127,   0, 0,  -1,   2, 0},
  { 0, 1, 127,  -1, 0,  -3,   4, 0}, { 0, 1, 126,  -2, 0,  -4,   6, 1},
  { 0, 1, 126,  -3, 0,  -5,   8, 1}, { 0, 1, 125,  -4, 0,  -6,  11, 1},
  { 0, 1, 124,  -4, 0,  -7,  13, 1}, { 0, 2, 123,  -5, 0,  -8,  15, 1},
  { 0, 2, 122,  -6, 0,  -9,  18, 1}, { 0, 2, 121,  -6, 0, -10,  20, 1},
  { 0, 2, 120,  -7, 0, -11,  22, 2}, { 0, 2, 119,  -8, 0, -12,  25, 2},
  { 0, 3, 117,  -8, 0, -13,  27, 2}, { 0, 3, 116,  -9, 0, -13,  29, 2},
  { 0, 3, 114, -10, 0, -14,  32, 3}, { 0, 3, 113, -10, 0, -15,  35, 2},
  { 0, 3, 111, -11, 0, -15,  37, 3}, { 0, 3, 109, -11, 0, -16,  40, 3},
  { 0, 3, 108, -12, 0, -16,  42, 3}, { 0, 4, 106, -13, 0, -17,  45, 3},
  { 0, 4, 104, -13, 0, -17,  47, 3}, { 0, 4, 102, -14, 0, -17,  50, 3},
  { 0, 4, 100, -14, 0, -17,  52, 3}, { 0, 4,  98, -15, 0, -18,  55, 4},
  { 0, 4,  96, -15, 0, -18,  58, 3}, { 0, 4,  94, -16, 0, -18,  60, 4},
  { 0, 4,  91, -16, 0, -18,  63, 4}, { 0, 4,  89, -16, 0, -18,  65, 4},
  { 0, 4,  87, -17, 0, -18,  68, 4}, { 0, 4,  85, -17, 0, -18,  70, 4},
  { 0, 4,  82, -17, 0, -18,  73, 4}, { 0, 4,  80, -17, 0, -18,  75, 4},
  { 0, 4,  78, -18, 0, -18,  78, 4}, { 0, 4,  75, -18, 0, -17,  80, 4},
  { 0, 4,  73, -18, 0, -17,  82, 4}, { 0, 4,  70, -18, 0, -17,  85, 4},
  { 0, 4,  68, -18, 0, -17,  87, 4}, { 0, 4,  65, -18, 0, -16,  89, 4},
  { 0, 4,  63, -18, 0, -16,  91, 4}, { 0, 4,  60, -18, 0, -16,  94, 4},
  { 0, 3,  58, -18, 0, -15,  96, 4}, { 0, 4,  55, -18, 0, -15,  98, 4},
  { 0, 3,  52, -17, 0, -14, 100, 4}, { 0, 3,  50, -17, 0, -14, 102, 4},
  { 0, 3,  47, -17, 0, -13, 104, 4}, { 0, 3,  45, -17, 0, -13, 106, 4},
  { 0, 3,  42, -16, 0, -12, 108, 3}, { 0, 3,  40, -16, 0, -11, 109, 3},
  { 0, 3,  37, -15, 0, -11, 111, 3}, { 0, 2,  35, -15, 0, -10, 113, 3},
  { 0, 3,  32, -14, 0, -10, 114, 3}, { 0, 2,  29, -13, 0,  -9, 116, 3},
  { 0, 2,  27, -13, 0,  -8, 117, 3}, { 0, 2,  25, -12, 0,  -8, 119, 2},
  { 0, 2,  22, -11, 0,  -7, 120, 2}, { 0, 1,  20, -10, 0,  -6, 121, 2},
  { 0, 1,  18,  -9, 0,  -6, 122, 2}, { 0, 1,  15,  -8, 0,  -5, 123, 2},
  { 0, 1,  13,  -7, 0,  -4, 124, 1}, { 0, 1,  11,  -6, 0,  -4, 125, 1},
  { 0, 1,   8,  -5, 0,  -3, 126, 1}, { 0, 1,   6,  -4, 0,  -2, 126, 1},
  { 0, 0,   4,  -3, 0,  -1, 127, 1}, { 0, 0,   2,  -1, 0,   0, 127, 0},
  // dummy (replicate row index 191)
  { 0, 0,   2,  -1, 0,   0, 127, 0},

