[FFmpeg-cvslog] avcodec/mips: Split uni mc optimizations to new file

Shivraj Patil git at videolan.org
Thu May 28 17:33:17 CEST 2015


ffmpeg | branch: master | Shivraj Patil <shivraj.patil at imgtec.com> | Thu May 28 20:28:58 2015 +0530| [10b77fbf0d1b4b3c79d41b49a81ce3c07beae8bb] | committer: Michael Niedermayer

avcodec/mips: Split uni mc optimizations to new file

This patch moves HEVC code of uni mc cases to new file hevc_mc_uni_msa.c.
(There are total 5 sub-modules of HEVC mc functions, if we add all these modules in one single file, its size would be huge (~750k) & difficult to maintain, so splitting it in multiple files)
This patch also adds new HEVC header file libavcodec/mips/hevc_macros_msa.h

Signed-off-by: Shivraj Patil <shivraj.patil at imgtec.com>
Signed-off-by: Michael Niedermayer <michaelni at gmx.at>

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=10b77fbf0d1b4b3c79d41b49a81ce3c07beae8bb
---

 libavcodec/mips/Makefile            |    3 +-
 libavcodec/mips/hevc_macros_msa.h   |   51 +
 libavcodec/mips/hevc_mc_uni_msa.c   | 1789 +++++++++++++++++++++++++
 libavcodec/mips/hevcdsp_msa.c       | 2482 ++---------------------------------
 libavutil/mips/generic_macros_msa.h |  217 +++
 5 files changed, 2133 insertions(+), 2409 deletions(-)

