Common/ASM_SSSE3/intrapred_ssse3.c

/*
 * 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 https://www.aomedia.org/license/software-license. 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 https://www.aomedia.org/license/patent-license.
 */

#include "EbDefinitions.h"
#include "common_dsp_rtcd.h"
#include <tmmintrin.h>

// Weights are quadratic from '1' to '1 / BlockSize', scaled by
// 2^sm_weight_log2_scale.
static const int32_t sm_weight_log2_scale = 8;

// max(block_size_wide[BLOCK_LARGEST], block_size_high[BLOCK_LARGEST])
#define MAX_BLOCK_DIM 64

/* clang-format off */
static const uint8_t sm_weight_arrays[2 * MAX_BLOCK_DIM] = {
    // Unused, because we always offset by bs, which is at least 2.
    0, 0,
    // bs = 2
    255, 128,
    // bs = 4
    255, 149, 85, 64,
    // bs = 8
    255, 197, 146, 105, 73, 50, 37, 32,
    // bs = 16
    255, 225, 196, 170, 145, 123, 102, 84, 68, 54, 43, 33, 26, 20, 17, 16,
    // bs = 32
    255, 240, 225, 210, 196, 182, 169, 157, 145, 133, 122, 111, 101, 92, 83, 74,
    66, 59, 52, 45, 39, 34, 29, 25, 21, 17, 14, 12, 10, 9, 8, 8,
    // bs = 64
    255, 248, 240, 233, 225, 218, 210, 203, 196, 189, 182, 176, 169, 163, 156,
    150, 144, 138, 133, 127, 121, 116, 111, 106, 101, 96, 91, 86, 82, 77, 73, 69,
    65, 61, 57, 54, 50, 47, 44, 41, 38, 35, 32, 29, 27, 25, 22, 20, 18, 16, 15,
    13, 12, 10, 9, 8, 7, 6, 6, 5, 5, 4, 4, 4,
};

// -----------------------------------------------------------------------------
// PAETH_PRED

// -----------------------------------------------------------------------------
// SMOOTH_PRED

// pixels[0]: above and below_pred interleave vector
// pixels[1]: left vector
// pixels[2]: right_pred vector
static INLINE void load_pixel_w4(const uint8_t *above, const uint8_t *left,
    int32_t height, __m128i *pixels) {
    __m128i d = _mm_cvtsi32_si128(((const uint32_t *)above)[0]);
    if (height == 4)
        pixels[1] = _mm_cvtsi32_si128(((const uint32_t *)left)[0]);
    else if (height == 8)
        pixels[1] = _mm_loadl_epi64(((const __m128i *)left));
    else
        pixels[1] = _mm_loadu_si128(((const __m128i *)left));

    pixels[2] = _mm_set1_epi16((uint16_t)above[3]);

    const __m128i bp = _mm_set1_epi16((uint16_t)left[height - 1]);
    const __m128i zero = _mm_setzero_si128();
    d = _mm_unpacklo_epi8(d, zero);
    pixels[0] = _mm_unpacklo_epi16(d, bp);
}

// weight_h[0]: weight_h vector
// weight_h[1]: scale - weight_h vector
// weight_h[2]: same as [0], second half for height = 16 only
// weight_h[3]: same as [1], second half for height = 16 only
// weight_w[0]: weights_w and scale - weights_w interleave vector
static INLINE void load_weight_w4(const uint8_t *weight_array, int32_t height,
    __m128i *weight_h, __m128i *weight_w) {
    const __m128i zero = _mm_setzero_si128();
    const __m128i d = _mm_set1_epi16((uint16_t)(1 << sm_weight_log2_scale));
    const __m128i t = _mm_cvtsi32_si128(((const uint32_t *)weight_array)[1]);
    weight_h[0] = _mm_unpacklo_epi8(t, zero);
    weight_h[1] = _mm_sub_epi16(d, weight_h[0]);
    weight_w[0] = _mm_unpacklo_epi16(weight_h[0], weight_h[1]);

    if (height == 8) {
        const __m128i weight = _mm_loadl_epi64((const __m128i *)&weight_array[8]);
        weight_h[0] = _mm_unpacklo_epi8(weight, zero);
        weight_h[1] = _mm_sub_epi16(d, weight_h[0]);
    }
    else if (height == 16) {
        const __m128i weight = _mm_loadu_si128((const __m128i *)&weight_array[16]);
        weight_h[0] = _mm_unpacklo_epi8(weight, zero);
        weight_h[1] = _mm_sub_epi16(d, weight_h[0]);
        weight_h[2] = _mm_unpackhi_epi8(weight, zero);
        weight_h[3] = _mm_sub_epi16(d, weight_h[2]);
    }
}

static INLINE void smooth_pred_4xh(const __m128i *pixel, const __m128i *wh,
    const __m128i *ww, int32_t h, uint8_t *dst,
    ptrdiff_t stride, int32_t second_half) {
    const __m128i round = _mm_set1_epi32((1 << sm_weight_log2_scale));
    const __m128i one = _mm_set1_epi16(1);
    const __m128i inc = _mm_set1_epi16(0x202);
    const __m128i gat = _mm_set1_epi32(0xc080400);
    __m128i rep = second_half ? _mm_set1_epi16((short)0x8008) : _mm_set1_epi16((short)0x8000);
    __m128i d = _mm_set1_epi16(0x100);

    for (int32_t i = 0; i < h; ++i) {
        const __m128i wg_wg = _mm_shuffle_epi8(wh[0], d);
        const __m128i sc_sc = _mm_shuffle_epi8(wh[1], d);
        const __m128i wh_sc = _mm_unpacklo_epi16(wg_wg, sc_sc);
        __m128i s = _mm_madd_epi16(pixel[0], wh_sc);