#else
  // [-1, 0)
  { 0, 127,   0, 0,   0,   1, 0, 0}, { 1, 127,  -1, 0,  -3,   4, 0, 0},
  { 1, 126,  -3, 0,  -5,   8, 1, 0}, { 1, 124,  -4, 0,  -7,  13, 1, 0},
  { 2, 122,  -6, 0,  -9,  18, 1, 0}, { 2, 120,  -7, 0, -11,  22, 2, 0},
  { 3, 117,  -8, 0, -13,  27, 2, 0}, { 3, 114, -10, 0, -14,  32, 3, 0},
  { 3, 111, -11, 0, -15,  37, 3, 0}, { 3, 108, -12, 0, -16,  42, 3, 0},
  { 4, 104, -13, 0, -17,  47, 3, 0}, { 4, 100, -14, 0, -17,  52, 3, 0},
  { 4,  96, -15, 0, -18,  58, 3, 0}, { 4,  91, -16, 0, -18,  63, 4, 0},
  { 4,  87, -17, 0, -18,  68, 4, 0}, { 4,  82, -17, 0, -18,  73, 4, 0},
  { 4,  78, -18, 0, -18,  78, 4, 0}, { 4,  73, -18, 0, -17,  82, 4, 0},
  { 4,  68, -18, 0, -17,  87, 4, 0}, { 4,  63, -18, 0, -16,  91, 4, 0},
  { 3,  58, -18, 0, -15,  96, 4, 0}, { 3,  52, -17, 0, -14, 100, 4, 0},
  { 3,  47, -17, 0, -13, 104, 4, 0}, { 3,  42, -16, 0, -12, 108, 3, 0},
  { 3,  37, -15, 0, -11, 111, 3, 0}, { 3,  32, -14, 0, -10, 114, 3, 0},
  { 2,  27, -13, 0,  -8, 117, 3, 0}, { 2,  22, -11, 0,  -7, 120, 2, 0},
  { 1,  18,  -9, 0,  -6, 122, 2, 0}, { 1,  13,  -7, 0,  -4, 124, 1, 0},
  { 1,   8,  -5, 0,  -3, 126, 1, 0}, { 0,   4,  -3, 0,  -1, 127, 1, 0},
  // [0, 1)
  { 0,   0,   1, 0, 0, 127,   0,  0}, { 0,  -3,   4, 1, 1, 127,  -2,  0},
  { 0,  -6,   8, 1, 2, 126,  -3,  0}, {-1,  -8,  13, 2, 3, 125,  -5, -1},
  {-1, -11,  18, 3, 4, 123,  -7, -1}, {-1, -13,  23, 3, 4, 121,  -8, -1},
  {-1, -15,  27, 4, 5, 119, -10, -1}, {-2, -17,  33, 5, 6, 116, -12, -1},
  {-2, -18,  38, 5, 6, 113, -13, -1}, {-2, -19,  43, 6, 7, 110, -15, -2},
  {-2, -20,  49, 6, 7, 106, -16, -2}, {-2, -21,  54, 7, 7, 102, -17, -2},
  {-2, -22,  59, 7, 8,  98, -18, -2}, {-2, -22,  64, 7, 8,  94, -19, -2},
  {-2, -22,  69, 8, 8,  89, -20, -2}, {-2, -21,  74, 8, 8,  84, -21, -2},
  {-2, -21,  79, 8, 8,  79, -21, -2}, {-2, -21,  84, 8, 8,  74, -21, -2},
