vp8[loongarch]: Optimize vp8_loop/sixtap, vpx_dc with LSX.

1. vp8_loop_filter_mbh, vp8_loop_filter_mbv
2. vp8_sixtap_predict16x16, vp8_sixtap_predict8x8
3. vpx_dc_predictor_16x16, vpx_dc_predictor_8x8

./vpxdec --progress -o YUV_1920X1080.yuv original_1200f/VP8_1920X1080.webm

before: 37.77fps
after : 220.90fps

Bug: webm:1755

Change-Id: I1a3ce16f0c872261d813b6531cfdf25bd59bb774

4 files changed, 1565 insertions, 4 deletions

diff --git a/vp8/common/loongarch/loopfilter_filters_lsx.c b/vp8/common/loongarch/loopfilter_filters_lsx.c
new file mode 100644
index 000000000..484b3d6ad
--- /dev/null
+++ b/vp8/common/loongarch/loopfilter_filters_lsx.c
@@ -0,0 +1,393 @@
+/*
+ * Copyright (c) 2021 Loongson Technology Corporation Limited
+ * Contributed by Lu Wang <wanglu@loongson.cn>
+ *
+ * Use of this source code is governed by a BSD-style license
+ * that can be found in the LICENSE file in the root of the source
+ * tree. An additional intellectual property rights grant can be found
+ * in the file PATENTS. All contributing project authors may
+ * be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "./vp8_rtcd.h"
+#include "vp8/common/loopfilter.h"
+#include "vpx_util/loongson_intrinsics.h"
+
+#define VP8_MBFILTER(p2, p1, p0, q0, q1, q2, mask, hev) \
+  {                                                     \
+    __m128i p2_m, p1_m, p0_m, q2_m, q1_m, q0_m;         \
+    __m128i u, filt, t1, t2, filt_sign, q0_sub_p0;      \
+    __m128i filt_r, filt_l;                             \
+    __m128i temp0, temp1, temp2, temp3;                 \
+    const __m128i cnst4b = __lsx_vldi(4);               \
+    const __m128i cnst3b = __lsx_vldi(3);               \
+    const __m128i cnst9h = __lsx_vldi(1033);            \
+    const __m128i cnst63h = __lsx_vldi(1087);           \
+                                                        \
+    p2_m = __lsx_vxori_b(p2, 0x80);                     \
+    p1_m = __lsx_vxori_b(p1, 0x80);                     \
+    p0_m = __lsx_vxori_b(p0, 0x80);                     \
+    q0_m = __lsx_vxori_b(q0, 0x80);                     \
+    q1_m = __lsx_vxori_b(q1, 0x80);                     \
+    q2_m = __lsx_vxori_b(q2, 0x80);                     \
+                                                        \
+    filt = __lsx_vssub_b(p1_m, q1_m);                   \
+    q0_sub_p0 = __lsx_vssub_b(q0_m, p0_m);              \
+    filt = __lsx_vsadd_b(filt, q0_sub_p0);              \
+    filt = __lsx_vsadd_b(filt, q0_sub_p0);              \
+    filt = __lsx_vsadd_b(filt, q0_sub_p0);              \
+    filt = __lsx_vand_v(filt, mask);                    \
+                                                        \
+    t2 = __lsx_vand_v(filt, hev);                       \
+    hev = __lsx_vxori_b(hev, 0xff);                     \
+    filt = __lsx_vand_v(hev, filt);                     \
+    t1 = __lsx_vsadd_b(t2, cnst4b);                     \
+    t1 = __lsx_vsra_b(t1, cnst3b);                      \
+    t2 = __lsx_vsadd_b(t2, cnst3b);                     \
+    t2 = __lsx_vsra_b(t2, cnst3b);                      \
+    q0_m = __lsx_vssub_b(q0_m, t1);                     \
+    p0_m = __lsx_vsadd_b(p0_m, t2);                     \
+    filt_sign = __lsx_vslti_b(filt, 0);                 \
+    filt_r = __lsx_vilvl_b(filt_sign, filt);            \
+    filt_l = __lsx_vilvh_b(filt_sign, filt);            \
+    temp0 = __lsx_vmul_h(filt_r, cnst9h);               \
+    temp1 = __lsx_vadd_h(temp0, cnst63h);               \
+    temp2 = __lsx_vmul_h(filt_l, cnst9h);               \
+    temp3 = __lsx_vadd_h(temp2, cnst63h);               \
+                                                        \
+    u = __lsx_vssrani_b_h(temp3, temp1, 7);             \
+    q2_m = __lsx_vssub_b(q2_m, u);                      \
+    p2_m = __lsx_vsadd_b(p2_m, u);                      \
+    q2 = __lsx_vxori_b(q2_m, 0x80);                     \
+    p2 = __lsx_vxori_b(p2_m, 0x80);                     \
+                                                        \
+    temp1 = __lsx_vadd_h(temp1, temp0);                 \
+    temp3 = __lsx_vadd_h(temp3, temp2);                 \
+                                                        \
+    u = __lsx_vssrani_b_h(temp3, temp1, 7);             \
+    q1_m = __lsx_vssub_b(q1_m, u);                      \
+    p1_m = __lsx_vsadd_b(p1_m, u);                      \
+    q1 = __lsx_vxori_b(q1_m, 0x80);                     \
+    p1 = __lsx_vxori_b(p1_m, 0x80);                     \
+                                                        \
+    temp1 = __lsx_vadd_h(temp1, temp0);                 \
+    temp3 = __lsx_vadd_h(temp3, temp2);                 \
+                                                        \
+    u = __lsx_vssrani_b_h(temp3, temp1, 7);             \
+    q0_m = __lsx_vssub_b(q0_m, u);                      \
+    p0_m = __lsx_vsadd_b(p0_m, u);                      \
+    q0 = __lsx_vxori_b(q0_m, 0x80);                     \
+    p0 = __lsx_vxori_b(p0_m, 0x80);                     \
+  }
+
+#define LPF_MASK_HEV(p3_in, p2_in, p1_in, p0_in, q0_in, q1_in, q2_in, q3_in, \
+                     limit_in, b_limit_in, thresh_in, hev_out, mask_out,     \
+                     flat_out)                                               \
+  {                                                                          \
+    __m128i p3_asub_p2_m, p2_asub_p1_m, p1_asub_p0_m, q1_asub_q0_m;          \
+    __m128i p1_asub_q1_m, p0_asub_q0_m, q3_asub_q2_m, q2_asub_q1_m;          \
+                                                                             \
+    p3_asub_p2_m = __lsx_vabsd_bu(p3_in, p2_in);                             \
+    p2_asub_p1_m = __lsx_vabsd_bu(p2_in, p1_in);                             \
+    p1_asub_p0_m = __lsx_vabsd_bu(p1_in, p0_in);                             \
+    q1_asub_q0_m = __lsx_vabsd_bu(q1_in, q0_in);                             \
+    q2_asub_q1_m = __lsx_vabsd_bu(q2_in, q1_in);                             \
+    q3_asub_q2_m = __lsx_vabsd_bu(q3_in, q2_in);                             \
+    p0_asub_q0_m = __lsx_vabsd_bu(p0_in, q0_in);                             \
+    p1_asub_q1_m = __lsx_vabsd_bu(p1_in, q1_in);                             \
+    flat_out = __lsx_vmax_bu(p1_asub_p0_m, q1_asub_q0_m);                    \
+    hev_out = __lsx_vslt_bu(thresh_in, flat_out);                            \
+    p0_asub_q0_m = __lsx_vsadd_bu(p0_asub_q0_m, p0_asub_q0_m);               \
+    p1_asub_q1_m = __lsx_vsrli_b(p1_asub_q1_m, 1);                           \
+    p0_asub_q0_m = __lsx_vsadd_bu(p0_asub_q0_m, p1_asub_q1_m);               \
+    mask_out = __lsx_vslt_bu(b_limit_in, p0_asub_q0_m);                      \
+    mask_out = __lsx_vmax_bu(flat_out, mask_out);                            \
+    p3_asub_p2_m = __lsx_vmax_bu(p3_asub_p2_m, p2_asub_p1_m);                \
+    mask_out = __lsx_vmax_bu(p3_asub_p2_m, mask_out);                        \
+    q2_asub_q1_m = __lsx_vmax_bu(q2_asub_q1_m, q3_asub_q2_m);                \
+    mask_out = __lsx_vmax_bu(q2_asub_q1_m, mask_out);                        \
+    mask_out = __lsx_vslt_bu(limit_in, mask_out);                            \
+    mask_out = __lsx_vxori_b(mask_out, 0xff);                                \
+  }
+
+#define VP8_ST6x1_B(in0, in0_idx, in1, in1_idx, pdst, stride) \
+  {                                                           \
+    __lsx_vstelm_w(in0, pdst, 0, in0_idx);                    \
+    __lsx_vstelm_h(in1, pdst + stride, 0, in1_idx);           \
+  }
+
+static inline void mbloop_filter_horizontal_edge_y_lsx(
+    uint8_t *src, int32_t pitch, const uint8_t b_limit_in,
+    const uint8_t limit_in, const uint8_t thresh_in) {
+  uint8_t *temp_src;
+  int32_t pitch_x2 = pitch << 1;
+  int32_t pitch_x3 = pitch_x2 + pitch;
+  int32_t pitch_x4 = pitch << 2;
+
+  __m128i p3, p2, p1, p0, q3, q2, q1, q0;
+  __m128i mask, hev, flat, thresh, limit, b_limit;
+
+  DUP2_ARG2(__lsx_vldrepl_b, &b_limit_in, 0, &limit_in, 0, b_limit, limit);
+  thresh = __lsx_vldrepl_b(&thresh_in, 0);
+
+  temp_src = src - pitch_x4;
+  DUP4_ARG2(__lsx_vldx, temp_src, 0, temp_src, pitch, temp_src, pitch_x2,
+            temp_src, pitch_x3, p3, p2, p1, p0);
+  temp_src += pitch_x4;
+  DUP4_ARG2(__lsx_vldx, temp_src, 0, temp_src, pitch, temp_src, pitch_x2,
+            temp_src, pitch_x3, q0, q1, q2, q3);
+  LPF_MASK_HEV(p3, p2, p1, p0, q0, q1, q2, q3, limit, b_limit, thresh, hev,
+               mask, flat);
+  VP8_MBFILTER(p2, p1, p0, q0, q1, q2, mask, hev);
+
+  temp_src = src - pitch_x3;
+  __lsx_vstx(p2, temp_src, 0);
+  __lsx_vstx(p1, temp_src, pitch);
+  __lsx_vstx(p0, temp_src, pitch_x2);
+  __lsx_vstx(q0, temp_src, pitch_x3);
+  temp_src += pitch_x4;
+  __lsx_vstx(q1, temp_src, 0);
+  __lsx_vstx(q2, temp_src, pitch);
+}
+
+static inline void mbloop_filter_horizontal_edge_uv_lsx(
+    uint8_t *src_u, uint8_t *src_v, int32_t pitch, const uint8_t b_limit_in,
+    const uint8_t limit_in, const uint8_t thresh_in) {
+  uint8_t *temp_src;
+  int32_t pitch_x2 = pitch << 1;
+  int32_t pitch_x3 = pitch_x2 + pitch;
+  int32_t pitch_x4 = pitch << 2;
+  __m128i p3, p2, p1, p0, q3, q2, q1, q0;
+  __m128i mask, hev, flat, thresh, limit, b_limit;