        __m128i b = _mm_shuffle_epi8(pixel[1], rep);
        b = _mm_unpacklo_epi16(b, pixel[2]);
        __m128i sum = _mm_madd_epi16(b, ww[0]);

        sum = _mm_add_epi32(s, sum);
        sum = _mm_add_epi32(sum, round);
        sum = _mm_srai_epi32(sum, 1 + sm_weight_log2_scale);

        sum = _mm_shuffle_epi8(sum, gat);
        *(uint32_t *)dst = _mm_cvtsi128_si32(sum);
        dst += stride;

        rep = _mm_add_epi16(rep, one);
        d = _mm_add_epi16(d, inc);
    }
}

void svt_aom_smooth_predictor_4x4_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above, const uint8_t *left) {
    __m128i pixels[3];
    load_pixel_w4(above, left, 4, pixels);

    __m128i wh[4], ww[2];
    load_weight_w4(sm_weight_arrays, 4, wh, ww);

    smooth_pred_4xh(pixels, wh, ww, 4, dst, stride, 0);
}

void svt_aom_smooth_predictor_4x8_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above, const uint8_t *left) {
    __m128i pixels[3];
    load_pixel_w4(above, left, 8, pixels);

    __m128i wh[4], ww[2];
    load_weight_w4(sm_weight_arrays, 8, wh, ww);

    smooth_pred_4xh(pixels, wh, ww, 8, dst, stride, 0);
}

void svt_aom_smooth_predictor_4x16_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels[3];
    load_pixel_w4(above, left, 16, pixels);

    __m128i wh[4], ww[2];
    load_weight_w4(sm_weight_arrays, 16, wh, ww);

    smooth_pred_4xh(pixels, wh, ww, 8, dst, stride, 0);
    dst += stride << 3;
    smooth_pred_4xh(pixels, &wh[2], ww, 8, dst, stride, 1);
}

// pixels[0]: above and below_pred interleave vector, first half
// pixels[1]: above and below_pred interleave vector, second half
// pixels[2]: left vector
// pixels[3]: right_pred vector
// pixels[4]: above and below_pred interleave vector, first half
// pixels[5]: above and below_pred interleave vector, second half
// pixels[6]: left vector + 16
// pixels[7]: right_pred vector
static INLINE void load_pixel_w8(const uint8_t *above, const uint8_t *left,
    int32_t height, __m128i *pixels) {
    const __m128i zero = _mm_setzero_si128();
    const __m128i bp = _mm_set1_epi16((uint16_t)left[height - 1]);
    __m128i d = _mm_loadl_epi64((const __m128i *)above);
    d = _mm_unpacklo_epi8(d, zero);
    pixels[0] = _mm_unpacklo_epi16(d, bp);
    pixels[1] = _mm_unpackhi_epi16(d, bp);

    pixels[3] = _mm_set1_epi16((uint16_t)above[7]);

    if (height == 4)
        pixels[2] = _mm_cvtsi32_si128(((const uint32_t *)left)[0]);
    else if (height == 8)
        pixels[2] = _mm_loadl_epi64((const __m128i *)left);
    else if (height == 16)
        pixels[2] = _mm_loadu_si128((const __m128i *)left);
    else {
        pixels[2] = _mm_loadu_si128((const __m128i *)left);
        pixels[4] = pixels[0];
        pixels[5] = pixels[1];
        pixels[6] = _mm_loadu_si128((const __m128i *)(left + 16));
        pixels[7] = pixels[3];
    }
}

// weight_h[0]: weight_h vector
// weight_h[1]: scale - weight_h vector
// weight_h[2]: same as [0], offset 8
// weight_h[3]: same as [1], offset 8
// weight_h[4]: same as [0], offset 16
// weight_h[5]: same as [1], offset 16
// weight_h[6]: same as [0], offset 24
// weight_h[7]: same as [1], offset 24
// weight_w[0]: weights_w and scale - weights_w interleave vector, first half
// weight_w[1]: weights_w and scale - weights_w interleave vector, second half
static INLINE void load_weight_w8(const uint8_t *weight_array, int32_t height,
    __m128i *weight_h, __m128i *weight_w) {
    const __m128i zero = _mm_setzero_si128();
    const int32_t we_offset = height < 8 ? 4 : 8;
    __m128i we = _mm_loadu_si128((const __m128i *)&weight_array[we_offset]);
    weight_h[0] = _mm_unpacklo_epi8(we, zero);
    const __m128i d = _mm_set1_epi16((uint16_t)(1 << sm_weight_log2_scale));
    weight_h[1] = _mm_sub_epi16(d, weight_h[0]);

    if (height == 4) {
        we = _mm_srli_si128(we, 4);
        __m128i tmp1 = _mm_unpacklo_epi8(we, zero);
        __m128i tmp2 = _mm_sub_epi16(d, tmp1);
        weight_w[0] = _mm_unpacklo_epi16(tmp1, tmp2);
        weight_w[1] = _mm_unpackhi_epi16(tmp1, tmp2);
    }
    else {
        weight_w[0] = _mm_unpacklo_epi16(weight_h[0], weight_h[1]);
        weight_w[1] = _mm_unpackhi_epi16(weight_h[0], weight_h[1]);
    }

    if (height == 16) {
        we = _mm_loadu_si128((const __m128i *)&weight_array[16]);
        weight_h[0] = _mm_unpacklo_epi8(we, zero);
        weight_h[1] = _mm_sub_epi16(d, weight_h[0]);
        weight_h[2] = _mm_unpackhi_epi8(we, zero);
        weight_h[3] = _mm_sub_epi16(d, weight_h[2]);
    }
    else if (height == 32) {
        const __m128i weight_lo =
            _mm_loadu_si128((const __m128i *)&weight_array[32]);
        weight_h[0] = _mm_unpacklo_epi8(weight_lo, zero);
        weight_h[1] = _mm_sub_epi16(d, weight_h[0]);
        weight_h[2] = _mm_unpackhi_epi8(weight_lo, zero);
        weight_h[3] = _mm_sub_epi16(d, weight_h[2]);
        const __m128i weight_hi =
            _mm_loadu_si128((const __m128i *)&weight_array[32 + 16]);
        weight_h[4] = _mm_unpacklo_epi8(weight_hi, zero);
        weight_h[5] = _mm_sub_epi16(d, weight_h[4]);
        weight_h[6] = _mm_unpackhi_epi8(weight_hi, zero);
        weight_h[7] = _mm_sub_epi16(d, weight_h[6]);
    }
}

static INLINE void smooth_pred_8xh(const __m128i *pixels, const __m128i *wh,
    const __m128i *ww, int32_t h, uint8_t *dst,
    ptrdiff_t stride, int32_t second_half) {
    const __m128i round = _mm_set1_epi32((1 << sm_weight_log2_scale));
    const __m128i one = _mm_set1_epi16(1);
    const __m128i inc = _mm_set1_epi16(0x202);
    const __m128i gat = _mm_set_epi32(0, 0, 0xe0c0a08, 0x6040200);