  {-2, -20,  89, 8, 8,  69, -22, -2}, {-2, -19,  94, 8, 7,  64, -22, -2},
  {-2, -18,  98, 8, 7,  59, -22, -2}, {-2, -17, 102, 7, 7,  54, -21, -2},
  {-2, -16, 106, 7, 6,  49, -20, -2}, {-2, -15, 110, 7, 6,  43, -19, -2},
  {-1, -13, 113, 6, 5,  38, -18, -2}, {-1, -12, 116, 6, 5,  33, -17, -2},
  {-1, -10, 119, 5, 4,  27, -15, -1}, {-1,  -8, 121, 4, 3,  23, -13, -1},
  {-1,  -7, 123, 4, 3,  18, -11, -1}, {-1,  -5, 125, 3, 2,  13,  -8, -1},
  { 0,  -3, 126, 2, 1,   8,  -6,  0}, { 0,  -2, 127, 1, 1,   4,  -3,  0},
  // [1, 2)
  { 0,  0, 127,   0, 0,   1,   0, 0}, { 0, 1, 127,  -1, 0,  -3,   4, 0},
  { 0,  1, 126,  -3, 0,  -5,   8, 1}, { 0, 1, 124,  -4, 0,  -7,  13, 1},
  { 0,  2, 122,  -6, 0,  -9,  18, 1}, { 0, 2, 120,  -7, 0, -11,  22, 2},
  { 0,  3, 117,  -8, 0, -13,  27, 2}, { 0, 3, 114, -10, 0, -14,  32, 3},
  { 0,  3, 111, -11, 0, -15,  37, 3}, { 0, 3, 108, -12, 0, -16,  42, 3},
  { 0,  4, 104, -13, 0, -17,  47, 3}, { 0, 4, 100, -14, 0, -17,  52, 3},
  { 0,  4,  96, -15, 0, -18,  58, 3}, { 0, 4,  91, -16, 0, -18,  63, 4},
  { 0,  4,  87, -17, 0, -18,  68, 4}, { 0, 4,  82, -17, 0, -18,  73, 4},
  { 0,  4,  78, -18, 0, -18,  78, 4}, { 0, 4,  73, -18, 0, -17,  82, 4},
  { 0,  4,  68, -18, 0, -17,  87, 4}, { 0, 4,  63, -18, 0, -16,  91, 4},
  { 0,  3,  58, -18, 0, -15,  96, 4}, { 0, 3,  52, -17, 0, -14, 100, 4},
  { 0,  3,  47, -17, 0, -13, 104, 4}, { 0, 3,  42, -16, 0, -12, 108, 3},
  { 0,  3,  37, -15, 0, -11, 111, 3}, { 0, 3,  32, -14, 0, -10, 114, 3},
  { 0,  2,  27, -13, 0,  -8, 117, 3}, { 0, 2,  22, -11, 0,  -7, 120, 2},
  { 0,  1,  18,  -9, 0,  -6, 122, 2}, { 0, 1,  13,  -7, 0,  -4, 124, 1},
  { 0,  1,   8,  -5, 0,  -3, 126, 1}, { 0, 0,   4,  -3, 0,  -1, 127, 1},
  // dummy (replicate row index 95)
  { 0, 0,   4,  -3, 0,  -1, 127, 1},
#endif  // WARPEDPIXEL_PREC_BITS == 6
};
/* clang-format on */