+  __m128i p3_u, p2_u, p1_u, p0_u, q3_u, q2_u, q1_u, q0_u;
+  __m128i p3_v, p2_v, p1_v, p0_v, q3_v, q2_v, q1_v, q0_v;
+
+  DUP2_ARG2(__lsx_vldrepl_b, &b_limit_in, 0, &limit_in, 0, b_limit, limit);
+  thresh = __lsx_vldrepl_b(&thresh_in, 0);
+
+  temp_src = src_u - pitch_x4;
+  DUP4_ARG2(__lsx_vldx, temp_src, 0, temp_src, pitch, temp_src, pitch_x2,
+            temp_src, pitch_x3, p3_u, p2_u, p1_u, p0_u);
+  temp_src += pitch_x4;
+  DUP4_ARG2(__lsx_vldx, temp_src, 0, temp_src, pitch, temp_src, pitch_x2,
+            temp_src, pitch_x3, q0_u, q1_u, q2_u, q3_u);
+  temp_src = src_v - pitch_x4;
+  DUP4_ARG2(__lsx_vldx, temp_src, 0, temp_src, pitch, temp_src, pitch_x2,
+            temp_src, pitch_x3, p3_v, p2_v, p1_v, p0_v);
+  temp_src += pitch_x4;
+  DUP4_ARG2(__lsx_vldx, temp_src, 0, temp_src, pitch, temp_src, pitch_x2,
+            temp_src, pitch_x3, q0_v, q1_v, q2_v, q3_v);
+
+  DUP4_ARG2(__lsx_vilvl_d, p3_v, p3_u, p2_v, p2_u, p1_v, p1_u, p0_v, p0_u, p3,
+            p2, p1, p0);
+  DUP4_ARG2(__lsx_vilvl_d, q0_v, q0_u, q1_v, q1_u, q2_v, q2_u, q3_v, q3_u, q0,
+            q1, q2, q3);
+  LPF_MASK_HEV(p3, p2, p1, p0, q0, q1, q2, q3, limit, b_limit, thresh, hev,
+               mask, flat);
+  VP8_MBFILTER(p2, p1, p0, q0, q1, q2, mask, hev);
+
+  src_u -= pitch_x3;
+  __lsx_vstelm_d(p2, src_u, 0, 0);
+  __lsx_vstelm_d(p1, src_u + pitch, 0, 0);
+  __lsx_vstelm_d(p0, src_u + pitch_x2, 0, 0);
+  __lsx_vstelm_d(q0, src_u + pitch_x3, 0, 0);
+  src_u += pitch_x4;
+  __lsx_vstelm_d(q1, src_u, 0, 0);
+  src_u += pitch;
+  __lsx_vstelm_d(q2, src_u, 0, 0);
+
+  src_v -= pitch_x3;
+  __lsx_vstelm_d(p2, src_v, 0, 1);
+  __lsx_vstelm_d(p1, src_v + pitch, 0, 1);
+  __lsx_vstelm_d(p0, src_v + pitch_x2, 0, 1);
+  __lsx_vstelm_d(q0, src_v + pitch_x3, 0, 1);
+  src_v += pitch_x4;
+  __lsx_vstelm_d(q1, src_v, 0, 1);
+  src_v += pitch;
+  __lsx_vstelm_d(q2, src_v, 0, 1);
+}
+
+static inline void mbloop_filter_vertical_edge_y_lsx(uint8_t *src,
+                                                     int32_t pitch,
+                                                     const uint8_t b_limit_in,
+                                                     const uint8_t limit_in,
+                                                     const uint8_t thresh_in) {
+  uint8_t *temp_src;
+  int32_t pitch_x2 = pitch << 1;
+  int32_t pitch_x3 = pitch_x2 + pitch;
+  int32_t pitch_x4 = pitch << 2;
+
+  __m128i p3, p2, p1, p0, q3, q2, q1, q0;
+  __m128i mask, hev, flat, thresh, limit, b_limit;
+  __m128i row0, row1, row2, row3, row4, row5, row6, row7, row8;
+  __m128i row9, row10, row11, row12, row13, row14, row15;
+  __m128i tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+
+  DUP2_ARG2(__lsx_vldrepl_b, &b_limit_in, 0, &limit_in, 0, b_limit, limit);
+  thresh = __lsx_vldrepl_b(&thresh_in, 0);
+  temp_src = src - 4;
+  DUP4_ARG2(__lsx_vldx, temp_src, 0, temp_src, pitch, temp_src, pitch_x2,
+            temp_src, pitch_x3, row0, row1, row2, row3);
+  temp_src += pitch_x4;
+  DUP4_ARG2(__lsx_vldx, temp_src, 0, temp_src, pitch, temp_src, pitch_x2,
+            temp_src, pitch_x3, row4, row5, row6, row7);
+  temp_src += pitch_x4;
+  DUP4_ARG2(__lsx_vldx, temp_src, 0, temp_src, pitch, temp_src, pitch_x2,
+            temp_src, pitch_x3, row8, row9, row10, row11);
+  temp_src += pitch_x4;
+  DUP4_ARG2(__lsx_vldx, temp_src, 0, temp_src, pitch, temp_src, pitch_x2,
+            temp_src, pitch_x3, row12, row13, row14, row15);
+  temp_src -= pitch_x4;
+  LSX_TRANSPOSE16x8_B(row0, row1, row2, row3, row4, row5, row6, row7, row8,
+                      row9, row10, row11, row12, row13, row14, row15, p3, p2,
+                      p1, p0, q0, q1, q2, q3);
+
+  LPF_MASK_HEV(p3, p2, p1, p0, q0, q1, q2, q3, limit, b_limit, thresh, hev,
+               mask, flat);
+  VP8_MBFILTER(p2, p1, p0, q0, q1, q2, mask, hev);
+  DUP2_ARG2(__lsx_vilvl_b, p1, p2, q0, p0, tmp0, tmp1);
+  tmp3 = __lsx_vilvl_h(tmp1, tmp0);
+  tmp4 = __lsx_vilvh_h(tmp1, tmp0);
+  DUP2_ARG2(__lsx_vilvh_b, p1, p2, q0, p0, tmp0, tmp1);
+  tmp6 = __lsx_vilvl_h(tmp1, tmp0);
+  tmp7 = __lsx_vilvh_h(tmp1, tmp0);
+  tmp2 = __lsx_vilvl_b(q2, q1);
+  tmp5 = __lsx_vilvh_b(q2, q1);
+
+  temp_src = src - 3;
+  VP8_ST6x1_B(tmp3, 0, tmp2, 0, temp_src, 4);
+  temp_src += pitch;
+  VP8_ST6x1_B(tmp3, 1, tmp2, 1, temp_src, 4);
+  temp_src += pitch;
+  VP8_ST6x1_B(tmp3, 2, tmp2, 2, temp_src, 4);
+  temp_src += pitch;
+  VP8_ST6x1_B(tmp3, 3, tmp2, 3, temp_src, 4);
+  temp_src += pitch;
+  VP8_ST6x1_B(tmp4, 0, tmp2, 4, temp_src, 4);
+  temp_src += pitch;
+  VP8_ST6x1_B(tmp4, 1, tmp2, 5, temp_src, 4);
+  temp_src += pitch;
+  VP8_ST6x1_B(tmp4, 2, tmp2, 6, temp_src, 4);
+  temp_src += pitch;
+  VP8_ST6x1_B(tmp4, 3, tmp2, 7, temp_src, 4);
+  temp_src += pitch;
+  VP8_ST6x1_B(tmp6, 0, tmp5, 0, temp_src, 4);
+  temp_src += pitch;
+  VP8_ST6x1_B(tmp6, 1, tmp5, 1, temp_src, 4);
+  temp_src += pitch;
+  VP8_ST6x1_B(tmp6, 2, tmp5, 2, temp_src, 4);
+  temp_src += pitch;
+  VP8_ST6x1_B(tmp6, 3, tmp5, 3, temp_src, 4);
+  temp_src += pitch;
+  VP8_ST6x1_B(tmp7, 0, tmp5, 4, temp_src, 4);
+  temp_src += pitch;
+  VP8_ST6x1_B(tmp7, 1, tmp5, 5, temp_src, 4);
+  temp_src += pitch;
+  VP8_ST6x1_B(tmp7, 2, tmp5, 6, temp_src, 4);
+  temp_src += pitch;
+  VP8_ST6x1_B(tmp7, 3, tmp5, 7, temp_src, 4);
+}
+
+static inline void mbloop_filter_vertical_edge_uv_lsx(
+    uint8_t *src_u, uint8_t *src_v, int32_t pitch, const uint8_t b_limit_in,
+    const uint8_t limit_in, const uint8_t thresh_in) {
+  int32_t pitch_x2 = pitch << 1;
+  int32_t pitch_x3 = pitch_x2 + pitch;
+  int32_t pitch_x4 = pitch << 2;
+  __m128i p3, p2, p1, p0, q3, q2, q1, q0;
+  __m128i mask, hev, flat, thresh, limit, b_limit;
+  __m128i row0, row1, row2, row3, row4, row5, row6, row7, row8;
+  __m128i row9, row10, row11, row12, row13, row14, row15;
+  __m128i tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+
+  DUP2_ARG2(__lsx_vldrepl_b, &b_limit_in, 0, &limit_in, 0, b_limit, limit);
+  thresh = __lsx_vldrepl_b(&thresh_in, 0);
+
+  src_u -= 4;
+  DUP4_ARG2(__lsx_vldx, src_u, 0, src_u, pitch, src_u, pitch_x2, src_u,
+            pitch_x3, row0, row1, row2, row3);
+  src_u += pitch_x4;
+  DUP4_ARG2(__lsx_vldx, src_u, 0, src_u, pitch, src_u, pitch_x2, src_u,
+            pitch_x3, row4, row5, row6, row7);
+  src_v -= 4;
+  DUP4_ARG2(__lsx_vldx, src_v, 0, src_v, pitch, src_v, pitch_x2, src_v,
+            pitch_x3, row8, row9, row10, row11);
+  src_v += pitch_x4;
+  DUP4_ARG2(__lsx_vldx, src_v, 0, src_v, pitch, src_v, pitch_x2, src_v,
+            pitch_x3, row12, row13, row14, row15);
+  LSX_TRANSPOSE16x8_B(row0, row1, row2, row3, row4, row5, row6, row7, row8,
+                      row9, row10, row11, row12, row13, row14, row15, p3, p2,
+                      p1, p0, q0, q1, q2, q3);
+
+  LPF_MASK_HEV(p3, p2, p1, p0, q0, q1, q2, q3, limit, b_limit, thresh, hev,
+               mask, flat);
+  VP8_MBFILTER(p2, p1, p0, q0, q1, q2, mask, hev);
+
+  DUP2_ARG2(__lsx_vilvl_b, p1, p2, q0, p0, tmp0, tmp1);
+  tmp3 = __lsx_vilvl_h(tmp1, tmp0);
+  tmp4 = __lsx_vilvh_h(tmp1, tmp0);
+  DUP2_ARG2(__lsx_vilvh_b, p1, p2, q0, p0, tmp0, tmp1);
+  tmp6 = __lsx_vilvl_h(tmp1, tmp0);
+  tmp7 = __lsx_vilvh_h(tmp1, tmp0);
+  tmp2 = __lsx_vilvl_b(q2, q1);
+  tmp5 = __lsx_vilvh_b(q2, q1);
+
+  src_u += 1 - pitch_x4;
+  VP8_ST6x1_B(tmp3, 0, tmp2, 0, src_u, 4);
+  src_u += pitch;
+  VP8_ST6x1_B(tmp3, 1, tmp2, 1, src_u, 4);
+  src_u += pitch;
+  VP8_ST6x1_B(tmp3, 2, tmp2, 2, src_u, 4);
+  src_u += pitch;
+  VP8_ST6x1_B(tmp3, 3, tmp2, 3, src_u, 4);
+  src_u += pitch;
+  VP8_ST6x1_B(tmp4, 0, tmp2, 4, src_u, 4);
+  src_u += pitch;
+  VP8_ST6x1_B(tmp4, 1, tmp2, 5, src_u, 4);
+  src_u += pitch;
+  VP8_ST6x1_B(tmp4, 2, tmp2, 6, src_u, 4);
+  src_u += pitch;
+  VP8_ST6x1_B(tmp4, 3, tmp2, 7, src_u, 4);
+
+  src_v += 1 - pitch_x4;
+  VP8_ST6x1_B(tmp6, 0, tmp5, 0, src_v, 4);
+  src_v += pitch;
+  VP8_ST6x1_B(tmp6, 1, tmp5, 1, src_v, 4);
+  src_v += pitch;
+  VP8_ST6x1_B(tmp6, 2, tmp5, 2, src_v, 4);
+  src_v += pitch;
+  VP8_ST6x1_B(tmp6, 3, tmp5, 3, src_v, 4);
+  src_v += pitch;
+  VP8_ST6x1_B(tmp7, 0, tmp5, 4, src_v, 4);
+  src_v += pitch;
+  VP8_ST6x1_B(tmp7, 1, tmp5, 5, src_v, 4);
+  src_v += pitch;
+  VP8_ST6x1_B(tmp7, 2, tmp5, 6, src_v, 4);
+  src_v += pitch;
+  VP8_ST6x1_B(tmp7, 3, tmp5, 7, src_v, 4);
+}
+
+void vp8_loop_filter_mbh_lsx(uint8_t *src_y, uint8_t *src_u, uint8_t *src_v,
+                             int32_t pitch_y, int32_t pitch_u_v,
+                             loop_filter_info *lpf_info_ptr) {
+  mbloop_filter_horizontal_edge_y_lsx(src_y, pitch_y, *lpf_info_ptr->mblim,
+                                      *lpf_info_ptr->lim,
+                                      *lpf_info_ptr->hev_thr);
+  if (src_u) {
+    mbloop_filter_horizontal_edge_uv_lsx(
+        src_u, src_v, pitch_u_v, *lpf_info_ptr->mblim, *lpf_info_ptr->lim,
+        *lpf_info_ptr->hev_thr);
+  }
+}
+
+void vp8_loop_filter_mbv_lsx(uint8_t *src_y, uint8_t *src_u, uint8_t *src_v,
+                             int32_t pitch_y, int32_t pitch_u_v,
+                             loop_filter_info *lpf_info_ptr) {
+  mbloop_filter_vertical_edge_y_lsx(src_y, pitch_y, *lpf_info_ptr->mblim,
+                                    *lpf_info_ptr->lim, *lpf_info_ptr->hev_thr);
+  if (src_u) {