diff --git a/libavcodec/mips/Makefile b/libavcodec/mips/Makefile
index 25813e7..9786175 100644
--- a/libavcodec/mips/Makefile
+++ b/libavcodec/mips/Makefile
@@ -20,6 +20,7 @@ MIPSDSPR1-OBJS-$(CONFIG_AAC_ENCODER)      += mips/aaccoder_mips.o
 MIPSFPU-OBJS-$(CONFIG_AAC_ENCODER)        += mips/iirfilter_mips.o
 OBJS-$(CONFIG_HEVC_DECODER)               += mips/hevcdsp_init_mips.o
 OBJS-$(CONFIG_H264DSP)                    += mips/h264dsp_init_mips.o
-MSA-OBJS-$(CONFIG_HEVC_DECODER)           += mips/hevcdsp_msa.o
+MSA-OBJS-$(CONFIG_HEVC_DECODER)           += mips/hevcdsp_msa.o            \
+                                             mips/hevc_mc_uni_msa.o
 MSA-OBJS-$(CONFIG_H264DSP)                += mips/h264dsp_msa.o
 LOONGSON3-OBJS-$(CONFIG_H264DSP)          += mips/h264dsp_mmi.o
diff --git a/libavcodec/mips/hevc_macros_msa.h b/libavcodec/mips/hevc_macros_msa.h
new file mode 100644
index 0000000..7a89f93
--- /dev/null
+++ b/libavcodec/mips/hevc_macros_msa.h
@@ -0,0 +1,51 @@
+/*
+ * Copyright (c) 2015 Manojkumar Bhosale (Manojkumar.Bhosale at imgtec.com)
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_MIPS_HEVC_MACROS_MSA_H
+#define AVCODEC_MIPS_HEVC_MACROS_MSA_H
+
+#define HEVC_PCK_SW_SB2(in0, in1, out)                            \
+{                                                                 \
+    v8i16 tmp0_m;                                                 \
+                                                                  \
+    tmp0_m = __msa_pckev_h((v8i16) in0, (v8i16) in1);             \
+    out = (v4i32) __msa_pckev_b((v16i8) tmp0_m, (v16i8) tmp0_m);  \
+}
+
+#define HEVC_PCK_SW_SB4(in0, in1, in2, in3, out)                  \
+{                                                                 \
+    v8i16 tmp0_m, tmp1_m;                                         \
+                                                                  \
+    PCKEV_H2_SH(in0, in1, in2, in3, tmp0_m, tmp1_m);              \
+    out = (v4i32) __msa_pckev_b((v16i8) tmp1_m, (v16i8) tmp0_m);  \
+}
+
+#define HEVC_FILT_8TAP(in0, in1, in2, in3,                       \
+                       filt0, filt1, filt2, filt3)               \
+( {                                                              \
+    v4i32 out_m;                                                 \
+                                                                 \
+    out_m = __msa_dotp_s_w((v8i16) in0, (v8i16) filt0);          \
+    out_m = __msa_dpadd_s_w(out_m, (v8i16) in1, (v8i16) filt1);  \
+    DPADD_SH2_SW(in2, in3, filt2, filt3, out_m, out_m);          \
+    out_m;                                                       \
+} )
+
+#endif  /* AVCODEC_MIPS_HEVC_MACROS_MSA_H */
diff --git a/libavcodec/mips/hevc_mc_uni_msa.c b/libavcodec/mips/hevc_mc_uni_msa.c
new file mode 100644
index 0000000..09179d2
--- /dev/null
+++ b/libavcodec/mips/hevc_mc_uni_msa.c
@@ -0,0 +1,1789 @@
+/*
+ * Copyright (c) 2015 Manojkumar Bhosale (Manojkumar.Bhosale at imgtec.com)
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/mips/generic_macros_msa.h"
+#include "libavcodec/mips/hevcdsp_mips.h"
+#include "libavcodec/mips/hevc_macros_msa.h"
+
+static void copy_width8_msa(uint8_t *src, int32_t src_stride,
+                            uint8_t *dst, int32_t dst_stride,
+                            int32_t height)
+{
+    int32_t cnt;
+    uint64_t out0, out1, out2, out3, out4, out5, out6, out7;
+    v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
+
+    if (0 == height % 12) {
+        for (cnt = (height / 12); cnt--;) {
+            LD_UB8(src, src_stride,
+                   src0, src1, src2, src3, src4, src5, src6, src7);
+            src += (8 * src_stride);
+
+            out0 = __msa_copy_u_d((v2i64) src0, 0);
+            out1 = __msa_copy_u_d((v2i64) src1, 0);
+            out2 = __msa_copy_u_d((v2i64) src2, 0);
+            out3 = __msa_copy_u_d((v2i64) src3, 0);
+            out4 = __msa_copy_u_d((v2i64) src4, 0);
+            out5 = __msa_copy_u_d((v2i64) src5, 0);
+            out6 = __msa_copy_u_d((v2i64) src6, 0);
+            out7 = __msa_copy_u_d((v2i64) src7, 0);
+
+            SD4(out0, out1, out2, out3, dst, dst_stride);
+            dst += (4 * dst_stride);
+            SD4(out4, out5, out6, out7, dst, dst_stride);
+            dst += (4 * dst_stride);
+
+            LD_UB4(src, src_stride, src0, src1, src2, src3);
+            src += (4 * src_stride);
+
+            out0 = __msa_copy_u_d((v2i64) src0, 0);
+            out1 = __msa_copy_u_d((v2i64) src1, 0);
+            out2 = __msa_copy_u_d((v2i64) src2, 0);
+            out3 = __msa_copy_u_d((v2i64) src3, 0);
+
+            SD4(out0, out1, out2, out3, dst, dst_stride);
+            dst += (4 * dst_stride);
+        }
+    } else if (0 == height % 8) {
+        for (cnt = height >> 3; cnt--;) {
+            LD_UB8(src, src_stride,
+                   src0, src1, src2, src3, src4, src5, src6, src7);
+            src += (8 * src_stride);
+
+            out0 = __msa_copy_u_d((v2i64) src0, 0);
+            out1 = __msa_copy_u_d((v2i64) src1, 0);
+            out2 = __msa_copy_u_d((v2i64) src2, 0);
+            out3 = __msa_copy_u_d((v2i64) src3, 0);
+            out4 = __msa_copy_u_d((v2i64) src4, 0);
+            out5 = __msa_copy_u_d((v2i64) src5, 0);
+            out6 = __msa_copy_u_d((v2i64) src6, 0);
+            out7 = __msa_copy_u_d((v2i64) src7, 0);