// Shuffle masks: we want to convert a sequence of bytes 0, 1, 2, ..., 15
// in an SSE register into two sequences:
// 0, 2, 2, 4, ..., 12, 12, 14, <don't care>
// 1, 3, 3, 5, ..., 13, 13, 15, <don't care>
static const uint8_t even_mask[16] = { 0, 2,  2,  4,  4,  6,  6,  8,
                                       8, 10, 10, 12, 12, 14, 14, 0 };
static const uint8_t odd_mask[16] = { 1, 3,  3,  5,  5,  7,  7,  9,
                                      9, 11, 11, 13, 13, 15, 15, 0 };

void av1_warp_affine_sse4_1(const int32_t *mat, const uint8_t *ref, int width,
                            int height, int stride, uint8_t *pred, int p_col,
                            int p_row, int p_width, int p_height, int p_stride,
                            int subsampling_x, int subsampling_y,
                            ConvolveParams *conv_params, int16_t alpha,
                            int16_t beta, int16_t gamma, int16_t delta) {
  int comp_avg = conv_params->do_average;
  __m128i tmp[15];
  int i, j, k;
  const int bd = 8;
  const int reduce_bits_horiz = conv_params->round_0;
  const int reduce_bits_vert = conv_params->is_compound
                                   ? conv_params->round_1
                                   : 2 * FILTER_BITS - reduce_bits_horiz;
  const int offset_bits_horiz = bd + FILTER_BITS - 1;
  assert(IMPLIES(conv_params->is_compound, conv_params->dst != NULL));

  const int w0 = conv_params->fwd_offset;
  const int w1 = conv_params->bck_offset;
  const __m128i wt0 = _mm_set1_epi32(w0);
  const __m128i wt1 = _mm_set1_epi32(w1);
  assert(FILTER_BITS == FILTER_BITS);

  /* Note: For this code to work, the left/right frame borders need to be
     extended by at least 13 pixels each. By the time we get here, other
     code will have set up this border, but we allow an explicit check
     for debugging purposes.
  */
  /*for (i = 0; i < height; ++i) {
    for (j = 0; j < 13; ++j) {
      assert(ref[i * stride - 13 + j] == ref[i * stride]);
      assert(ref[i * stride + width + j] == ref[i * stride + (width - 1)]);
    }
  }*/

  for (i = 0; i < p_height; i += 8) {
    for (j = 0; j < p_width; j += 8) {
      const int32_t src_x = (p_col + j + 4) << subsampling_x;
      const int32_t src_y = (p_row + i + 4) << subsampling_y;
      const int32_t dst_x = mat[2] * src_x + mat[3] * src_y + mat[0];
      const int32_t dst_y = mat[4] * src_x + mat[5] * src_y + mat[1];
      const int32_t x4 = dst_x >> subsampling_x;
      const int32_t y4 = dst_y >> subsampling_y;

      int32_t ix4 = x4 >> WARPEDMODEL_PREC_BITS;
      int32_t sx4 = x4 & ((1 << WARPEDMODEL_PREC_BITS) - 1);
      int32_t iy4 = y4 >> WARPEDMODEL_PREC_BITS;
      int32_t sy4 = y4 & ((1 << WARPEDMODEL_PREC_BITS) - 1);

      // Add in all the constant terms, including rounding and offset
      sx4 += alpha * (-4) + beta * (-4) + (1 << (WARPEDDIFF_PREC_BITS - 1)) +
             (WARPEDPIXEL_PREC_SHIFTS << WARPEDDIFF_PREC_BITS);
      sy4 += gamma * (-4) + delta * (-4) + (1 << (WARPEDDIFF_PREC_BITS - 1)) +
             (WARPEDPIXEL_PREC_SHIFTS << WARPEDDIFF_PREC_BITS);

      sx4 &= ~((1 << WARP_PARAM_REDUCE_BITS) - 1);
      sy4 &= ~((1 << WARP_PARAM_REDUCE_BITS) - 1);

      // Horizontal filter
      // If the block is aligned such that, after clamping, every sample
      // would be taken from the leftmost/rightmost column, then we can
      // skip the expensive horizontal filter.
      if (ix4 <= -7) {
        for (k = -7; k < AOMMIN(8, p_height - i); ++k) {
          int iy = iy4 + k;
          if (iy < 0)
            iy = 0;
          else if (iy > height - 1)
            iy = height - 1;
          tmp[k + 7] = _mm_set1_epi16(
              (1 << (bd + FILTER_BITS - reduce_bits_horiz - 1)) +
              ref[iy * stride] * (1 << (FILTER_BITS - reduce_bits_horiz)));
        }
      } else if (ix4 >= width + 6) {
        for (k = -7; k < AOMMIN(8, p_height - i); ++k) {
          int iy = iy4 + k;
          if (iy < 0)
            iy = 0;
          else if (iy > height - 1)
            iy = height - 1;
          tmp[k + 7] =
              _mm_set1_epi16((1 << (bd + FILTER_BITS - reduce_bits_horiz - 1)) +
                             ref[iy * stride + (width - 1)] *
                                 (1 << (FILTER_BITS - reduce_bits_horiz)));
        }
      } else {
        for (k = -7; k < AOMMIN(8, p_height - i); ++k) {
          int iy = iy4 + k;
          if (iy < 0)
            iy = 0;
          else if (iy > height - 1)
            iy = height - 1;
          int sx = sx4 + beta * (k + 4);