    __m128i rep = second_half ? _mm_set1_epi16((short)0x8008) : _mm_set1_epi16((short)0x8000);
    __m128i d = _mm_set1_epi16(0x100);

    int32_t i;
    for (i = 0; i < h; ++i) {
        const __m128i wg_wg = _mm_shuffle_epi8(wh[0], d);
        const __m128i sc_sc = _mm_shuffle_epi8(wh[1], d);
        const __m128i wh_sc = _mm_unpacklo_epi16(wg_wg, sc_sc);
        __m128i s0 = _mm_madd_epi16(pixels[0], wh_sc);
        __m128i s1 = _mm_madd_epi16(pixels[1], wh_sc);

        __m128i b = _mm_shuffle_epi8(pixels[2], rep);
        b = _mm_unpacklo_epi16(b, pixels[3]);
        __m128i sum0 = _mm_madd_epi16(b, ww[0]);
        __m128i sum1 = _mm_madd_epi16(b, ww[1]);

        s0 = _mm_add_epi32(s0, sum0);
        s0 = _mm_add_epi32(s0, round);
        s0 = _mm_srai_epi32(s0, 1 + sm_weight_log2_scale);

        s1 = _mm_add_epi32(s1, sum1);
        s1 = _mm_add_epi32(s1, round);
        s1 = _mm_srai_epi32(s1, 1 + sm_weight_log2_scale);

        sum0 = _mm_packus_epi16(s0, s1);
        sum0 = _mm_shuffle_epi8(sum0, gat);
        _mm_storel_epi64((__m128i *)dst, sum0);
        dst += stride;

        rep = _mm_add_epi16(rep, one);
        d = _mm_add_epi16(d, inc);
    }
}

void svt_aom_smooth_predictor_8x4_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above, const uint8_t *left) {
    __m128i pixels[4];
    load_pixel_w8(above, left, 4, pixels);

    __m128i wh[4], ww[2];
    load_weight_w8(sm_weight_arrays, 4, wh, ww);

    smooth_pred_8xh(pixels, wh, ww, 4, dst, stride, 0);
}

void svt_aom_smooth_predictor_8x8_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above, const uint8_t *left) {
    __m128i pixels[4];
    load_pixel_w8(above, left, 8, pixels);

    __m128i wh[4], ww[2];
    load_weight_w8(sm_weight_arrays, 8, wh, ww);

    smooth_pred_8xh(pixels, wh, ww, 8, dst, stride, 0);
}

void svt_aom_smooth_predictor_8x16_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels[4];
    load_pixel_w8(above, left, 16, pixels);

    __m128i wh[4], ww[2];
    load_weight_w8(sm_weight_arrays, 16, wh, ww);

    smooth_pred_8xh(pixels, wh, ww, 8, dst, stride, 0);
    dst += stride << 3;
    smooth_pred_8xh(pixels, &wh[2], ww, 8, dst, stride, 1);
}

void svt_aom_smooth_predictor_8x32_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels[8];
    load_pixel_w8(above, left, 32, pixels);

    __m128i wh[8], ww[2];
    load_weight_w8(sm_weight_arrays, 32, wh, ww);

    smooth_pred_8xh(&pixels[0], wh, ww, 8, dst, stride, 0);
    dst += stride << 3;
    smooth_pred_8xh(&pixels[0], &wh[2], ww, 8, dst, stride, 1);
    dst += stride << 3;
    smooth_pred_8xh(&pixels[4], &wh[4], ww, 8, dst, stride, 0);
    dst += stride << 3;
    smooth_pred_8xh(&pixels[4], &wh[6], ww, 8, dst, stride, 1);
}

static INLINE void smooth_predictor_wxh(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left, uint32_t bw,
    uint32_t bh) {
    const uint8_t *const sm_weights_w = sm_weight_arrays + bw;
    const uint8_t *const sm_weights_h = sm_weight_arrays + bh;
    const __m128i zero = _mm_setzero_si128();
    const __m128i scale_value =
        _mm_set1_epi16((uint16_t)(1 << sm_weight_log2_scale));
    const __m128i bottom_left = _mm_cvtsi32_si128((uint32_t)left[bh - 1]);
    const __m128i dup16 = _mm_set1_epi32(0x01000100);
    const __m128i top_right =
        _mm_shuffle_epi8(_mm_cvtsi32_si128((uint32_t)above[bw - 1]), dup16);
    const __m128i gat = _mm_set_epi32(0, 0, 0xe0c0a08, 0x6040200);
    const __m128i round = _mm_set1_epi32((uint16_t)(1 << sm_weight_log2_scale));

    for (uint32_t y = 0; y < bh; ++y) {
        const __m128i weights_y = _mm_cvtsi32_si128((uint32_t)sm_weights_h[y]);
        const __m128i left_y = _mm_cvtsi32_si128((uint32_t)left[y]);
        const __m128i scale_m_weights_y = _mm_sub_epi16(scale_value, weights_y);
        __m128i pred_scaled_bl = _mm_mullo_epi16(scale_m_weights_y, bottom_left);
        const __m128i wl_y =
            _mm_shuffle_epi32(_mm_unpacklo_epi16(weights_y, left_y), 0);
        pred_scaled_bl = _mm_add_epi32(pred_scaled_bl, round);
        pred_scaled_bl = _mm_shuffle_epi32(pred_scaled_bl, 0);

        for (uint32_t x = 0; x < bw; x += 8) {
            const __m128i top_x = _mm_loadl_epi64((const __m128i *)(above + x));
            const __m128i weights_x =
                _mm_loadl_epi64((const __m128i *)(sm_weights_w + x));
            const __m128i tw_x = _mm_unpacklo_epi8(top_x, weights_x);
            const __m128i tw_x_lo = _mm_unpacklo_epi8(tw_x, zero);
            const __m128i tw_x_hi = _mm_unpackhi_epi8(tw_x, zero);