+    mbloop_filter_vertical_edge_uv_lsx(src_u, src_v, pitch_u_v,
+                                       *lpf_info_ptr->mblim, *lpf_info_ptr->lim,
+                                       *lpf_info_ptr->hev_thr);
+  }
+}
diff --git a/vp8/common/loongarch/sixtap_filter_lsx.c b/vp8/common/loongarch/sixtap_filter_lsx.c
new file mode 100644
index 000000000..75fe533d9
--- /dev/null
+++ b/vp8/common/loongarch/sixtap_filter_lsx.c
@@ -0,0 +1,1164 @@
+/*
+ * Copyright (c) 2021 Loongson Technology Corporation Limited
+ * Contributed by Lu Wang <wanglu@loongson.cn>
+ *
+ * Use of this source code is governed by a BSD-style license
+ * that can be found in the LICENSE file in the root of the source
+ * tree. An additional intellectual property rights grant can be found
+ * in the file PATENTS.  All contributing project authors may
+ * be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "./vp8_rtcd.h"
+#include "vp8/common/filter.h"
+#include "vpx_ports/mem.h"
+#include "vpx_util/loongson_intrinsics.h"
+
+DECLARE_ALIGNED(16, static const int8_t, vp8_subpel_filters_lsx[7][8]) = {
+  { 0, -6, 123, 12, -1, 0, 0, 0 },
+  { 2, -11, 108, 36, -8, 1, 0, 0 }, /* New 1/4 pel 6 tap filter */
+  { 0, -9, 93, 50, -6, 0, 0, 0 },
+  { 3, -16, 77, 77, -16, 3, 0, 0 }, /* New 1/2 pel 6 tap filter */
+  { 0, -6, 50, 93, -9, 0, 0, 0 },
+  { 1, -8, 36, 108, -11, 2, 0, 0 }, /* New 1/4 pel 6 tap filter */
+  { 0, -1, 12, 123, -6, 0, 0, 0 },
+};
+
+static const uint8_t vp8_mc_filt_mask_arr[16 * 3] = {
+  /* 8 width cases */
+  0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8,
+  /* 4 width cases */
+  0, 1, 1, 2, 2, 3, 3, 4, 16, 17, 17, 18, 18, 19, 19, 20,
+  /* 4 width cases */
+  8, 9, 9, 10, 10, 11, 11, 12, 24, 25, 25, 26, 26, 27, 27, 28
+};
+
+#define DPADD_H3(in0, in1, in2, coeff0, coeff1, coeff2) \
+  ({                                                    \
+    __m128i out0_m;                                     \
+                                                        \
+    out0_m = __lsx_vdp2_h_b(in0, coeff0);               \
+    out0_m = __lsx_vdp2add_h_b(out0_m, in1, coeff1);    \
+    out0_m = __lsx_vdp2add_h_b(out0_m, in2, coeff2);    \
+                                                        \
+    out0_m;                                             \
+  })
+
+#define HORIZ_6TAP_FILT(src0, src1, mask0, mask1, mask2, filt_h0, filt_h1,  \
+                        filt_h2)                                            \
+  ({                                                                        \
+    __m128i vec0_m, vec1_m, vec2_m;                                         \
+    __m128i hz_out_m;                                                       \
+                                                                            \
+    DUP2_ARG3(__lsx_vshuf_b, src0, src1, mask0, src0, src1, mask1, vec0_m,  \
+              vec1_m);                                                      \
+    vec2_m = __lsx_vshuf_b(src0, src1, mask2);                              \
+    hz_out_m = DPADD_H3(vec0_m, vec1_m, vec2_m, filt_h0, filt_h1, filt_h2); \
+                                                                            \
+    hz_out_m = __lsx_vsrari_h(hz_out_m, VP8_FILTER_SHIFT);                  \
+    hz_out_m = __lsx_vsat_h(hz_out_m, 7);                                   \
+                                                                            \
+    hz_out_m;                                                               \
+  })
+
+#define HORIZ_6TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1,    \
+                                   mask2, filt0, filt1, filt2, out0, out1,  \
+                                   out2, out3)                              \
+  ({                                                                        \
+    __m128i vec0_m, vec1_m, vec2_m, vec3_m, vec4_m, vec5_m, vec6_m, vec7_m; \
+                                                                            \
+    DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask0, src1, src1, mask0, vec0_m,  \
+              vec1_m);                                                      \
+    DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask0, src3, src3, mask0, vec2_m,  \
+              vec3_m);                                                      \
+    DUP4_ARG2(__lsx_vdp2_h_b, vec0_m, filt0, vec1_m, filt0, vec2_m, filt0,  \
+              vec3_m, filt0, out0, out1, out2, out3);                       \
+    DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask1, src1, src1, mask1, vec0_m,  \
+              vec1_m);                                                      \
+    DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask1, src3, src3, mask1, vec2_m,  \
+              vec3_m);                                                      \
+    DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask2, src1, src1, mask2, vec4_m,  \
+              vec5_m);                                                      \
+    DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask2, src3, src3, mask2, vec6_m,  \
+              vec7_m);                                                      \
+    DUP4_ARG3(__lsx_vdp2add_h_b, out0, vec0_m, filt1, out1, vec1_m, filt1,  \
+              out2, vec2_m, filt1, out3, vec3_m, filt1, out0, out1, out2,   \
+              out3);                                                        \
+    DUP4_ARG3(__lsx_vdp2add_h_b, out0, vec4_m, filt2, out1, vec5_m, filt2,  \
+              out2, vec6_m, filt2, out3, vec7_m, filt2, out0, out1, out2,   \
+              out3);                                                        \
+  })
+
+#define FILT_4TAP_DPADD_H(vec0, vec1, filt0, filt1) \
+  ({                                                \
+    __m128i tmp0;                                   \
+                                                    \
+    tmp0 = __lsx_vdp2_h_b(vec0, filt0);             \
+    tmp0 = __lsx_vdp2add_h_b(tmp0, vec1, filt1);    \
+                                                    \
+    tmp0;                                           \
+  })
+
+#define HORIZ_4TAP_FILT(src0, src1, mask0, mask1, filt_h0, filt_h1)        \
+  ({                                                                       \
+    __m128i vec0_m, vec1_m;                                                \
+    __m128i hz_out_m;                                                      \
+                                                                           \
+    DUP2_ARG3(__lsx_vshuf_b, src0, src1, mask0, src0, src1, mask1, vec0_m, \
+              vec1_m);                                                     \
+    hz_out_m = FILT_4TAP_DPADD_H(vec0_m, vec1_m, filt_h0, filt_h1);        \
+    hz_out_m = __lsx_vsrari_h(hz_out_m, VP8_FILTER_SHIFT);                 \
+    hz_out_m = __lsx_vsat_h(hz_out_m, 7);                                  \
+                                                                           \
+    hz_out_m;                                                              \
+  })
+
+#define HORIZ_4TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1,   \
+                                   filt0, filt1, out0, out1, out2, out3)   \
+  ({                                                                       \
+    __m128i vec0_m, vec1_m, vec2_m, vec3_m;                                \
+                                                                           \
+    DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask0, src1, src1, mask0, vec0_m, \
+              vec1_m);                                                     \
+    DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask0, src3, src3, mask0, vec2_m, \
+              vec3_m);                                                     \
+    DUP4_ARG2(__lsx_vdp2_h_b, vec0_m, filt0, vec1_m, filt0, vec2_m, filt0, \
+              vec3_m, filt0, out0, out1, out2, out3);                      \
+    DUP2_ARG3(__lsx_vshuf_b, src0, src0, mask1, src1, src1, mask1, vec0_m, \
+              vec1_m);                                                     \
+    DUP2_ARG3(__lsx_vshuf_b, src2, src2, mask1, src3, src3, mask1, vec2_m, \
+              vec3_m);                                                     \
+    DUP4_ARG3(__lsx_vdp2add_h_b, out0, vec0_m, filt1, out1, vec1_m, filt1, \
+              out2, vec2_m, filt1, out3, vec3_m, filt1, out0, out1, out2,  \
+              out3);                                                       \
+  })
+
+static void common_hz_6t_8w_lsx(uint8_t *RESTRICT src, int32_t src_stride,
+                                uint8_t *RESTRICT dst, int32_t dst_stride,
+                                const int8_t *filter, int32_t height) {
+  uint32_t loop_cnt;
+  int32_t src_stride_x2 = src_stride << 1;
+  int32_t src_stride_x3 = src_stride_x2 + src_stride;
+  int32_t src_stride_x4 = src_stride << 2;
+  int32_t dst_stride_x2 = dst_stride << 1;
+  int32_t dst_stride_x3 = dst_stride_x2 + dst_stride;
+  int32_t dst_stride_x4 = dst_stride << 2;
+  __m128i src0, src1, src2, src3, filt0, filt1, filt2;
+  __m128i mask0, mask1, mask2, tmp0, tmp1;
+  __m128i filt, out0, out1, out2, out3;
+
+  mask0 = __lsx_vld(vp8_mc_filt_mask_arr, 0);
+  src -= 2;
+
+  filt = __lsx_vld(filter, 0);
+  DUP2_ARG2(__lsx_vreplvei_h, filt, 0, filt, 1, filt0, filt1);