          // Load source pixels
          const __m128i src =
              _mm_loadu_si128((__m128i *)(ref + iy * stride + ix4 - 7));
          const __m128i src_even =
              _mm_shuffle_epi8(src, _mm_loadu_si128((__m128i *)even_mask));
          const __m128i src_odd =
              _mm_shuffle_epi8(src, _mm_loadu_si128((__m128i *)odd_mask));

          // Filter even-index pixels
          const __m128i tmp_0 = _mm_loadl_epi64((
              __m128i *)&filter_8bit[(sx + 0 * alpha) >> WARPEDDIFF_PREC_BITS]);
          const __m128i tmp_1 = _mm_loadl_epi64((
              __m128i *)&filter_8bit[(sx + 1 * alpha) >> WARPEDDIFF_PREC_BITS]);
          const __m128i tmp_2 = _mm_loadl_epi64((
              __m128i *)&filter_8bit[(sx + 2 * alpha) >> WARPEDDIFF_PREC_BITS]);
          const __m128i tmp_3 = _mm_loadl_epi64((
              __m128i *)&filter_8bit[(sx + 3 * alpha) >> WARPEDDIFF_PREC_BITS]);
          const __m128i tmp_4 = _mm_loadl_epi64((
              __m128i *)&filter_8bit[(sx + 4 * alpha) >> WARPEDDIFF_PREC_BITS]);
          const __m128i tmp_5 = _mm_loadl_epi64((
              __m128i *)&filter_8bit[(sx + 5 * alpha) >> WARPEDDIFF_PREC_BITS]);
          const __m128i tmp_6 = _mm_loadl_epi64((
              __m128i *)&filter_8bit[(sx + 6 * alpha) >> WARPEDDIFF_PREC_BITS]);
          const __m128i tmp_7 = _mm_loadl_epi64((
              __m128i *)&filter_8bit[(sx + 7 * alpha) >> WARPEDDIFF_PREC_BITS]);

          // Coeffs 0 2 0 2 4 6 4 6 1 3 1 3 5 7 5 7 for pixels 0 2
          const __m128i tmp_8 = _mm_unpacklo_epi16(tmp_0, tmp_2);
          // Coeffs 0 2 0 2 4 6 4 6 1 3 1 3 5 7 5 7 for pixels 1 3
          const __m128i tmp_9 = _mm_unpacklo_epi16(tmp_1, tmp_3);
          // Coeffs 0 2 0 2 4 6 4 6 1 3 1 3 5 7 5 7 for pixels 4 6
          const __m128i tmp_10 = _mm_unpacklo_epi16(tmp_4, tmp_6);
          // Coeffs 0 2 0 2 4 6 4 6 1 3 1 3 5 7 5 7 for pixels 5 7
          const __m128i tmp_11 = _mm_unpacklo_epi16(tmp_5, tmp_7);

          // Coeffs 0 2 0 2 0 2 0 2 4 6 4 6 4 6 4 6 for pixels 0 2 4 6
          const __m128i tmp_12 = _mm_unpacklo_epi32(tmp_8, tmp_10);
          // Coeffs 1 3 1 3 1 3 1 3 5 7 5 7 5 7 5 7 for pixels 0 2 4 6
          const __m128i tmp_13 = _mm_unpackhi_epi32(tmp_8, tmp_10);
          // Coeffs 0 2 0 2 0 2 0 2 4 6 4 6 4 6 4 6 for pixels 1 3 5 7
          const __m128i tmp_14 = _mm_unpacklo_epi32(tmp_9, tmp_11);
          // Coeffs 1 3 1 3 1 3 1 3 5 7 5 7 5 7 5 7 for pixels 1 3 5 7
          const __m128i tmp_15 = _mm_unpackhi_epi32(tmp_9, tmp_11);