            __m128i pred_lo = _mm_madd_epi16(tw_x_lo, wl_y);
            __m128i pred_hi = _mm_madd_epi16(tw_x_hi, wl_y);

            const __m128i scale_m_weights_x =
                _mm_sub_epi16(scale_value, _mm_unpacklo_epi8(weights_x, zero));
            const __m128i swxtr = _mm_mullo_epi16(scale_m_weights_x, top_right);
            const __m128i swxtr_lo = _mm_unpacklo_epi16(swxtr, zero);
            const __m128i swxtr_hi = _mm_unpackhi_epi16(swxtr, zero);

            pred_lo = _mm_add_epi32(pred_lo, pred_scaled_bl);
            pred_hi = _mm_add_epi32(pred_hi, pred_scaled_bl);

            pred_lo = _mm_add_epi32(pred_lo, swxtr_lo);
            pred_hi = _mm_add_epi32(pred_hi, swxtr_hi);

            pred_lo = _mm_srai_epi32(pred_lo, (1 + sm_weight_log2_scale));
            pred_hi = _mm_srai_epi32(pred_hi, (1 + sm_weight_log2_scale));

            __m128i pred = _mm_packus_epi16(pred_lo, pred_hi);
            pred = _mm_shuffle_epi8(pred, gat);
            _mm_storel_epi64((__m128i *)(dst + x), pred);
        }
        dst += stride;
    }
}

void svt_aom_smooth_predictor_16x4_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_predictor_wxh(dst, stride, above, left, 16, 4);
}

void svt_aom_smooth_predictor_16x8_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_predictor_wxh(dst, stride, above, left, 16, 8);
}

void svt_aom_smooth_predictor_16x16_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_predictor_wxh(dst, stride, above, left, 16, 16);
}

void svt_aom_smooth_predictor_16x32_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_predictor_wxh(dst, stride, above, left, 16, 32);
}

void svt_aom_smooth_predictor_32x8_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_predictor_wxh(dst, stride, above, left, 32, 8);
}

void svt_aom_smooth_predictor_32x16_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_predictor_wxh(dst, stride, above, left, 32, 16);
}

void svt_aom_smooth_predictor_32x32_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_predictor_wxh(dst, stride, above, left, 32, 32);
}

void svt_aom_smooth_predictor_32x64_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_predictor_wxh(dst, stride, above, left, 32, 64);
}

void svt_aom_smooth_predictor_64x64_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_predictor_wxh(dst, stride, above, left, 64, 64);
}

void svt_aom_smooth_predictor_64x32_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_predictor_wxh(dst, stride, above, left, 64, 32);
}

void svt_aom_smooth_predictor_64x16_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_predictor_wxh(dst, stride, above, left, 64, 16);
}

void svt_aom_smooth_predictor_16x64_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_predictor_wxh(dst, stride, above, left, 16, 64);
}

// -----------------------------------------------------------------------------
// SMOOTH_V_PRED

// pixels[0]: above and below_pred interleave vector
static INLINE void load_pixel_v_w4(const uint8_t *above, const uint8_t *left,
    int32_t height, __m128i *pixels) {
    const __m128i zero = _mm_setzero_si128();
    __m128i d = _mm_cvtsi32_si128(((const uint32_t *)above)[0]);
    const __m128i bp = _mm_set1_epi16((uint16_t)left[height - 1]);
    d = _mm_unpacklo_epi8(d, zero);
    pixels[0] = _mm_unpacklo_epi16(d, bp);
}

// weights[0]: weights_h vector
// weights[1]: scale - weights_h vector
static INLINE void load_weight_v_w4(const uint8_t *weight_array, int32_t height,
    __m128i *weights) {
    const __m128i zero = _mm_setzero_si128();
    const __m128i d = _mm_set1_epi16((uint16_t)(1 << sm_weight_log2_scale));

    if (height == 4) {
        const __m128i weight =
            _mm_cvtsi32_si128(((const uint32_t *)weight_array)[1]);
        weights[0] = _mm_unpacklo_epi8(weight, zero);
        weights[1] = _mm_sub_epi16(d, weights[0]);
    }
    else if (height == 8) {
        const __m128i weight = _mm_loadl_epi64((const __m128i *)&weight_array[8]);
        weights[0] = _mm_unpacklo_epi8(weight, zero);
        weights[1] = _mm_sub_epi16(d, weights[0]);
    }
    else {
        const __m128i weight = _mm_loadu_si128((const __m128i *)&weight_array[16]);
        weights[0] = _mm_unpacklo_epi8(weight, zero);
        weights[1] = _mm_sub_epi16(d, weights[0]);
        weights[2] = _mm_unpackhi_epi8(weight, zero);
        weights[3] = _mm_sub_epi16(d, weights[2]);
    }
}

static INLINE void smooth_v_pred_4xh(const __m128i *pixel,
    const __m128i *weight, int32_t h, uint8_t *dst,
    ptrdiff_t stride) {
    const __m128i pred_round = _mm_set1_epi32((1 << (sm_weight_log2_scale - 1)));
    const __m128i inc = _mm_set1_epi16(0x202);
    const __m128i gat = _mm_set1_epi32(0xc080400);
    __m128i d = _mm_set1_epi16(0x100);

    for (int32_t i = 0; i < h; ++i) {
        const __m128i wg_wg = _mm_shuffle_epi8(weight[0], d);
        const __m128i sc_sc = _mm_shuffle_epi8(weight[1], d);
        const __m128i wh_sc = _mm_unpacklo_epi16(wg_wg, sc_sc);
        __m128i sum = _mm_madd_epi16(pixel[0], wh_sc);
        sum = _mm_add_epi32(sum, pred_round);
        sum = _mm_srai_epi32(sum, sm_weight_log2_scale);
        sum = _mm_shuffle_epi8(sum, gat);
        *(uint32_t *)dst = _mm_cvtsi128_si32(sum);
        dst += stride;
        d = _mm_add_epi16(d, inc);
    }
}

void svt_aom_smooth_v_predictor_4x4_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels;
    load_pixel_v_w4(above, left, 4, &pixels);