+  filt2 = __lsx_vreplvei_h(filt, 2);
+  DUP2_ARG2(__lsx_vaddi_bu, mask0, 2, mask0, 4, mask1, mask2);
+
+  DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+            src_stride_x3, src0, src1, src2, src3);
+  DUP4_ARG2(__lsx_vxori_b, src0, 128, src1, 128, src2, 128, src3, 128, src0,
+            src1, src2, src3);
+  src += src_stride_x4;
+  HORIZ_6TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2, filt0,
+                             filt1, filt2, out0, out1, out2, out3);
+  DUP2_ARG3(__lsx_vssrarni_b_h, out1, out0, VP8_FILTER_SHIFT, out3, out2,
+            VP8_FILTER_SHIFT, tmp0, tmp1);
+  DUP2_ARG2(__lsx_vxori_b, tmp0, 128, tmp1, 128, tmp0, tmp1);
+  __lsx_vstelm_d(tmp0, dst, 0, 0);
+  __lsx_vstelm_d(tmp0, dst + dst_stride, 0, 1);
+  __lsx_vstelm_d(tmp1, dst + dst_stride_x2, 0, 0);
+  __lsx_vstelm_d(tmp1, dst + dst_stride_x3, 0, 1);
+  dst += dst_stride_x4;
+
+  for (loop_cnt = (height >> 2) - 1; loop_cnt--;) {
+    DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+              src_stride_x3, src0, src1, src2, src3);
+    DUP4_ARG2(__lsx_vxori_b, src0, 128, src1, 128, src2, 128, src3, 128, src0,
+              src1, src2, src3);
+    src += src_stride_x4;
+    HORIZ_6TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
+                               filt0, filt1, filt2, out0, out1, out2, out3);
+    DUP2_ARG3(__lsx_vssrarni_b_h, out1, out0, VP8_FILTER_SHIFT, out3, out2,
+              VP8_FILTER_SHIFT, tmp0, tmp1);
+    DUP2_ARG2(__lsx_vxori_b, tmp0, 128, tmp1, 128, tmp0, tmp1);
+    __lsx_vstelm_d(tmp0, dst, 0, 0);
+    __lsx_vstelm_d(tmp0, dst + dst_stride, 0, 1);
+    __lsx_vstelm_d(tmp1, dst + dst_stride_x2, 0, 0);
+    __lsx_vstelm_d(tmp1, dst + dst_stride_x3, 0, 1);
+    dst += dst_stride_x4;
+  }
+}
+
+static void common_hz_6t_16w_lsx(uint8_t *RESTRICT src, int32_t src_stride,
+                                 uint8_t *RESTRICT dst, int32_t dst_stride,
+                                 const int8_t *filter, int32_t height) {
+  uint32_t loop_cnt;
+  int32_t src_stride_x2 = src_stride << 1;
+  int32_t src_stride_x3 = src_stride_x2 + src_stride;
+  int32_t src_stride_x4 = src_stride << 2;
+  int32_t dst_stride_x2 = dst_stride << 1;
+  int32_t dst_stride_x3 = dst_stride_x2 + dst_stride;
+  int32_t dst_stride_x4 = dst_stride << 2;
+  __m128i src0, src1, src2, src3, src4, src5, src6, src7, filt0, filt1, filt2;
+  __m128i mask0, mask1, mask2, out;
+  __m128i filt, out0, out1, out2, out3, out4, out5, out6, out7;
+
+  mask0 = __lsx_vld(vp8_mc_filt_mask_arr, 0);
+  src -= 2;
+
+  filt = __lsx_vld(filter, 0);
+  DUP2_ARG2(__lsx_vreplvei_h, filt, 0, filt, 1, filt0, filt1);
+  filt2 = __lsx_vreplvei_h(filt, 2);
+  DUP2_ARG2(__lsx_vaddi_bu, mask0, 2, mask0, 4, mask1, mask2);
+
+  for (loop_cnt = (height >> 2); loop_cnt--;) {
+    DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+              src_stride_x3, src0, src2, src4, src6);
+    src += 8;
+    DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+              src_stride_x3, src1, src3, src5, src7);
+    DUP4_ARG2(__lsx_vxori_b, src0, 128, src1, 128, src2, 128, src3, 128, src0,
+              src1, src2, src3);
+    DUP4_ARG2(__lsx_vxori_b, src4, 128, src5, 128, src6, 128, src7, 128, src4,
+              src5, src6, src7);
+    src += src_stride_x4 - 8;
+
+    HORIZ_6TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
+                               filt0, filt1, filt2, out0, out1, out2, out3);
+    HORIZ_6TAP_8WID_4VECS_FILT(src4, src5, src6, src7, mask0, mask1, mask2,
+                               filt0, filt1, filt2, out4, out5, out6, out7);
+    DUP4_ARG2(__lsx_vsrari_h, out0, VP8_FILTER_SHIFT, out1, VP8_FILTER_SHIFT,
+              out2, VP8_FILTER_SHIFT, out3, VP8_FILTER_SHIFT, out0, out1, out2,
+              out3);
+    DUP4_ARG2(__lsx_vsrari_h, out4, VP8_FILTER_SHIFT, out5, VP8_FILTER_SHIFT,
+              out6, VP8_FILTER_SHIFT, out7, VP8_FILTER_SHIFT, out4, out5, out6,
+              out7);
+    DUP4_ARG2(__lsx_vsat_h, out0, 7, out1, 7, out2, 7, out3, 7, out0, out1,
+              out2, out3);
+    DUP4_ARG2(__lsx_vsat_h, out4, 7, out5, 7, out6, 7, out7, 7, out4, out5,
+              out6, out7);
+    out = __lsx_vpickev_b(out1, out0);
+    out = __lsx_vxori_b(out, 128);
+    __lsx_vst(out, dst, 0);
+    out = __lsx_vpickev_b(out3, out2);
+    out = __lsx_vxori_b(out, 128);
+    __lsx_vstx(out, dst, dst_stride);
+    out = __lsx_vpickev_b(out5, out4);
+    out = __lsx_vxori_b(out, 128);
+    __lsx_vstx(out, dst, dst_stride_x2);
+    out = __lsx_vpickev_b(out7, out6);
+    out = __lsx_vxori_b(out, 128);
+    __lsx_vstx(out, dst, dst_stride_x3);
+    dst += dst_stride_x4;
+  }
+}
+
+static void common_vt_6t_8w_lsx(uint8_t *RESTRICT src, int32_t src_stride,
+                                uint8_t *RESTRICT dst, int32_t dst_stride,
+                                const int8_t *filter, int32_t height) {
+  uint32_t loop_cnt;
+  int32_t src_stride_x2 = src_stride << 1;
+  int32_t src_stride_x3 = src_stride_x2 + src_stride;
+  int32_t src_stride_x4 = src_stride << 2;
+  int32_t dst_stride_x2 = dst_stride << 1;
+  int32_t dst_stride_x3 = dst_stride_x2 + dst_stride;
+  int32_t dst_stride_x4 = dst_stride << 2;
+  __m128i src0, src1, src2, src3, src4, src7, src8, src9, src10;
+  __m128i src10_r, src32_r, src76_r, src98_r, src21_r, src43_r, src87_r;
+  __m128i src109_r, filt0, filt1, filt2;
+  __m128i tmp0, tmp1;
+  __m128i filt, out0_r, out1_r, out2_r, out3_r;
+
+  src -= src_stride_x2;
+  filt = __lsx_vld(filter, 0);
+  DUP2_ARG2(__lsx_vreplvei_h, filt, 0, filt, 1, filt0, filt1);
+  filt2 = __lsx_vreplvei_h(filt, 2);
+
+  DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+            src_stride_x3, src0, src1, src2, src3);
+  src += src_stride_x4;
+  src4 = __lsx_vld(src, 0);
+  src += src_stride;
+
+  DUP4_ARG2(__lsx_vxori_b, src0, 128, src1, 128, src2, 128, src3, 128, src0,
+            src1, src2, src3);
+  src4 = __lsx_vxori_b(src4, 128);
+  DUP4_ARG2(__lsx_vilvl_b, src1, src0, src3, src2, src2, src1, src4, src3,
+            src10_r, src32_r, src21_r, src43_r);
+
+  for (loop_cnt = (height >> 2); loop_cnt--;) {
+    DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+              src_stride_x3, src7, src8, src9, src10);
+    DUP4_ARG2(__lsx_vxori_b, src7, 128, src8, 128, src9, 128, src10, 128, src7,
+              src8, src9, src10);
+    src += src_stride_x4;
+
+    DUP4_ARG2(__lsx_vilvl_b, src7, src4, src8, src7, src9, src8, src10, src9,
+              src76_r, src87_r, src98_r, src109_r);
+    out0_r = DPADD_H3(src10_r, src32_r, src76_r, filt0, filt1, filt2);
+    out1_r = DPADD_H3(src21_r, src43_r, src87_r, filt0, filt1, filt2);
+    out2_r = DPADD_H3(src32_r, src76_r, src98_r, filt0, filt1, filt2);
+    out3_r = DPADD_H3(src43_r, src87_r, src109_r, filt0, filt1, filt2);
+    DUP2_ARG3(__lsx_vssrarni_b_h, out1_r, out0_r, VP8_FILTER_SHIFT, out3_r,
+              out2_r, VP8_FILTER_SHIFT, tmp0, tmp1);
+    DUP2_ARG2(__lsx_vxori_b, tmp0, 128, tmp1, 128, tmp0, tmp1);
+    __lsx_vstelm_d(tmp0, dst, 0, 0);
+    __lsx_vstelm_d(tmp0, dst + dst_stride, 0, 1);
+    __lsx_vstelm_d(tmp1, dst + dst_stride_x2, 0, 0);
+    __lsx_vstelm_d(tmp1, dst + dst_stride_x3, 0, 1);
+    dst += dst_stride_x4;
+
+    src10_r = src76_r;
+    src32_r = src98_r;
+    src21_r = src87_r;
+    src43_r = src109_r;
+    src4 = src10;
+  }
+}
+
+static void common_vt_6t_16w_lsx(uint8_t *RESTRICT src, int32_t src_stride,
+                                 uint8_t *RESTRICT dst, int32_t dst_stride,
+                                 const int8_t *filter, int32_t height) {
+  uint32_t loop_cnt;
+  int32_t src_stride_x2 = src_stride << 1;
+  int32_t src_stride_x3 = src_stride_x2 + src_stride;
+  int32_t src_stride_x4 = src_stride << 2;
+  int32_t dst_stride_x2 = dst_stride << 1;
+  int32_t dst_stride_x3 = dst_stride_x2 + dst_stride;
+  int32_t dst_stride_x4 = dst_stride << 2;
+  __m128i src0, src1, src2, src3, src4, src5, src6, src7, src8;
+  __m128i src10_r, src32_r, src54_r, src76_r, src21_r, src43_r, src65_r;
+  __m128i src87_r, src10_l, src32_l, src54_l, src76_l, src21_l, src43_l;
+  __m128i src65_l, src87_l, filt0, filt1, filt2;
+  __m128i tmp0, tmp1, tmp2, tmp3;
+  __m128i filt, out0_r, out1_r, out2_r, out3_r, out0_l, out1_l, out2_l, out3_l;
+
+  src -= src_stride_x2;
+  filt = __lsx_vld(filter, 0);
+  DUP2_ARG2(__lsx_vreplvei_h, filt, 0, filt, 1, filt0, filt1);
+  filt2 = __lsx_vreplvei_h(filt, 2);
+
+  DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+            src_stride_x3, src0, src1, src2, src3);
+  src += src_stride_x4;
+  src4 = __lsx_vldx(src, 0);
+  src += src_stride;
+
+  DUP4_ARG2(__lsx_vxori_b, src0, 128, src1, 128, src2, 128, src3, 128, src0,
+            src1, src2, src3);
+  src4 = __lsx_vxori_b(src4, 128);
+  DUP4_ARG2(__lsx_vilvl_b, src1, src0, src3, src2, src4, src3, src2, src1,
+            src10_r, src32_r, src43_r, src21_r);
+  DUP4_ARG2(__lsx_vilvh_b, src1, src0, src3, src2, src4, src3, src2, src1,
+            src10_l, src32_l, src43_l, src21_l);
+
+  for (loop_cnt = (height >> 2); loop_cnt--;) {
+    DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+              src_stride_x3, src5, src6, src7, src8);
+    src += src_stride_x4;
+