          // Coeffs 0 2 for pixels 0 2 4 6 1 3 5 7
          const __m128i coeff_02 = _mm_unpacklo_epi64(tmp_12, tmp_14);
          // Coeffs 4 6 for pixels 0 2 4 6 1 3 5 7
          const __m128i coeff_46 = _mm_unpackhi_epi64(tmp_12, tmp_14);
          // Coeffs 1 3 for pixels 0 2 4 6 1 3 5 7
          const __m128i coeff_13 = _mm_unpacklo_epi64(tmp_13, tmp_15);
          // Coeffs 5 7 for pixels 0 2 4 6 1 3 5 7
          const __m128i coeff_57 = _mm_unpackhi_epi64(tmp_13, tmp_15);

          // The pixel order we need for 'src' is:
          // 0 2 2 4 4 6 6 8 1 3 3 5 5 7 7 9
          const __m128i src_02 = _mm_unpacklo_epi64(src_even, src_odd);
          const __m128i res_02 = _mm_maddubs_epi16(src_02, coeff_02);
          // 4 6 6 8 8 10 10 12 5 7 7 9 9 11 11 13
          const __m128i src_46 = _mm_unpacklo_epi64(_mm_srli_si128(src_even, 4),
                                                    _mm_srli_si128(src_odd, 4));
          const __m128i res_46 = _mm_maddubs_epi16(src_46, coeff_46);
          // 1 3 3 5 5 7 7 9 2 4 4 6 6 8 8 10
          const __m128i src_13 =
              _mm_unpacklo_epi64(src_odd, _mm_srli_si128(src_even, 2));
          const __m128i res_13 = _mm_maddubs_epi16(src_13, coeff_13);
          // 5 7 7 9 9 11 11 13 6 8 8 10 10 12 12 14
          const __m128i src_57 = _mm_unpacklo_epi64(
              _mm_srli_si128(src_odd, 4), _mm_srli_si128(src_even, 6));
          const __m128i res_57 = _mm_maddubs_epi16(src_57, coeff_57);

          const __m128i round_const = _mm_set1_epi16(
              (1 << offset_bits_horiz) + ((1 << reduce_bits_horiz) >> 1));

          // Note: The values res_02 + res_46 and res_13 + res_57 both
          // fit into int16s at this point, but their sum may be too wide to fit
          // into an int16. However, once we also add round_const, the sum of
          // all of these fits into a uint16.
          //
          // The wrapping behaviour of _mm_add_* is used here to make sure we
          // get the correct result despite converting between different
          // (implicit) types.
          const __m128i res_even = _mm_add_epi16(res_02, res_46);
          const __m128i res_odd = _mm_add_epi16(res_13, res_57);
          const __m128i res =
              _mm_add_epi16(_mm_add_epi16(res_even, res_odd), round_const);
          tmp[k + 7] = _mm_srl_epi16(res, _mm_cvtsi32_si128(reduce_bits_horiz));
        }
      }

      // Vertical filter
      for (k = -4; k < AOMMIN(4, p_height - i - 4); ++k) {
        int sy = sy4 + delta * (k + 4);

        // Load from tmp and rearrange pairs of consecutive rows into the
        // column order 0 0 2 2 4 4 6 6; 1 1 3 3 5 5 7 7
        const __m128i *src = tmp + (k + 4);
        const __m128i src_0 = _mm_unpacklo_epi16(src[0], src[1]);
        const __m128i src_2 = _mm_unpacklo_epi16(src[2], src[3]);
        const __m128i src_4 = _mm_unpacklo_epi16(src[4], src[5]);
        const __m128i src_6 = _mm_unpacklo_epi16(src[6], src[7]);