    __m128i weights[2];
    load_weight_v_w4(sm_weight_arrays, 4, weights);

    smooth_v_pred_4xh(&pixels, weights, 4, dst, stride);
}

void svt_aom_smooth_v_predictor_4x8_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels;
    load_pixel_v_w4(above, left, 8, &pixels);

    __m128i weights[2];
    load_weight_v_w4(sm_weight_arrays, 8, weights);

    smooth_v_pred_4xh(&pixels, weights, 8, dst, stride);
}

void svt_aom_smooth_v_predictor_4x16_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels;
    load_pixel_v_w4(above, left, 16, &pixels);

    __m128i weights[4];
    load_weight_v_w4(sm_weight_arrays, 16, weights);

    smooth_v_pred_4xh(&pixels, weights, 8, dst, stride);
    dst += stride << 3;
    smooth_v_pred_4xh(&pixels, &weights[2], 8, dst, stride);
}

// pixels[0]: above and below_pred interleave vector, first half
// pixels[1]: above and below_pred interleave vector, second half
static INLINE void load_pixel_v_w8(const uint8_t *above, const uint8_t *left,
    int32_t height, __m128i *pixels) {
    const __m128i zero = _mm_setzero_si128();
    __m128i d = _mm_loadl_epi64((const __m128i *)above);
    const __m128i bp = _mm_set1_epi16((uint16_t)left[height - 1]);
    d = _mm_unpacklo_epi8(d, zero);
    pixels[0] = _mm_unpacklo_epi16(d, bp);
    pixels[1] = _mm_unpackhi_epi16(d, bp);
}

// weight_h[0]: weight_h vector
// weight_h[1]: scale - weight_h vector
// weight_h[2]: same as [0], offset 8
// weight_h[3]: same as [1], offset 8
// weight_h[4]: same as [0], offset 16
// weight_h[5]: same as [1], offset 16
// weight_h[6]: same as [0], offset 24
// weight_h[7]: same as [1], offset 24
static INLINE void load_weight_v_w8(const uint8_t *weight_array, int32_t height,
    __m128i *weight_h) {
    const __m128i zero = _mm_setzero_si128();
    const __m128i d = _mm_set1_epi16((uint16_t)(1 << sm_weight_log2_scale));

    if (height < 16) {
        const int32_t offset = height < 8 ? 4 : 8;
        const __m128i weight =
            _mm_loadu_si128((const __m128i *)&weight_array[offset]);
        weight_h[0] = _mm_unpacklo_epi8(weight, zero);
        weight_h[1] = _mm_sub_epi16(d, weight_h[0]);
    }
    else if (height == 16) {
        const __m128i weight = _mm_loadu_si128((const __m128i *)&weight_array[16]);
        weight_h[0] = _mm_unpacklo_epi8(weight, zero);
        weight_h[1] = _mm_sub_epi16(d, weight_h[0]);
        weight_h[2] = _mm_unpackhi_epi8(weight, zero);
        weight_h[3] = _mm_sub_epi16(d, weight_h[2]);
    }
    else {
        const __m128i weight_lo =
            _mm_loadu_si128((const __m128i *)&weight_array[32]);
        weight_h[0] = _mm_unpacklo_epi8(weight_lo, zero);
        weight_h[1] = _mm_sub_epi16(d, weight_h[0]);
        weight_h[2] = _mm_unpackhi_epi8(weight_lo, zero);
        weight_h[3] = _mm_sub_epi16(d, weight_h[2]);
        const __m128i weight_hi =
            _mm_loadu_si128((const __m128i *)&weight_array[32 + 16]);
        weight_h[4] = _mm_unpacklo_epi8(weight_hi, zero);
        weight_h[5] = _mm_sub_epi16(d, weight_h[4]);
        weight_h[6] = _mm_unpackhi_epi8(weight_hi, zero);
        weight_h[7] = _mm_sub_epi16(d, weight_h[6]);
    }
}

static INLINE void smooth_v_pred_8xh(const __m128i *pixels, const __m128i *wh,
    int32_t h, uint8_t *dst, ptrdiff_t stride) {
    const __m128i pred_round = _mm_set1_epi32((1 << (sm_weight_log2_scale - 1)));
    const __m128i inc = _mm_set1_epi16(0x202);
    const __m128i gat = _mm_set_epi32(0, 0, 0xe0c0a08, 0x6040200);
    __m128i d = _mm_set1_epi16(0x100);

    for (int32_t i = 0; i < h; ++i) {
        const __m128i wg_wg = _mm_shuffle_epi8(wh[0], d);
        const __m128i sc_sc = _mm_shuffle_epi8(wh[1], d);
        const __m128i wh_sc = _mm_unpacklo_epi16(wg_wg, sc_sc);
        __m128i s0 = _mm_madd_epi16(pixels[0], wh_sc);
        __m128i s1 = _mm_madd_epi16(pixels[1], wh_sc);

        s0 = _mm_add_epi32(s0, pred_round);
        s0 = _mm_srai_epi32(s0, sm_weight_log2_scale);

        s1 = _mm_add_epi32(s1, pred_round);
        s1 = _mm_srai_epi32(s1, sm_weight_log2_scale);

        __m128i sum01 = _mm_packus_epi16(s0, s1);
        sum01 = _mm_shuffle_epi8(sum01, gat);
        _mm_storel_epi64((__m128i *)dst, sum01);
        dst += stride;

        d = _mm_add_epi16(d, inc);
    }
}

void svt_aom_smooth_v_predictor_8x4_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels[2];
    load_pixel_v_w8(above, left, 4, pixels);