+    DUP4_ARG2(__lsx_vxori_b, src5, 128, src6, 128, src7, 128, src8, 128, src5,
+              src6, src7, src8);
+    DUP4_ARG2(__lsx_vilvl_b, src5, src4, src6, src5, src7, src6, src8, src7,
+              src54_r, src65_r, src76_r, src87_r);
+    DUP4_ARG2(__lsx_vilvh_b, src5, src4, src6, src5, src7, src6, src8, src7,
+              src54_l, src65_l, src76_l, src87_l);
+    out0_r = DPADD_H3(src10_r, src32_r, src54_r, filt0, filt1, filt2);
+    out1_r = DPADD_H3(src21_r, src43_r, src65_r, filt0, filt1, filt2);
+    out2_r = DPADD_H3(src32_r, src54_r, src76_r, filt0, filt1, filt2);
+    out3_r = DPADD_H3(src43_r, src65_r, src87_r, filt0, filt1, filt2);
+    out0_l = DPADD_H3(src10_l, src32_l, src54_l, filt0, filt1, filt2);
+    out1_l = DPADD_H3(src21_l, src43_l, src65_l, filt0, filt1, filt2);
+    out2_l = DPADD_H3(src32_l, src54_l, src76_l, filt0, filt1, filt2);
+    out3_l = DPADD_H3(src43_l, src65_l, src87_l, filt0, filt1, filt2);
+    DUP4_ARG3(__lsx_vssrarni_b_h, out0_l, out0_r, VP8_FILTER_SHIFT, out1_l,
+              out1_r, VP8_FILTER_SHIFT, out2_l, out2_r, VP8_FILTER_SHIFT,
+              out3_l, out3_r, VP8_FILTER_SHIFT, tmp0, tmp1, tmp2, tmp3);
+    DUP4_ARG2(__lsx_vxori_b, tmp0, 128, tmp1, 128, tmp2, 128, tmp3, 128, tmp0,
+              tmp1, tmp2, tmp3);
+    __lsx_vstx(tmp0, dst, 0);
+    __lsx_vstx(tmp1, dst, dst_stride);
+    __lsx_vstx(tmp2, dst, dst_stride_x2);
+    __lsx_vstx(tmp3, dst, dst_stride_x3);
+    dst += dst_stride_x4;
+
+    src10_r = src54_r;
+    src32_r = src76_r;
+    src21_r = src65_r;
+    src43_r = src87_r;
+    src10_l = src54_l;
+    src32_l = src76_l;
+    src21_l = src65_l;
+    src43_l = src87_l;
+    src4 = src8;
+  }
+}
+
+static void common_hv_6ht_6vt_8w_lsx(uint8_t *RESTRICT src, int32_t src_stride,
+                                     uint8_t *RESTRICT dst, int32_t dst_stride,
+                                     const int8_t *filter_horiz,
+                                     const int8_t *filter_vert,
+                                     int32_t height) {
+  uint32_t loop_cnt;
+  int32_t src_stride_x2 = src_stride << 1;
+  int32_t src_stride_x3 = src_stride_x2 + src_stride;
+  int32_t src_stride_x4 = src_stride << 2;
+  int32_t dst_stride_x2 = dst_stride << 1;
+  int32_t dst_stride_x3 = dst_stride_x2 + dst_stride;
+  int32_t dst_stride_x4 = dst_stride << 2;
+  __m128i src0, src1, src2, src3, src4, src5, src6, src7, src8;
+  __m128i filt_hz0, filt_hz1, filt_hz2;
+  __m128i mask0, mask1, mask2, vec0, vec1;
+  __m128i filt, filt_vt0, filt_vt1, filt_vt2;
+  __m128i hz_out0, hz_out1, hz_out2, hz_out3, hz_out4, hz_out5, hz_out6;
+  __m128i hz_out7, hz_out8, out0, out1, out2, out3, out4, out5, out6, out7;
+  __m128i tmp0, tmp1, tmp2, tmp3;
+
+  mask0 = __lsx_vld(vp8_mc_filt_mask_arr, 0);
+  src -= (2 + src_stride_x2);
+
+  filt = __lsx_vld(filter_horiz, 0);
+  DUP2_ARG2(__lsx_vreplvei_h, filt, 0, filt, 1, filt_hz0, filt_hz1);
+  filt_hz2 = __lsx_vreplvei_h(filt, 2);
+
+  DUP2_ARG2(__lsx_vaddi_bu, mask0, 2, mask0, 4, mask1, mask2);
+  DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+            src_stride_x3, src0, src1, src2, src3);
+  src += src_stride_x4;
+  src4 = __lsx_vldx(src, 0);
+  src += src_stride;
+
+  DUP4_ARG2(__lsx_vxori_b, src0, 128, src1, 128, src2, 128, src3, 128, src0,
+            src1, src2, src3);
+  src4 = __lsx_vxori_b(src4, 128);
+
+  hz_out0 = HORIZ_6TAP_FILT(src0, src0, mask0, mask1, mask2, filt_hz0, filt_hz1,
+                            filt_hz2);
+  hz_out1 = HORIZ_6TAP_FILT(src1, src1, mask0, mask1, mask2, filt_hz0, filt_hz1,
+                            filt_hz2);
+  hz_out2 = HORIZ_6TAP_FILT(src2, src2, mask0, mask1, mask2, filt_hz0, filt_hz1,
+                            filt_hz2);
+  hz_out3 = HORIZ_6TAP_FILT(src3, src3, mask0, mask1, mask2, filt_hz0, filt_hz1,
+                            filt_hz2);
+  hz_out4 = HORIZ_6TAP_FILT(src4, src4, mask0, mask1, mask2, filt_hz0, filt_hz1,
+                            filt_hz2);
+  filt = __lsx_vld(filter_vert, 0);
+  DUP2_ARG2(__lsx_vreplvei_h, filt, 0, filt, 1, filt_vt0, filt_vt1);
+  filt_vt2 = __lsx_vreplvei_h(filt, 2);
+
+  DUP4_ARG2(__lsx_vpackev_b, hz_out1, hz_out0, hz_out3, hz_out2, hz_out2,
+            hz_out1, hz_out4, hz_out3, out0, out1, out3, out4);
+
+  for (loop_cnt = (height >> 2); loop_cnt--;) {
+    DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+              src_stride_x3, src5, src6, src7, src8);
+    src += src_stride_x4;
+
+    DUP4_ARG2(__lsx_vxori_b, src5, 128, src6, 128, src7, 128, src8, 128, src5,
+              src6, src7, src8);
+    hz_out5 = HORIZ_6TAP_FILT(src5, src5, mask0, mask1, mask2, filt_hz0,
+                              filt_hz1, filt_hz2);
+    out2 = __lsx_vpackev_b(hz_out5, hz_out4);
+    tmp0 = DPADD_H3(out0, out1, out2, filt_vt0, filt_vt1, filt_vt2);
+
+    hz_out6 = HORIZ_6TAP_FILT(src6, src6, mask0, mask1, mask2, filt_hz0,
+                              filt_hz1, filt_hz2);
+    out5 = __lsx_vpackev_b(hz_out6, hz_out5);
+    tmp1 = DPADD_H3(out3, out4, out5, filt_vt0, filt_vt1, filt_vt2);
+
+    hz_out7 = HORIZ_6TAP_FILT(src7, src7, mask0, mask1, mask2, filt_hz0,
+                              filt_hz1, filt_hz2);
+    out7 = __lsx_vpackev_b(hz_out7, hz_out6);
+    tmp2 = DPADD_H3(out1, out2, out7, filt_vt0, filt_vt1, filt_vt2);
+
+    hz_out8 = HORIZ_6TAP_FILT(src8, src8, mask0, mask1, mask2, filt_hz0,
+                              filt_hz1, filt_hz2);
+    out6 = __lsx_vpackev_b(hz_out8, hz_out7);
+    tmp3 = DPADD_H3(out4, out5, out6, filt_vt0, filt_vt1, filt_vt2);
+
+    DUP2_ARG3(__lsx_vssrarni_b_h, tmp1, tmp0, VP8_FILTER_SHIFT, tmp3, tmp2,
+              VP8_FILTER_SHIFT, vec0, vec1);
+    DUP2_ARG2(__lsx_vxori_b, vec0, 128, vec1, 128, vec0, vec1);
+
+    __lsx_vstelm_d(vec0, dst, 0, 0);
+    __lsx_vstelm_d(vec0, dst + dst_stride, 0, 1);
+    __lsx_vstelm_d(vec1, dst + dst_stride_x2, 0, 0);
+    __lsx_vstelm_d(vec1, dst + dst_stride_x3, 0, 1);
+    dst += dst_stride_x4;
+
+    hz_out4 = hz_out8;
+    out0 = out2;
+    out1 = out7;
+    out3 = out5;
+    out4 = out6;
+  }
+}
+
+static void common_hv_6ht_6vt_16w_lsx(uint8_t *RESTRICT src, int32_t src_stride,
+                                      uint8_t *RESTRICT dst, int32_t dst_stride,
+                                      const int8_t *filter_horiz,
+                                      const int8_t *filter_vert,
+                                      int32_t height) {
+  common_hv_6ht_6vt_8w_lsx(src, src_stride, dst, dst_stride, filter_horiz,
+                           filter_vert, height);
+  common_hv_6ht_6vt_8w_lsx(src + 8, src_stride, dst + 8, dst_stride,
+                           filter_horiz, filter_vert, height);
+}
+
+static void common_hz_4t_8w_lsx(uint8_t *RESTRICT src, int32_t src_stride,
+                                uint8_t *RESTRICT dst, int32_t dst_stride,
+                                const int8_t *filter, int32_t height) {
+  uint32_t loop_cnt;
+  int32_t src_stride_x2 = src_stride << 1;
+  int32_t src_stride_x3 = src_stride_x2 + src_stride;
+  int32_t src_stride_x4 = src_stride << 2;
+  int32_t dst_stride_x2 = dst_stride << 1;
+  int32_t dst_stride_x3 = dst_stride_x2 + dst_stride;
+  int32_t dst_stride_x4 = dst_stride << 2;
+  __m128i src0, src1, src2, src3, filt0, filt1, mask0, mask1;
+  __m128i tmp0, tmp1;
+  __m128i filt, out0, out1, out2, out3;
+
+  mask0 = __lsx_vld(vp8_mc_filt_mask_arr, 0);
+  src -= 1;
+
+  filt = __lsx_vld(filter, 0);
+  DUP2_ARG2(__lsx_vreplvei_h, filt, 0, filt, 1, filt0, filt1);
+  mask1 = __lsx_vaddi_bu(mask0, 2);
+
+  for (loop_cnt = (height >> 2); loop_cnt--;) {
+    DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+              src_stride_x3, src0, src1, src2, src3);
+    src += src_stride_x4;
+
+    DUP4_ARG2(__lsx_vxori_b, src0, 128, src1, 128, src2, 128, src3, 128, src0,
+              src1, src2, src3);
+    HORIZ_4TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, filt0,
+                               filt1, out0, out1, out2, out3);
+    DUP2_ARG3(__lsx_vssrarni_b_h, out1, out0, VP8_FILTER_SHIFT, out3, out2,
+              VP8_FILTER_SHIFT, tmp0, tmp1);
+    DUP2_ARG2(__lsx_vxori_b, tmp0, 128, tmp1, 128, tmp0, tmp1);
+    __lsx_vstelm_d(tmp0, dst, 0, 0);
+    __lsx_vstelm_d(tmp0, dst + dst_stride, 0, 1);
+    __lsx_vstelm_d(tmp1, dst + dst_stride_x2, 0, 0);
+    __lsx_vstelm_d(tmp1, dst + dst_stride_x3, 0, 1);
+    dst += dst_stride_x4;
+  }
+}
+
+static void common_hz_4t_16w_lsx(uint8_t *RESTRICT src, int32_t src_stride,
+                                 uint8_t *RESTRICT dst, int32_t dst_stride,
+                                 const int8_t *filter, int32_t height) {
+  uint32_t loop_cnt;
+  int32_t src_stride_x2 = src_stride << 1;
+  int32_t src_stride_x3 = src_stride_x2 + src_stride;
+  int32_t src_stride_x4 = src_stride << 2;
+  int32_t dst_stride_x2 = dst_stride << 1;
+  int32_t dst_stride_x3 = dst_stride_x2 + dst_stride;
+  int32_t dst_stride_x4 = dst_stride << 2;
+  __m128i src0, src1, src2, src3, src4, src5, src6, src7;
+  __m128i filt0, filt1, mask0, mask1;
+  __m128i filt, out0, out1, out2, out3, out4, out5, out6, out7;
+
+  mask0 = __lsx_vld(vp8_mc_filt_mask_arr, 0);
+  src -= 1;
+
+  filt = __lsx_vld(filter, 0);
+  DUP2_ARG2(__lsx_vreplvei_h, filt, 0, filt, 1, filt0, filt1);