        // Filter even-index pixels
        const __m128i tmp_0 = _mm_loadu_si128(
            (__m128i *)(warped_filter +
                        ((sy + 0 * gamma) >> WARPEDDIFF_PREC_BITS)));
        const __m128i tmp_2 = _mm_loadu_si128(
            (__m128i *)(warped_filter +
                        ((sy + 2 * gamma) >> WARPEDDIFF_PREC_BITS)));
        const __m128i tmp_4 = _mm_loadu_si128(
            (__m128i *)(warped_filter +
                        ((sy + 4 * gamma) >> WARPEDDIFF_PREC_BITS)));
        const __m128i tmp_6 = _mm_loadu_si128(
            (__m128i *)(warped_filter +
                        ((sy + 6 * gamma) >> WARPEDDIFF_PREC_BITS)));

        const __m128i tmp_8 = _mm_unpacklo_epi32(tmp_0, tmp_2);
        const __m128i tmp_10 = _mm_unpacklo_epi32(tmp_4, tmp_6);
        const __m128i tmp_12 = _mm_unpackhi_epi32(tmp_0, tmp_2);
        const __m128i tmp_14 = _mm_unpackhi_epi32(tmp_4, tmp_6);

        const __m128i coeff_0 = _mm_unpacklo_epi64(tmp_8, tmp_10);
        const __m128i coeff_2 = _mm_unpackhi_epi64(tmp_8, tmp_10);
        const __m128i coeff_4 = _mm_unpacklo_epi64(tmp_12, tmp_14);
        const __m128i coeff_6 = _mm_unpackhi_epi64(tmp_12, tmp_14);

        const __m128i res_0 = _mm_madd_epi16(src_0, coeff_0);
        const __m128i res_2 = _mm_madd_epi16(src_2, coeff_2);
        const __m128i res_4 = _mm_madd_epi16(src_4, coeff_4);
        const __m128i res_6 = _mm_madd_epi16(src_6, coeff_6);

        const __m128i res_even = _mm_add_epi32(_mm_add_epi32(res_0, res_2),
                                               _mm_add_epi32(res_4, res_6));

        // Filter odd-index pixels
        const __m128i src_1 = _mm_unpackhi_epi16(src[0], src[1]);
        const __m128i src_3 = _mm_unpackhi_epi16(src[2], src[3]);
        const __m128i src_5 = _mm_unpackhi_epi16(src[4], src[5]);
        const __m128i src_7 = _mm_unpackhi_epi16(src[6], src[7]);

        const __m128i tmp_1 = _mm_loadu_si128(
            (__m128i *)(warped_filter +
                        ((sy + 1 * gamma) >> WARPEDDIFF_PREC_BITS)));
        const __m128i tmp_3 = _mm_loadu_si128(
            (__m128i *)(warped_filter +
                        ((sy + 3 * gamma) >> WARPEDDIFF_PREC_BITS)));
        const __m128i tmp_5 = _mm_loadu_si128(
            (__m128i *)(warped_filter +
                        ((sy + 5 * gamma) >> WARPEDDIFF_PREC_BITS)));
        const __m128i tmp_7 = _mm_loadu_si128(
            (__m128i *)(warped_filter +
                        ((sy + 7 * gamma) >> WARPEDDIFF_PREC_BITS)));

        const __m128i tmp_9 = _mm_unpacklo_epi32(tmp_1, tmp_3);
        const __m128i tmp_11 = _mm_unpacklo_epi32(tmp_5, tmp_7);
        const __m128i tmp_13 = _mm_unpackhi_epi32(tmp_1, tmp_3);
        const __m128i tmp_15 = _mm_unpackhi_epi32(tmp_5, tmp_7);

        const __m128i coeff_1 = _mm_unpacklo_epi64(tmp_9, tmp_11);
        const __m128i coeff_3 = _mm_unpackhi_epi64(tmp_9, tmp_11);
        const __m128i coeff_5 = _mm_unpacklo_epi64(tmp_13, tmp_15);
        const __m128i coeff_7 = _mm_unpackhi_epi64(tmp_13, tmp_15);

        const __m128i res_1 = _mm_madd_epi16(src_1, coeff_1);
        const __m128i res_3 = _mm_madd_epi16(src_3, coeff_3);
        const __m128i res_5 = _mm_madd_epi16(src_5, coeff_5);
        const __m128i res_7 = _mm_madd_epi16(src_7, coeff_7);

        const __m128i res_odd = _mm_add_epi32(_mm_add_epi32(res_1, res_3),
                                              _mm_add_epi32(res_5, res_7));