    __m128i wh[2];
    load_weight_v_w8(sm_weight_arrays, 4, wh);

    smooth_v_pred_8xh(pixels, wh, 4, dst, stride);
}

void svt_aom_smooth_v_predictor_8x8_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels[2];
    load_pixel_v_w8(above, left, 8, pixels);

    __m128i wh[2];
    load_weight_v_w8(sm_weight_arrays, 8, wh);

    smooth_v_pred_8xh(pixels, wh, 8, dst, stride);
}

void svt_aom_smooth_v_predictor_8x16_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels[2];
    load_pixel_v_w8(above, left, 16, pixels);

    __m128i wh[4];
    load_weight_v_w8(sm_weight_arrays, 16, wh);

    smooth_v_pred_8xh(pixels, wh, 8, dst, stride);
    dst += stride << 3;
    smooth_v_pred_8xh(pixels, &wh[2], 8, dst, stride);
}

void svt_aom_smooth_v_predictor_8x32_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels[2];
    load_pixel_v_w8(above, left, 32, pixels);

    __m128i wh[8];
    load_weight_v_w8(sm_weight_arrays, 32, wh);

    smooth_v_pred_8xh(pixels, &wh[0], 8, dst, stride);
    dst += stride << 3;
    smooth_v_pred_8xh(pixels, &wh[2], 8, dst, stride);
    dst += stride << 3;
    smooth_v_pred_8xh(pixels, &wh[4], 8, dst, stride);
    dst += stride << 3;
    smooth_v_pred_8xh(pixels, &wh[6], 8, dst, stride);
}

static INLINE void smooth_v_predictor_wxh(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left, uint32_t bw,
    uint32_t bh) {
    const uint8_t *const sm_weights_h = sm_weight_arrays + bh;
    const __m128i zero = _mm_setzero_si128();
    const __m128i scale_value =
        _mm_set1_epi16((uint16_t)(1 << sm_weight_log2_scale));
    const __m128i dup16 = _mm_set1_epi32(0x01000100);
    const __m128i bottom_left =
        _mm_shuffle_epi8(_mm_cvtsi32_si128((uint32_t)left[bh - 1]), dup16);
    const __m128i gat = _mm_set_epi32(0, 0, 0xe0c0a08, 0x6040200);
    const __m128i round =
        _mm_set1_epi32((uint16_t)(1 << (sm_weight_log2_scale - 1)));

    for (uint32_t y = 0; y < bh; ++y) {
        const __m128i weights_y = _mm_cvtsi32_si128((uint32_t)sm_weights_h[y]);
        const __m128i scale_m_weights_y =
            _mm_shuffle_epi8(_mm_sub_epi16(scale_value, weights_y), dup16);
        const __m128i wl_y =
            _mm_shuffle_epi32(_mm_unpacklo_epi16(weights_y, bottom_left), 0);

        for (uint32_t x = 0; x < bw; x += 8) {
            const __m128i top_x = _mm_loadl_epi64((const __m128i *)(above + x));
            // 8 -> 16
            const __m128i tw_x = _mm_unpacklo_epi8(top_x, zero);
            const __m128i tw_x_lo = _mm_unpacklo_epi16(tw_x, scale_m_weights_y);
            const __m128i tw_x_hi = _mm_unpackhi_epi16(tw_x, scale_m_weights_y);
            // top_x * weights_y + scale_m_weights_y * bottom_left
            __m128i pred_lo = _mm_madd_epi16(tw_x_lo, wl_y);
            __m128i pred_hi = _mm_madd_epi16(tw_x_hi, wl_y);

            pred_lo = _mm_add_epi32(pred_lo, round);
            pred_hi = _mm_add_epi32(pred_hi, round);
            pred_lo = _mm_srai_epi32(pred_lo, sm_weight_log2_scale);
            pred_hi = _mm_srai_epi32(pred_hi, sm_weight_log2_scale);

            __m128i pred = _mm_packus_epi16(pred_lo, pred_hi);
            pred = _mm_shuffle_epi8(pred, gat);
            _mm_storel_epi64((__m128i *)(dst + x), pred);
        }
        dst += stride;
    }
}

void svt_aom_smooth_v_predictor_16x4_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_v_predictor_wxh(dst, stride, above, left, 16, 4);
}

void svt_aom_smooth_v_predictor_16x8_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_v_predictor_wxh(dst, stride, above, left, 16, 8);
}

void svt_aom_smooth_v_predictor_16x16_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_v_predictor_wxh(dst, stride, above, left, 16, 16);
}

void svt_aom_smooth_v_predictor_16x32_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_v_predictor_wxh(dst, stride, above, left, 16, 32);
}

void svt_aom_smooth_v_predictor_32x8_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_v_predictor_wxh(dst, stride, above, left, 32, 8);
}

void svt_aom_smooth_v_predictor_32x16_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_v_predictor_wxh(dst, stride, above, left, 32, 16);
}

void svt_aom_smooth_v_predictor_32x32_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_v_predictor_wxh(dst, stride, above, left, 32, 32);
}

void svt_aom_smooth_v_predictor_32x64_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_v_predictor_wxh(dst, stride, above, left, 32, 64);
}

void svt_aom_smooth_v_predictor_64x64_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_v_predictor_wxh(dst, stride, above, left, 64, 64);
}

void svt_aom_smooth_v_predictor_64x32_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_v_predictor_wxh(dst, stride, above, left, 64, 32);
}

void svt_aom_smooth_v_predictor_64x16_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_v_predictor_wxh(dst, stride, above, left, 64, 16);
}

void svt_aom_smooth_v_predictor_16x64_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_v_predictor_wxh(dst, stride, above, left, 16, 64);
}

// -----------------------------------------------------------------------------
// SMOOTH_H_PRED

// pixels[0]: left vector
// pixels[1]: right_pred vector
static INLINE void load_pixel_h_w4(const uint8_t *above, const uint8_t *left,
    int32_t height, __m128i *pixels) {
    if (height == 4)
        pixels[0] = _mm_cvtsi32_si128(((const uint32_t *)left)[0]);
    else if (height == 8)
        pixels[0] = _mm_loadl_epi64(((const __m128i *)left));
    else
        pixels[0] = _mm_loadu_si128(((const __m128i *)left));
    pixels[1] = _mm_set1_epi16((uint16_t)above[3]);
}