+  mask1 = __lsx_vaddi_bu(mask0, 2);
+
+  for (loop_cnt = (height >> 2); loop_cnt--;) {
+    DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+              src_stride_x3, src0, src2, src4, src6);
+    src += 8;
+    DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+              src_stride_x3, src1, src3, src5, src7);
+    src += src_stride_x4 - 8;
+
+    DUP4_ARG2(__lsx_vxori_b, src0, 128, src1, 128, src2, 128, src3, 128, src0,
+              src1, src2, src3);
+    DUP4_ARG2(__lsx_vxori_b, src4, 128, src5, 128, src6, 128, src7, 128, src4,
+              src5, src6, src7);
+    HORIZ_4TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, filt0,
+                               filt1, out0, out1, out2, out3);
+    HORIZ_4TAP_8WID_4VECS_FILT(src4, src5, src6, src7, mask0, mask1, filt0,
+                               filt1, out4, out5, out6, out7);
+    DUP4_ARG3(__lsx_vssrarni_b_h, out1, out0, VP8_FILTER_SHIFT, out3, out2,
+              VP8_FILTER_SHIFT, out5, out4, VP8_FILTER_SHIFT, out7, out6,
+              VP8_FILTER_SHIFT, out0, out1, out2, out3);
+    DUP4_ARG2(__lsx_vxori_b, out0, 128, out1, 128, out2, 128, out3, 128, out0,
+              out1, out2, out3);
+    __lsx_vstx(out0, dst, 0);
+    __lsx_vstx(out1, dst, dst_stride);
+    __lsx_vstx(out2, dst, dst_stride_x2);
+    __lsx_vstx(out3, dst, dst_stride_x3);
+    dst += dst_stride_x4;
+  }
+}
+
+static void common_vt_4t_8w_lsx(uint8_t *RESTRICT src, int32_t src_stride,
+                                uint8_t *RESTRICT dst, int32_t dst_stride,
+                                const int8_t *filter, int32_t height) {
+  uint32_t loop_cnt;
+  int32_t src_stride_x2 = src_stride << 1;
+  int32_t src_stride_x3 = src_stride_x2 + src_stride;
+  int32_t src_stride_x4 = src_stride << 2;
+  int32_t dst_stride_x2 = dst_stride << 1;
+  int32_t dst_stride_x3 = dst_stride_x2 + dst_stride;
+  int32_t dst_stride_x4 = dst_stride << 2;
+  __m128i src0, src1, src2, src7, src8, src9, src10;
+  __m128i src10_r, src72_r, src98_r, src21_r, src87_r, src109_r, filt0, filt1;
+  __m128i tmp0, tmp1;
+  __m128i filt, out0_r, out1_r, out2_r, out3_r;
+
+  src -= src_stride;
+  filt = __lsx_vld(filter, 0);
+  DUP2_ARG2(__lsx_vreplvei_h, filt, 0, filt, 1, filt0, filt1);
+
+  DUP2_ARG2(__lsx_vldx, src, 0, src, src_stride, src0, src1);
+  src2 = __lsx_vldx(src, src_stride_x2);
+  src += src_stride_x3;
+
+  DUP2_ARG2(__lsx_vxori_b, src0, 128, src1, 128, src0, src1);
+  src2 = __lsx_vxori_b(src2, 128);
+  DUP2_ARG2(__lsx_vilvl_b, src1, src0, src2, src1, src10_r, src21_r);
+
+  for (loop_cnt = (height >> 2); loop_cnt--;) {
+    DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+              src_stride_x3, src7, src8, src9, src10);
+    src += src_stride_x4;
+
+    DUP4_ARG2(__lsx_vxori_b, src7, 128, src8, 128, src9, 128, src10, 128, src7,
+              src8, src9, src10);
+    DUP4_ARG2(__lsx_vilvl_b, src7, src2, src8, src7, src9, src8, src10, src9,
+              src72_r, src87_r, src98_r, src109_r);
+    out0_r = FILT_4TAP_DPADD_H(src10_r, src72_r, filt0, filt1);
+    out1_r = FILT_4TAP_DPADD_H(src21_r, src87_r, filt0, filt1);
+    out2_r = FILT_4TAP_DPADD_H(src72_r, src98_r, filt0, filt1);
+    out3_r = FILT_4TAP_DPADD_H(src87_r, src109_r, filt0, filt1);
+    DUP2_ARG3(__lsx_vssrarni_b_h, out1_r, out0_r, VP8_FILTER_SHIFT, out3_r,
+              out2_r, VP8_FILTER_SHIFT, tmp0, tmp1);
+    DUP2_ARG2(__lsx_vxori_b, tmp0, 128, tmp1, 128, tmp0, tmp1);
+    __lsx_vstelm_d(tmp0, dst, 0, 0);
+    __lsx_vstelm_d(tmp0, dst + dst_stride, 0, 1);
+    __lsx_vstelm_d(tmp1, dst + dst_stride_x2, 0, 0);
+    __lsx_vstelm_d(tmp1, dst + dst_stride_x3, 0, 1);
+    dst += dst_stride_x4;
+
+    src10_r = src98_r;
+    src21_r = src109_r;
+    src2 = src10;
+  }
+}
+
+static void common_vt_4t_16w_lsx(uint8_t *RESTRICT src, int32_t src_stride,
+                                 uint8_t *RESTRICT dst, int32_t dst_stride,
+                                 const int8_t *filter, int32_t height) {
+  uint32_t loop_cnt;
+  int32_t src_stride_x2 = src_stride << 1;
+  int32_t src_stride_x3 = src_stride_x2 + src_stride;
+  int32_t src_stride_x4 = src_stride << 2;
+  int32_t dst_stride_x2 = dst_stride << 1;
+  int32_t dst_stride_x3 = dst_stride_x2 + dst_stride;
+  int32_t dst_stride_x4 = dst_stride << 2;
+  __m128i src0, src1, src2, src3, src4, src5, src6;
+  __m128i src10_r, src32_r, src54_r, src21_r, src43_r, src65_r, src10_l;
+  __m128i src32_l, src54_l, src21_l, src43_l, src65_l, filt0, filt1;
+  __m128i tmp0, tmp1, tmp2, tmp3;
+  __m128i filt, out0_r, out1_r, out2_r, out3_r, out0_l, out1_l, out2_l, out3_l;
+
+  src -= src_stride;
+  filt = __lsx_vld(filter, 0);
+  DUP2_ARG2(__lsx_vreplvei_h, filt, 0, filt, 1, filt0, filt1);
+
+  DUP2_ARG2(__lsx_vldx, src, 0, src, src_stride, src0, src1);
+  src2 = __lsx_vldx(src, src_stride_x2);
+  src += src_stride_x3;
+
+  DUP2_ARG2(__lsx_vxori_b, src0, 128, src1, 128, src0, src1);
+  src2 = __lsx_vxori_b(src2, 128);
+  DUP2_ARG2(__lsx_vilvl_b, src1, src0, src2, src1, src10_r, src21_r);
+  DUP2_ARG2(__lsx_vilvh_b, src1, src0, src2, src1, src10_l, src21_l);
+
+  for (loop_cnt = (height >> 2); loop_cnt--;) {
+    DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+              src_stride_x3, src3, src4, src5, src6);
+    src += src_stride_x4;
+
+    DUP4_ARG2(__lsx_vxori_b, src3, 128, src4, 128, src5, 128, src6, 128, src3,
+              src4, src5, src6);
+    DUP4_ARG2(__lsx_vilvl_b, src3, src2, src4, src3, src5, src4, src6, src5,
+              src32_r, src43_r, src54_r, src65_r);
+    DUP4_ARG2(__lsx_vilvh_b, src3, src2, src4, src3, src5, src4, src6, src5,
+              src32_l, src43_l, src54_l, src65_l);
+    out0_r = FILT_4TAP_DPADD_H(src10_r, src32_r, filt0, filt1);
+    out1_r = FILT_4TAP_DPADD_H(src21_r, src43_r, filt0, filt1);
+    out2_r = FILT_4TAP_DPADD_H(src32_r, src54_r, filt0, filt1);
+    out3_r = FILT_4TAP_DPADD_H(src43_r, src65_r, filt0, filt1);
+    out0_l = FILT_4TAP_DPADD_H(src10_l, src32_l, filt0, filt1);
+    out1_l = FILT_4TAP_DPADD_H(src21_l, src43_l, filt0, filt1);
+    out2_l = FILT_4TAP_DPADD_H(src32_l, src54_l, filt0, filt1);
+    out3_l = FILT_4TAP_DPADD_H(src43_l, src65_l, filt0, filt1);
+    DUP4_ARG3(__lsx_vssrarni_b_h, out0_l, out0_r, VP8_FILTER_SHIFT, out1_l,
+              out1_r, VP8_FILTER_SHIFT, out2_l, out2_r, VP8_FILTER_SHIFT,
+              out3_l, out3_r, VP8_FILTER_SHIFT, tmp0, tmp1, tmp2, tmp3);
+    DUP4_ARG2(__lsx_vxori_b, tmp0, 128, tmp1, 128, tmp2, 128, tmp3, 128, tmp0,
+              tmp1, tmp2, tmp3);
+    __lsx_vstx(tmp0, dst, 0);
+    __lsx_vstx(tmp1, dst, dst_stride);
+    __lsx_vstx(tmp2, dst, dst_stride_x2);
+    __lsx_vstx(tmp3, dst, dst_stride_x3);
+    dst += dst_stride_x4;
+
+    src10_r = src54_r;
+    src21_r = src65_r;
+    src10_l = src54_l;
+    src21_l = src65_l;
+    src2 = src6;
+  }
+}
+
+static inline void common_hv_4ht_4vt_8w_lsx(
+    uint8_t *RESTRICT src, int32_t src_stride, uint8_t *RESTRICT dst,
+    int32_t dst_stride, const int8_t *filter_horiz, const int8_t *filter_vert,
+    int32_t height) {
+  uint32_t loop_cnt;
+  int32_t src_stride_x2 = src_stride << 1;
+  int32_t src_stride_x3 = src_stride_x2 + src_stride;
+  int32_t src_stride_x4 = src_stride << 2;
+  int32_t dst_stride_x2 = dst_stride << 1;
+  int32_t dst_stride_x3 = dst_stride_x2 + dst_stride;
+  int32_t dst_stride_x4 = dst_stride << 2;
+  __m128i src0, src1, src2, src3, src4, src5, src6, filt_hz0, filt_hz1;
+  __m128i mask0, mask1, out0, out1;
+  __m128i filt, filt_vt0, filt_vt1, tmp0, tmp1, tmp2, tmp3;
+  __m128i hz_out0, hz_out1, hz_out2, hz_out3;
+  __m128i vec0, vec1, vec2, vec3, vec4;
+
+  mask0 = __lsx_vld(vp8_mc_filt_mask_arr, 0);
+  src -= 1 + src_stride;
+
+  filt = __lsx_vld(filter_horiz, 0);
+  DUP2_ARG2(__lsx_vreplvei_h, filt, 0, filt, 1, filt_hz0, filt_hz1);
+  mask1 = __lsx_vaddi_bu(mask0, 2);
+
+  DUP2_ARG2(__lsx_vldx, src, 0, src, src_stride, src0, src1);
+  src2 = __lsx_vldx(src, src_stride_x2);
+  src += src_stride_x3;
+
+  DUP2_ARG2(__lsx_vxori_b, src0, 128, src1, 128, src0, src1);
+  src2 = __lsx_vxori_b(src2, 128);
+  hz_out0 = HORIZ_4TAP_FILT(src0, src0, mask0, mask1, filt_hz0, filt_hz1);
+  hz_out1 = HORIZ_4TAP_FILT(src1, src1, mask0, mask1, filt_hz0, filt_hz1);
+  hz_out2 = HORIZ_4TAP_FILT(src2, src2, mask0, mask1, filt_hz0, filt_hz1);
+  DUP2_ARG2(__lsx_vpackev_b, hz_out1, hz_out0, hz_out2, hz_out1, vec0, vec2);
+
+  filt = __lsx_vld(filter_vert, 0);
+  DUP2_ARG2(__lsx_vreplvei_h, filt, 0, filt, 1, filt_vt0, filt_vt1);
+
+  for (loop_cnt = (height >> 2); loop_cnt--;) {
+    DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+              src_stride_x3, src3, src4, src5, src6);
+    src += src_stride_x4;
+
+    DUP4_ARG2(__lsx_vxori_b, src3, 128, src4, 128, src5, 128, src6, 128, src3,
+              src4, src5, src6);
+    hz_out3 = HORIZ_4TAP_FILT(src3, src3, mask0, mask1, filt_hz0, filt_hz1);
+    vec1 = __lsx_vpackev_b(hz_out3, hz_out2);
+    tmp0 = FILT_4TAP_DPADD_H(vec0, vec1, filt_vt0, filt_vt1);
+
+    hz_out0 = HORIZ_4TAP_FILT(src4, src4, mask0, mask1, filt_hz0, filt_hz1);
+    vec3 = __lsx_vpackev_b(hz_out0, hz_out3);
+    tmp1 = FILT_4TAP_DPADD_H(vec2, vec3, filt_vt0, filt_vt1);
+
+    hz_out1 = HORIZ_4TAP_FILT(src5, src5, mask0, mask1, filt_hz0, filt_hz1);