        // Rearrange pixels back into the order 0 ... 7
        __m128i res_lo = _mm_unpacklo_epi32(res_even, res_odd);
        __m128i res_hi = _mm_unpackhi_epi32(res_even, res_odd);

        if (conv_params->is_compound) {
          __m128i *const p =
              (__m128i *)&conv_params
                  ->dst[(i + k + 4) * conv_params->dst_stride + j];
          const __m128i round_const = _mm_set1_epi32(
              -(1 << (bd + 2 * FILTER_BITS - reduce_bits_horiz - 1)) +
              ((1 << (reduce_bits_vert)) >> 1));
          res_lo = _mm_add_epi32(res_lo, round_const);
          res_lo = _mm_sra_epi32(res_lo, _mm_cvtsi32_si128(reduce_bits_vert));
          if (conv_params->use_jnt_comp_avg) {
            if (comp_avg) {
              res_lo = _mm_add_epi32(_mm_mullo_epi32(_mm_loadu_si128(p), wt0),
                                     _mm_mullo_epi32(res_lo, wt1));
              res_lo = _mm_srai_epi32(res_lo, DIST_PRECISION_BITS);
            }
          } else {
            if (comp_avg)
              res_lo =
                  _mm_srai_epi32(_mm_add_epi32(_mm_loadu_si128(p), res_lo), 1);
          }

          _mm_storeu_si128(p, res_lo);
          if (p_width > 4) {
            res_hi = _mm_add_epi32(res_hi, round_const);
            res_hi = _mm_sra_epi32(res_hi, _mm_cvtsi32_si128(reduce_bits_vert));
            if (conv_params->use_jnt_comp_avg) {
              if (comp_avg) {
                res_hi =
                    _mm_add_epi32(_mm_mullo_epi32(_mm_loadu_si128(p + 1), wt0),
                                  _mm_mullo_epi32(res_hi, wt1));
                res_hi = _mm_srai_epi32(res_hi, DIST_PRECISION_BITS);
              }
            } else {
              if (comp_avg)
                res_hi = _mm_srai_epi32(
                    _mm_add_epi32(_mm_loadu_si128(p + 1), res_hi), 1);
            }

            _mm_storeu_si128(p + 1, res_hi);
          }
        } else {
          // Round and pack into 8 bits
          const __m128i round_const =
              _mm_set1_epi32(-(1 << (bd + reduce_bits_vert - 1)) +
                             ((1 << reduce_bits_vert) >> 1));

          const __m128i res_lo_round = _mm_srai_epi32(
              _mm_add_epi32(res_lo, round_const), reduce_bits_vert);
          const __m128i res_hi_round = _mm_srai_epi32(
              _mm_add_epi32(res_hi, round_const), reduce_bits_vert);

          const __m128i res_16bit = _mm_packs_epi32(res_lo_round, res_hi_round);
          __m128i res_8bit = _mm_packus_epi16(res_16bit, res_16bit);

          // Store, blending with 'pred' if needed
          __m128i *const p = (__m128i *)&pred[(i + k + 4) * p_stride + j];

          // Note: If we're outputting a 4x4 block, we need to be very careful
          // to only output 4 pixels at this point, to avoid encode/decode
          // mismatches when encoding with multiple threads.
          if (p_width == 4) {
            if (comp_avg) {
              const __m128i orig = _mm_cvtsi32_si128(*(uint32_t *)p);
              res_8bit = _mm_avg_epu8(res_8bit, orig);
            }
            *(uint32_t *)p = _mm_cvtsi128_si32(res_8bit);
          } else {
            if (comp_avg) res_8bit = _mm_avg_epu8(res_8bit, _mm_loadl_epi64(p));
            _mm_storel_epi64(p, res_8bit);
          }
        }
      }
    }
  }
}