// weights[0]: weights_w and scale - weights_w interleave vector
static INLINE void load_weight_h_w4(const uint8_t *weight_array, int32_t height,
    __m128i *weights) {
    (void)height;
    const __m128i t = _mm_loadu_si128((const __m128i *)&weight_array[4]);
    const __m128i zero = _mm_setzero_si128();

    const __m128i weights_0 = _mm_unpacklo_epi8(t, zero);
    const __m128i d = _mm_set1_epi16((uint16_t)(1 << sm_weight_log2_scale));
    const __m128i weights_1 = _mm_sub_epi16(d, weights_0);
    weights[0] = _mm_unpacklo_epi16(weights_0, weights_1);
}

static INLINE void smooth_h_pred_4xh(const __m128i *pixel,
    const __m128i *weight, int32_t h, uint8_t *dst,
    ptrdiff_t stride) {
    const __m128i pred_round = _mm_set1_epi32((1 << (sm_weight_log2_scale - 1)));
    const __m128i one = _mm_set1_epi16(1);
    const __m128i gat = _mm_set1_epi32(0xc080400);
    __m128i rep = _mm_set1_epi16((short)0x8000);

    for (int32_t i = 0; i < h; ++i) {
        __m128i b = _mm_shuffle_epi8(pixel[0], rep);
        b = _mm_unpacklo_epi16(b, pixel[1]);
        __m128i sum = _mm_madd_epi16(b, weight[0]);

        sum = _mm_add_epi32(sum, pred_round);
        sum = _mm_srai_epi32(sum, sm_weight_log2_scale);

        sum = _mm_shuffle_epi8(sum, gat);
        *(uint32_t *)dst = _mm_cvtsi128_si32(sum);
        dst += stride;

        rep = _mm_add_epi16(rep, one);
    }
}

void svt_aom_smooth_h_predictor_4x4_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels[2];
    load_pixel_h_w4(above, left, 4, pixels);

    __m128i weights;
    load_weight_h_w4(sm_weight_arrays, 4, &weights);

    smooth_h_pred_4xh(pixels, &weights, 4, dst, stride);
}

void svt_aom_smooth_h_predictor_4x8_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels[2];
    load_pixel_h_w4(above, left, 8, pixels);

    __m128i weights;
    load_weight_h_w4(sm_weight_arrays, 8, &weights);

    smooth_h_pred_4xh(pixels, &weights, 8, dst, stride);
}

void svt_aom_smooth_h_predictor_4x16_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels[2];
    load_pixel_h_w4(above, left, 16, pixels);

    __m128i weights;
    load_weight_h_w4(sm_weight_arrays, 8, &weights);

    smooth_h_pred_4xh(pixels, &weights, 8, dst, stride);
    dst += stride << 3;

    pixels[0] = _mm_srli_si128(pixels[0], 8);
    smooth_h_pred_4xh(pixels, &weights, 8, dst, stride);
}

// pixels[0]: left vector
// pixels[1]: right_pred vector
// pixels[2]: left vector + 16
// pixels[3]: right_pred vector
static INLINE void load_pixel_h_w8(const uint8_t *above, const uint8_t *left,
    int32_t height, __m128i *pixels) {
    pixels[1] = _mm_set1_epi16((uint16_t)above[7]);

    if (height == 4)
        pixels[0] = _mm_cvtsi32_si128(((const uint32_t *)left)[0]);
    else if (height == 8)
        pixels[0] = _mm_loadl_epi64((const __m128i *)left);
    else if (height == 16)
        pixels[0] = _mm_loadu_si128((const __m128i *)left);
    else {
        pixels[0] = _mm_loadu_si128((const __m128i *)left);
        pixels[2] = _mm_loadu_si128((const __m128i *)(left + 16));
        pixels[3] = pixels[1];
    }
}

// weight_w[0]: weights_w and scale - weights_w interleave vector, first half
// weight_w[1]: weights_w and scale - weights_w interleave vector, second half
static INLINE void load_weight_h_w8(const uint8_t *weight_array, int32_t height,
    __m128i *weight_w) {
    (void)height;
    const __m128i zero = _mm_setzero_si128();
    const __m128i d = _mm_set1_epi16((uint16_t)(1 << sm_weight_log2_scale));
    const __m128i we = _mm_loadu_si128((const __m128i *)&weight_array[8]);
    const __m128i tmp1 = _mm_unpacklo_epi8(we, zero);
    const __m128i tmp2 = _mm_sub_epi16(d, tmp1);
    weight_w[0] = _mm_unpacklo_epi16(tmp1, tmp2);
    weight_w[1] = _mm_unpackhi_epi16(tmp1, tmp2);
}

static INLINE void smooth_h_pred_8xh(const __m128i *pixels, const __m128i *ww,
    int32_t h, uint8_t *dst, ptrdiff_t stride,
    int32_t second_half) {
    const __m128i pred_round = _mm_set1_epi32((1 << (sm_weight_log2_scale - 1)));
    const __m128i one = _mm_set1_epi16(1);
    const __m128i gat = _mm_set_epi32(0, 0, 0xe0c0a08, 0x6040200);
    __m128i rep = second_half ? _mm_set1_epi16((short)0x8008) : _mm_set1_epi16((short)0x8000);

    for (int32_t i = 0; i < h; ++i) {
        __m128i b = _mm_shuffle_epi8(pixels[0], rep);
        b = _mm_unpacklo_epi16(b, pixels[1]);
        __m128i sum0 = _mm_madd_epi16(b, ww[0]);
        __m128i sum1 = _mm_madd_epi16(b, ww[1]);

        sum0 = _mm_add_epi32(sum0, pred_round);
        sum0 = _mm_srai_epi32(sum0, sm_weight_log2_scale);

        sum1 = _mm_add_epi32(sum1, pred_round);
        sum1 = _mm_srai_epi32(sum1, sm_weight_log2_scale);

        sum0 = _mm_packus_epi16(sum0, sum1);
        sum0 = _mm_shuffle_epi8(sum0, gat);
        _mm_storel_epi64((__m128i *)dst, sum0);
        dst += stride;

        rep = _mm_add_epi16(rep, one);
    }
}

void svt_aom_smooth_h_predictor_8x4_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels[2];
    load_pixel_h_w8(above, left, 4, pixels);