+    vec4 = __lsx_vpackev_b(hz_out1, hz_out0);
+    tmp2 = FILT_4TAP_DPADD_H(vec1, vec4, filt_vt0, filt_vt1);
+
+    hz_out2 = HORIZ_4TAP_FILT(src6, src6, mask0, mask1, filt_hz0, filt_hz1);
+    DUP2_ARG2(__lsx_vpackev_b, hz_out0, hz_out3, hz_out2, hz_out1, vec0, vec1);
+    tmp3 = FILT_4TAP_DPADD_H(vec0, vec1, filt_vt0, filt_vt1);
+
+    DUP2_ARG3(__lsx_vssrarni_b_h, tmp1, tmp0, 7, tmp3, tmp2, 7, out0, out1);
+    DUP2_ARG2(__lsx_vxori_b, out0, 128, out1, 128, out0, out1);
+    __lsx_vstelm_d(out0, dst, 0, 0);
+    __lsx_vstelm_d(out0, dst + dst_stride, 0, 1);
+    __lsx_vstelm_d(out1, dst + dst_stride_x2, 0, 0);
+    __lsx_vstelm_d(out1, dst + dst_stride_x3, 0, 1);
+    dst += dst_stride_x4;
+
+    vec0 = vec4;
+    vec2 = vec1;
+  }
+}
+
+static void common_hv_4ht_4vt_16w_lsx(uint8_t *RESTRICT src, int32_t src_stride,
+                                      uint8_t *RESTRICT dst, int32_t dst_stride,
+                                      const int8_t *filter_horiz,
+                                      const int8_t *filter_vert,
+                                      int32_t height) {
+  common_hv_4ht_4vt_8w_lsx(src, src_stride, dst, dst_stride, filter_horiz,
+                           filter_vert, height);
+  common_hv_4ht_4vt_8w_lsx(src + 8, src_stride, dst + 8, dst_stride,
+                           filter_horiz, filter_vert, height);
+}
+
+static inline void common_hv_6ht_4vt_8w_lsx(
+    uint8_t *RESTRICT src, int32_t src_stride, uint8_t *RESTRICT dst,
+    int32_t dst_stride, const int8_t *filter_horiz, const int8_t *filter_vert,
+    int32_t height) {
+  uint32_t loop_cnt;
+  int32_t src_stride_x2 = src_stride << 1;
+  int32_t src_stride_x3 = src_stride_x2 + src_stride;
+  int32_t src_stride_x4 = src_stride << 2;
+  int32_t dst_stride_x2 = dst_stride << 1;
+  int32_t dst_stride_x3 = dst_stride_x2 + dst_stride;
+  int32_t dst_stride_x4 = dst_stride << 2;
+
+  __m128i src0, src1, src2, src3, src4, src5, src6;
+  __m128i filt_hz0, filt_hz1, filt_hz2, mask0, mask1, mask2;
+  __m128i filt, filt_vt0, filt_vt1, hz_out0, hz_out1, hz_out2, hz_out3;
+  __m128i tmp0, tmp1, tmp2, tmp3, vec0, vec1, vec2, vec3;
+  __m128i out0, out1;
+
+  mask0 = __lsx_vld(vp8_mc_filt_mask_arr, 0);
+  src -= (2 + src_stride);
+
+  filt = __lsx_vld(filter_horiz, 0);
+  DUP2_ARG2(__lsx_vreplvei_h, filt, 0, filt, 1, filt_hz0, filt_hz1);
+  filt_hz2 = __lsx_vreplvei_h(filt, 2);
+  DUP2_ARG2(__lsx_vaddi_bu, mask0, 2, mask0, 4, mask1, mask2);
+
+  DUP2_ARG2(__lsx_vldx, src, 0, src, src_stride, src0, src1);
+  src2 = __lsx_vldx(src, src_stride_x2);
+  src += src_stride_x3;
+
+  DUP2_ARG2(__lsx_vxori_b, src0, 128, src1, 128, src0, src1);
+  src2 = __lsx_vxori_b(src2, 128);
+  hz_out0 = HORIZ_6TAP_FILT(src0, src0, mask0, mask1, mask2, filt_hz0, filt_hz1,
+                            filt_hz2);
+  hz_out1 = HORIZ_6TAP_FILT(src1, src1, mask0, mask1, mask2, filt_hz0, filt_hz1,
+                            filt_hz2);
+  hz_out2 = HORIZ_6TAP_FILT(src2, src2, mask0, mask1, mask2, filt_hz0, filt_hz1,
+                            filt_hz2);
+  DUP2_ARG2(__lsx_vpackev_b, hz_out1, hz_out0, hz_out2, hz_out1, vec0, vec2);
+
+  filt = __lsx_vld(filter_vert, 0);
+  DUP2_ARG2(__lsx_vreplvei_h, filt, 0, filt, 1, filt_vt0, filt_vt1);
+
+  for (loop_cnt = (height >> 2); loop_cnt--;) {
+    DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+              src_stride_x3, src3, src4, src5, src6);
+    src += src_stride_x4;
+    DUP4_ARG2(__lsx_vxori_b, src3, 128, src4, 128, src5, 128, src6, 128, src3,
+              src4, src5, src6);
+
+    hz_out3 = HORIZ_6TAP_FILT(src3, src3, mask0, mask1, mask2, filt_hz0,
+                              filt_hz1, filt_hz2);
+    vec1 = __lsx_vpackev_b(hz_out3, hz_out2);
+    tmp0 = FILT_4TAP_DPADD_H(vec0, vec1, filt_vt0, filt_vt1);
+
+    hz_out0 = HORIZ_6TAP_FILT(src4, src4, mask0, mask1, mask2, filt_hz0,
+                              filt_hz1, filt_hz2);
+    vec3 = __lsx_vpackev_b(hz_out0, hz_out3);
+    tmp1 = FILT_4TAP_DPADD_H(vec2, vec3, filt_vt0, filt_vt1);
+
+    hz_out1 = HORIZ_6TAP_FILT(src5, src5, mask0, mask1, mask2, filt_hz0,
+                              filt_hz1, filt_hz2);
+    vec0 = __lsx_vpackev_b(hz_out1, hz_out0);
+    tmp2 = FILT_4TAP_DPADD_H(vec1, vec0, filt_vt0, filt_vt1);
+
+    hz_out2 = HORIZ_6TAP_FILT(src6, src6, mask0, mask1, mask2, filt_hz0,
+                              filt_hz1, filt_hz2);
+    DUP2_ARG2(__lsx_vpackev_b, hz_out0, hz_out3, hz_out2, hz_out1, vec1, vec2);
+    tmp3 = FILT_4TAP_DPADD_H(vec1, vec2, filt_vt0, filt_vt1);
+
+    DUP2_ARG3(__lsx_vssrarni_b_h, tmp1, tmp0, 7, tmp3, tmp2, 7, out0, out1);
+    DUP2_ARG2(__lsx_vxori_b, out0, 128, out1, 128, out0, out1);
+    __lsx_vstelm_d(out0, dst, 0, 0);
+    __lsx_vstelm_d(out0, dst + dst_stride, 0, 1);
+    __lsx_vstelm_d(out1, dst + dst_stride_x2, 0, 0);
+    __lsx_vstelm_d(out1, dst + dst_stride_x3, 0, 1);
+    dst += dst_stride_x4;
+  }
+}
+
+static void common_hv_6ht_4vt_16w_lsx(uint8_t *RESTRICT src, int32_t src_stride,
+                                      uint8_t *RESTRICT dst, int32_t dst_stride,
+                                      const int8_t *filter_horiz,
+                                      const int8_t *filter_vert,
+                                      int32_t height) {
+  common_hv_6ht_4vt_8w_lsx(src, src_stride, dst, dst_stride, filter_horiz,
+                           filter_vert, height);
+  common_hv_6ht_4vt_8w_lsx(src + 8, src_stride, dst + 8, dst_stride,
+                           filter_horiz, filter_vert, height);
+}
+
+static inline void common_hv_4ht_6vt_8w_lsx(
+    uint8_t *RESTRICT src, int32_t src_stride, uint8_t *RESTRICT dst,
+    int32_t dst_stride, const int8_t *filter_horiz, const int8_t *filter_vert,
+    int32_t height) {
+  uint32_t loop_cnt;
+  int32_t src_stride_x2 = src_stride << 1;
+  int32_t src_stride_x3 = src_stride_x2 + src_stride;
+  int32_t src_stride_x4 = src_stride << 2;
+  int32_t dst_stride_x2 = dst_stride << 1;
+  int32_t dst_stride_x3 = dst_stride_x2 + dst_stride;
+  int32_t dst_stride_x4 = dst_stride << 2;
+  __m128i src0, src1, src2, src3, src4, src5, src6, src7, src8;
+  __m128i filt_hz0, filt_hz1, mask0, mask1;
+  __m128i filt, filt_vt0, filt_vt1, filt_vt2, tmp0, tmp1, tmp2, tmp3;
+  __m128i hz_out0, hz_out1, hz_out2, hz_out3, hz_out4, hz_out5, hz_out6;
+  __m128i hz_out7, hz_out8, out0, out1, out2, out3, out4, out5, out6, out7;
+  __m128i vec0, vec1;
+
+  mask0 = __lsx_vld(vp8_mc_filt_mask_arr, 0);
+  src -= 1 + src_stride_x2;
+
+  filt = __lsx_vld(filter_horiz, 0);
+  DUP2_ARG2(__lsx_vreplvei_h, filt, 0, filt, 1, filt_hz0, filt_hz1);
+  mask1 = __lsx_vaddi_bu(mask0, 2);
+
+  DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+            src_stride_x3, src0, src1, src2, src3);
+  src += src_stride_x4;
+  src4 = __lsx_vld(src, 0);
+  src += src_stride;
+
+  DUP4_ARG2(__lsx_vxori_b, src0, 128, src1, 128, src2, 128, src3, 128, src0,
+            src1, src2, src3);
+  src4 = __lsx_vxori_b(src4, 128);
+  hz_out0 = HORIZ_4TAP_FILT(src0, src0, mask0, mask1, filt_hz0, filt_hz1);
+  hz_out1 = HORIZ_4TAP_FILT(src1, src1, mask0, mask1, filt_hz0, filt_hz1);
+  hz_out2 = HORIZ_4TAP_FILT(src2, src2, mask0, mask1, filt_hz0, filt_hz1);
+  hz_out3 = HORIZ_4TAP_FILT(src3, src3, mask0, mask1, filt_hz0, filt_hz1);
+  hz_out4 = HORIZ_4TAP_FILT(src4, src4, mask0, mask1, filt_hz0, filt_hz1);
+  DUP2_ARG2(__lsx_vpackev_b, hz_out1, hz_out0, hz_out3, hz_out2, out0, out1);
+  DUP2_ARG2(__lsx_vpackev_b, hz_out2, hz_out1, hz_out4, hz_out3, out3, out4);
+
+  filt = __lsx_vld(filter_vert, 0);
+  DUP2_ARG2(__lsx_vreplvei_h, filt, 0, filt, 1, filt_vt0, filt_vt1);
+  filt_vt2 = __lsx_vreplvei_h(filt, 2);
+
+  for (loop_cnt = (height >> 2); loop_cnt--;) {
+    DUP4_ARG2(__lsx_vldx, src, 0, src, src_stride, src, src_stride_x2, src,
+              src_stride_x3, src5, src6, src7, src8);
+    src += src_stride_x4;
+
+    DUP4_ARG2(__lsx_vxori_b, src5, 128, src6, 128, src7, 128, src8, 128, src5,
+              src6, src7, src8);
+    hz_out5 = HORIZ_4TAP_FILT(src5, src5, mask0, mask1, filt_hz0, filt_hz1);
+    out2 = __lsx_vpackev_b(hz_out5, hz_out4);
+    tmp0 = DPADD_H3(out0, out1, out2, filt_vt0, filt_vt1, filt_vt2);
+
+    hz_out6 = HORIZ_4TAP_FILT(src6, src6, mask0, mask1, filt_hz0, filt_hz1);
+    out5 = __lsx_vpackev_b(hz_out6, hz_out5);
+    tmp1 = DPADD_H3(out3, out4, out5, filt_vt0, filt_vt1, filt_vt2);
+
+    hz_out7 = HORIZ_4TAP_FILT(src7, src7, mask0, mask1, filt_hz0, filt_hz1);
+    out6 = __lsx_vpackev_b(hz_out7, hz_out6);
+    tmp2 = DPADD_H3(out1, out2, out6, filt_vt0, filt_vt1, filt_vt2);
+
+    hz_out8 = HORIZ_4TAP_FILT(src8, src8, mask0, mask1, filt_hz0, filt_hz1);
+    out7 = __lsx_vpackev_b(hz_out8, hz_out7);
+    tmp3 = DPADD_H3(out4, out5, out7, filt_vt0, filt_vt1, filt_vt2);
+    DUP2_ARG3(__lsx_vssrarni_b_h, tmp1, tmp0, 7, tmp3, tmp2, 7, vec0, vec1);
+    DUP2_ARG2(__lsx_vxori_b, vec0, 128, vec1, 128, vec0, vec1);
+    __lsx_vstelm_d(vec0, dst, 0, 0);
+    __lsx_vstelm_d(vec0, dst + dst_stride, 0, 1);
+    __lsx_vstelm_d(vec1, dst + dst_stride_x2, 0, 0);
+    __lsx_vstelm_d(vec1, dst + dst_stride_x3, 0, 1);
+    dst += dst_stride_x4;
+    hz_out4 = hz_out8;
+    out0 = out2;
+    out1 = out6;
+    out3 = out5;
+    out4 = out7;
+  }
+}
+
+static void common_hv_4ht_6vt_16w_lsx(uint8_t *RESTRICT src, int32_t src_stride,
+                                      uint8_t *RESTRICT dst, int32_t dst_stride,
+                                      const int8_t *filter_horiz,
+                                      const int8_t *filter_vert,
+                                      int32_t height) {
+  common_hv_4ht_6vt_8w_lsx(src, src_stride, dst, dst_stride, filter_horiz,
+                           filter_vert, height);
+  common_hv_4ht_6vt_8w_lsx(src + 8, src_stride, dst + 8, dst_stride,
+                           filter_horiz, filter_vert, height);
+}
+
+typedef void (*PVp8SixtapPredictFunc1)(
+    uint8_t *RESTRICT src, int32_t src_stride, uint8_t *RESTRICT dst,
+    int32_t dst_stride, const int8_t *filter_horiz, const int8_t *filter_vert,
+    int32_t height);
+
+typedef void (*PVp8SixtapPredictFunc2)(uint8_t *RESTRICT src,
+                                       int32_t src_stride,
+                                       uint8_t *RESTRICT dst,
+                                       int32_t dst_stride, const int8_t *filter,
+                                       int32_t height);
+
+void vp8_sixtap_predict8x8_lsx(uint8_t *RESTRICT src, int32_t src_stride,
+                               int32_t xoffset, int32_t yoffset,
+                               uint8_t *RESTRICT dst, int32_t dst_stride) {
+  const int8_t *h_filter = vp8_subpel_filters_lsx[xoffset - 1];
+  const int8_t *v_filter = vp8_subpel_filters_lsx[yoffset - 1];
+
+  static PVp8SixtapPredictFunc1 Predict8x8Funcs1[4] = {
+    common_hv_6ht_6vt_8w_lsx,
+    common_hv_6ht_4vt_8w_lsx,
+    common_hv_4ht_6vt_8w_lsx,
+    common_hv_4ht_4vt_8w_lsx,
+  };
+
+  static PVp8SixtapPredictFunc2 Predict8x8Funcs2[4] = { common_vt_6t_8w_lsx,
+                                                        common_vt_4t_8w_lsx,
+                                                        common_hz_6t_8w_lsx,
+                                                        common_hz_4t_8w_lsx };
+
+  if (yoffset < 8 && xoffset < 8) {
+    if (yoffset) {
+      if (xoffset) {
+        switch (xoffset & 1) {
+          case 0:
+            switch (yoffset & 1) {
+              case 0:
+                Predict8x8Funcs1[0](src, src_stride, dst, dst_stride, h_filter,
+                                    v_filter, 8);
+                break;
+
+              case 1:
+                Predict8x8Funcs1[1](src, src_stride, dst, dst_stride, h_filter,
+                                    v_filter + 1, 8);
+                break;
+            }
+            break;
+
+          case 1:
+            switch (yoffset & 1) {
+              case 0:
+                Predict8x8Funcs1[2](src, src_stride, dst, dst_stride,
+                                    h_filter + 1, v_filter, 8);
+                break;
+
+              case 1:
+                Predict8x8Funcs1[3](src, src_stride, dst, dst_stride,
+                                    h_filter + 1, v_filter + 1, 8);
+                break;
+            }
+            break;
+        }
+      } else {
+        switch (yoffset & 1) {
+          case 0:
+            Predict8x8Funcs2[0](src, src_stride, dst, dst_stride, v_filter, 8);
+            break;
+
+          case 1:
+            Predict8x8Funcs2[1](src, src_stride, dst, dst_stride, v_filter + 1,
+                                8);
+            break;
+        }
+      }
+    } else {
+      switch (xoffset & 1) {
+        case 1:
+          Predict8x8Funcs2[3](src, src_stride, dst, dst_stride, h_filter + 1,
+                              8);
+          break;
+      }
+      switch (xoffset) {
+        case 0: vp8_copy_mem8x8(src, src_stride, dst, dst_stride); break;
+        case 2:
+        case 4:
+        case 6:
+          Predict8x8Funcs2[2](src, src_stride, dst, dst_stride, h_filter, 8);
+          break;
+      }
+    }
+  }
+}
+
+void vp8_sixtap_predict16x16_lsx(uint8_t *RESTRICT src, int32_t src_stride,
+                                 int32_t xoffset, int32_t yoffset,
+                                 uint8_t *RESTRICT dst, int32_t dst_stride) {
+  const int8_t *h_filter = vp8_subpel_filters_lsx[xoffset - 1];
+  const int8_t *v_filter = vp8_subpel_filters_lsx[yoffset - 1];
+
+  static PVp8SixtapPredictFunc1 Predict16x16Funcs1[4] = {
+    common_hv_6ht_6vt_16w_lsx,
+    common_hv_6ht_4vt_16w_lsx,
+    common_hv_4ht_6vt_16w_lsx,
+    common_hv_4ht_4vt_16w_lsx,
+  };
+
+  static PVp8SixtapPredictFunc2 Predict16x16Funcs2[4] = {
+    common_vt_6t_16w_lsx, common_vt_4t_16w_lsx, common_hz_6t_16w_lsx,
+    common_hz_4t_16w_lsx
+  };
+
+  if (yoffset < 8 && xoffset < 8) {
+    if (yoffset) {
+      if (xoffset) {
+        switch (xoffset & 1) {
+          case 0:
+            switch (yoffset & 1) {
+              case 0:
+                Predict16x16Funcs1[0](src, src_stride, dst, dst_stride,
+                                      h_filter, v_filter, 16);
+                break;
+
+              case 1:
+                Predict16x16Funcs1[1](src, src_stride, dst, dst_stride,
+                                      h_filter, v_filter + 1, 16);
+                break;
+            }
+            break;
+
+          case 1:
+            switch (yoffset & 1) {
+              case 0:
+                Predict16x16Funcs1[2](src, src_stride, dst, dst_stride,
+                                      h_filter + 1, v_filter, 16);
+                break;
+
+              case 1:
+                Predict16x16Funcs1[3](src, src_stride, dst, dst_stride,
+                                      h_filter, v_filter + 1, 16);
+                break;
+            }
+            break;
+        }
+      } else {
+        switch (yoffset & 1) {
+          case 0:
+            Predict16x16Funcs2[0](src, src_stride, dst, dst_stride, v_filter,
+                                  16);
+            break;
+
+          case 1:
+            Predict16x16Funcs2[1](src, src_stride, dst, dst_stride,
+                                  v_filter + 1, 16);
+            break;
+        }
+      }
+    } else {
+      switch (xoffset & 1) {
+        case 1:
+          Predict16x16Funcs2[3](src, src_stride, dst, dst_stride, h_filter + 1,
+                                16);
+          break;
+      }
+      switch (xoffset) {
+        case 0: vp8_copy_mem16x16(src, src_stride, dst, dst_stride); break;
+        case 2:
+        case 4:
+        case 6:
+          Predict16x16Funcs2[2](src, src_stride, dst, dst_stride, h_filter, 16);
+          break;
+      }
+    }
+  }
+}
diff --git a/vp8/common/rtcd_defs.pl b/vp8/common/rtcd_defs.pl
index 8452b5e85..40117e367 100644
--- a/vp8/common/rtcd_defs.pl
+++ b/vp8/common/rtcd_defs.pl
@@ -47,13 +47,13 @@ specialize qw/vp8_dequant_idct_add_uv_block sse2 neon dspr2 msa mmi/;
 # Loopfilter
 #
 add_proto qw/void vp8_loop_filter_mbv/, "unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr, int y_stride, int uv_stride, struct loop_filter_info *lfi";
-specialize qw/vp8_loop_filter_mbv sse2 neon dspr2 msa mmi/;
+specialize qw/vp8_loop_filter_mbv sse2 neon dspr2 msa mmi lsx/;
 
 add_proto qw/void vp8_loop_filter_bv/, "unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr, int y_stride, int uv_stride, struct loop_filter_info *lfi";
 specialize qw/vp8_loop_filter_bv sse2 neon dspr2 msa mmi/;
 
 add_proto qw/void vp8_loop_filter_mbh/, "unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr, int y_stride, int uv_stride, struct loop_filter_info *lfi";
-specialize qw/vp8_loop_filter_mbh sse2 neon dspr2 msa mmi/;
+specialize qw/vp8_loop_filter_mbh sse2 neon dspr2 msa mmi lsx/;
 
 add_proto qw/void vp8_loop_filter_bh/, "unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr, int y_stride, int uv_stride, struct loop_filter_info *lfi";
 specialize qw/vp8_loop_filter_bh sse2 neon dspr2 msa mmi/;
@@ -146,10 +146,10 @@ if (vpx_config("CONFIG_POSTPROC") eq "yes") {
 # Subpixel
 #
 add_proto qw/void vp8_sixtap_predict16x16/, "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch";
-specialize qw/vp8_sixtap_predict16x16 sse2 ssse3 neon dspr2 msa mmi/;
+specialize qw/vp8_sixtap_predict16x16 sse2 ssse3 neon dspr2 msa mmi lsx/;
 
 add_proto qw/void vp8_sixtap_predict8x8/, "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch";
-specialize qw/vp8_sixtap_predict8x8 sse2 ssse3 neon dspr2 msa mmi/;
+specialize qw/vp8_sixtap_predict8x8 sse2 ssse3 neon dspr2 msa mmi lsx/;
 
 add_proto qw/void vp8_sixtap_predict8x4/, "unsigned char *src_ptr, int src_pixels_per_line, int xoffset, int yoffset, unsigned char *dst_ptr, int dst_pitch";
 specialize qw/vp8_sixtap_predict8x4 sse2 ssse3 neon dspr2 msa mmi/;
diff --git a/vp8/vp8_common.mk b/vp8/vp8_common.mk
index 286a93a05..909924ce8 100644
--- a/vp8/vp8_common.mk
+++ b/vp8/vp8_common.mk
@@ -124,6 +124,10 @@ ifeq ($(CONFIG_POSTPROC),yes)
 VP8_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/mfqe_msa.c
 endif
 
+# common (loongarch LSX intrinsics)
+VP8_COMMON_SRCS-$(HAVE_LSX) += common/loongarch/loopfilter_filters_lsx.c
+VP8_COMMON_SRCS-$(HAVE_LSX) += common/loongarch/sixtap_filter_lsx.c
+
 # common (neon intrinsics)
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/loopfilter_arm.c
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/loopfilter_arm.h