    __m128i ww[2];
    load_weight_h_w8(sm_weight_arrays, 4, ww);

    smooth_h_pred_8xh(pixels, ww, 4, dst, stride, 0);
}

void svt_aom_smooth_h_predictor_8x8_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels[2];
    load_pixel_h_w8(above, left, 8, pixels);

    __m128i ww[2];
    load_weight_h_w8(sm_weight_arrays, 8, ww);

    smooth_h_pred_8xh(pixels, ww, 8, dst, stride, 0);
}

void svt_aom_smooth_h_predictor_8x16_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels[2];
    load_pixel_h_w8(above, left, 16, pixels);

    __m128i ww[2];
    load_weight_h_w8(sm_weight_arrays, 16, ww);

    smooth_h_pred_8xh(pixels, ww, 8, dst, stride, 0);
    dst += stride << 3;
    smooth_h_pred_8xh(pixels, ww, 8, dst, stride, 1);
}

void svt_aom_smooth_h_predictor_8x32_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    __m128i pixels[4];
    load_pixel_h_w8(above, left, 32, pixels);

    __m128i ww[2];
    load_weight_h_w8(sm_weight_arrays, 32, ww);

    smooth_h_pred_8xh(&pixels[0], ww, 8, dst, stride, 0);
    dst += stride << 3;
    smooth_h_pred_8xh(&pixels[0], ww, 8, dst, stride, 1);
    dst += stride << 3;
    smooth_h_pred_8xh(&pixels[2], ww, 8, dst, stride, 0);
    dst += stride << 3;
    smooth_h_pred_8xh(&pixels[2], ww, 8, dst, stride, 1);
}

static INLINE void smooth_h_predictor_wxh(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left, uint32_t bw,
    uint32_t bh) {
    const uint8_t *const sm_weights_w = sm_weight_arrays + bw;
    const __m128i zero = _mm_setzero_si128();
    const __m128i scale_value =
        _mm_set1_epi16((uint16_t)(1 << sm_weight_log2_scale));
    const __m128i top_right = _mm_cvtsi32_si128((uint32_t)above[bw - 1]);
    const __m128i gat = _mm_set_epi32(0, 0, 0xe0c0a08, 0x6040200);
    const __m128i pred_round = _mm_set1_epi32((1 << (sm_weight_log2_scale - 1)));

    for (uint32_t y = 0; y < bh; ++y) {
        const __m128i left_y = _mm_cvtsi32_si128((uint32_t)left[y]);
        const __m128i tr_ly =
            _mm_shuffle_epi32(_mm_unpacklo_epi16(top_right, left_y), 0);

        for (uint32_t x = 0; x < bw; x += 8) {
            const __m128i weights_x =
                _mm_loadl_epi64((const __m128i *)(sm_weights_w + x));
            const __m128i weights_xw = _mm_unpacklo_epi8(weights_x, zero);
            const __m128i scale_m_weights_x = _mm_sub_epi16(scale_value, weights_xw);
            const __m128i wx_lo = _mm_unpacklo_epi16(scale_m_weights_x, weights_xw);
            const __m128i wx_hi = _mm_unpackhi_epi16(scale_m_weights_x, weights_xw);
            __m128i pred_lo = _mm_madd_epi16(wx_lo, tr_ly);
            __m128i pred_hi = _mm_madd_epi16(wx_hi, tr_ly);

            pred_lo = _mm_add_epi32(pred_lo, pred_round);
            pred_hi = _mm_add_epi32(pred_hi, pred_round);

            pred_lo = _mm_srai_epi32(pred_lo, sm_weight_log2_scale);
            pred_hi = _mm_srai_epi32(pred_hi, sm_weight_log2_scale);

            __m128i pred = _mm_packus_epi16(pred_lo, pred_hi);
            pred = _mm_shuffle_epi8(pred, gat);
            _mm_storel_epi64((__m128i *)(dst + x), pred);
        }
        dst += stride;
    }
}

void svt_aom_smooth_h_predictor_16x4_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_h_predictor_wxh(dst, stride, above, left, 16, 4);
}

void svt_aom_smooth_h_predictor_16x8_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_h_predictor_wxh(dst, stride, above, left, 16, 8);
}

void svt_aom_smooth_h_predictor_16x16_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_h_predictor_wxh(dst, stride, above, left, 16, 16);
}

void svt_aom_smooth_h_predictor_16x32_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_h_predictor_wxh(dst, stride, above, left, 16, 32);
}

void svt_aom_smooth_h_predictor_16x64_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_h_predictor_wxh(dst, stride, above, left, 16, 64);
}

void svt_aom_smooth_h_predictor_32x8_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_h_predictor_wxh(dst, stride, above, left, 32, 8);
}

void svt_aom_smooth_h_predictor_32x16_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_h_predictor_wxh(dst, stride, above, left, 32, 16);
}

void svt_aom_smooth_h_predictor_32x32_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_h_predictor_wxh(dst, stride, above, left, 32, 32);
}

void svt_aom_smooth_h_predictor_32x64_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_h_predictor_wxh(dst, stride, above, left, 32, 64);
}

void svt_aom_smooth_h_predictor_64x64_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_h_predictor_wxh(dst, stride, above, left, 64, 64);
}

void svt_aom_smooth_h_predictor_64x32_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_h_predictor_wxh(dst, stride, above, left, 64, 32);
}

void svt_aom_smooth_h_predictor_64x16_ssse3(uint8_t *dst, ptrdiff_t stride,
    const uint8_t *above,
    const uint8_t *left) {
    smooth_h_predictor_wxh(dst, stride, above, left, 64, 16);
}