+
+            SD4(out0, out1, out2, out3, dst, dst_stride);
+            dst += (4 * dst_stride);
+            SD4(out4, out5, out6, out7, dst, dst_stride);
+            dst += (4 * dst_stride);
+        }
+    } else if (0 == height % 4) {
+        for (cnt = (height / 4); cnt--;) {
+            LD_UB4(src, src_stride, src0, src1, src2, src3);
+            src += (4 * src_stride);
+            out0 = __msa_copy_u_d((v2i64) src0, 0);
+            out1 = __msa_copy_u_d((v2i64) src1, 0);
+            out2 = __msa_copy_u_d((v2i64) src2, 0);
+            out3 = __msa_copy_u_d((v2i64) src3, 0);
+
+            SD4(out0, out1, out2, out3, dst, dst_stride);
+            dst += (4 * dst_stride);
+        }
+    } else if (0 == height % 2) {
+        for (cnt = (height / 2); cnt--;) {
+            LD_UB2(src, src_stride, src0, src1);
+            src += (2 * src_stride);
+            out0 = __msa_copy_u_d((v2i64) src0, 0);
+            out1 = __msa_copy_u_d((v2i64) src1, 0);
+
+            SD(out0, dst);
+            dst += dst_stride;
+            SD(out1, dst);
+            dst += dst_stride;
+        }
+    }
+}
+
+static void copy_width12_msa(uint8_t *src, int32_t src_stride,
+                             uint8_t *dst, int32_t dst_stride,
+                             int32_t height)
+{
+    v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
+
+    LD_UB8(src, src_stride, src0, src1, src2, src3, src4, src5, src6, src7);
+    src += (8 * src_stride);
+    ST12x8_UB(src0, src1, src2, src3, src4, src5, src6, src7, dst, dst_stride);
+    dst += (8 * dst_stride);
+    LD_UB8(src, src_stride, src0, src1, src2, src3, src4, src5, src6, src7);
+    ST12x8_UB(src0, src1, src2, src3, src4, src5, src6, src7, dst, dst_stride);
+}
+
+static void copy_16multx8mult_msa(uint8_t *src, int32_t src_stride,
+                                  uint8_t *dst, int32_t dst_stride,
+                                  int32_t height, int32_t width)
+{
+    int32_t cnt, loop_cnt;
+    uint8_t *src_tmp, *dst_tmp;
+    v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
+
+    for (cnt = (width >> 4); cnt--;) {
+        src_tmp = src;
+        dst_tmp = dst;
+
+        for (loop_cnt = (height >> 3); loop_cnt--;) {
+            LD_UB8(src_tmp, src_stride,
+                   src0, src1, src2, src3, src4, src5, src6, src7);
+            src_tmp += (8 * src_stride);
+
+            ST_UB8(src0, src1, src2, src3, src4, src5, src6, src7,
+                   dst_tmp, dst_stride);
+            dst_tmp += (8 * dst_stride);
+        }
+
+        src += 16;
+        dst += 16;
+    }
+}
+
+static void copy_width16_msa(uint8_t *src, int32_t src_stride,
+                             uint8_t *dst, int32_t dst_stride,
+                             int32_t height)
+{
+    int32_t cnt;
+    v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
+
+    if (0 == height % 12) {
+        for (cnt = (height / 12); cnt--;) {
+            LD_UB8(src, src_stride,
+                   src0, src1, src2, src3, src4, src5, src6, src7);
+            src += (8 * src_stride);
+            ST_UB8(src0, src1, src2, src3, src4, src5, src6, src7,
+                   dst, dst_stride);
+            dst += (8 * dst_stride);
+
+            LD_UB4(src, src_stride, src0, src1, src2, src3);
+            src += (4 * src_stride);
+            ST_UB4(src0, src1, src2, src3, dst, dst_stride);
+            dst += (4 * dst_stride);
+        }
+    } else if (0 == height % 8) {
+        copy_16multx8mult_msa(src, src_stride, dst, dst_stride, height, 16);
+    } else if (0 == height % 4) {
+        for (cnt = (height >> 2); cnt--;) {
+            LD_UB4(src, src_stride, src0, src1, src2, src3);
+            src += (4 * src_stride);
+
+            ST_UB4(src0, src1, src2, src3, dst, dst_stride);
+            dst += (4 * dst_stride);
+        }
+    }
+}
+
+static void copy_width24_msa(uint8_t *src, int32_t src_stride,
+                             uint8_t *dst, int32_t dst_stride,
+                             int32_t height)
+{
+    copy_16multx8mult_msa(src, src_stride, dst, dst_stride, height, 16);
+    copy_width8_msa(src + 16, src_stride, dst + 16, dst_stride, height);
+}
+
+static void copy_width32_msa(uint8_t *src, int32_t src_stride,
+                             uint8_t *dst, int32_t dst_stride,
+                             int32_t height)
+{
+    int32_t cnt;
+    v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
+
+    if (0 == height % 12) {
+        for (cnt = (height / 12); cnt--;) {
+            LD_UB4(src, src_stride, src0, src1, src2, src3);
+            LD_UB4(src + 16, src_stride, src4, src5, src6, src7);
+            src += (4 * src_stride);
+            ST_UB4(src0, src1, src2, src3, dst, dst_stride);
+            ST_UB4(src4, src5, src6, src7, dst + 16, dst_stride);
+            dst += (4 * dst_stride);
+
+            LD_UB4(src, src_stride, src0, src1, src2, src3);
+            LD_UB4(src + 16, src_stride, src4, src5, src6, src7);
+            src += (4 * src_stride);
+            ST_UB4(src0, src1, src2, src3, dst, dst_stride);
+            ST_UB4(src4, src5, src6, src7, dst + 16, dst_stride);
+            dst += (4 * dst_stride);
+
+            LD_UB4(src, src_stride, src0, src1, src2, src3);
+            LD_UB4(src + 16, src_stride, src4, src5, src6, src7);
+            src += (4 * src_stride);
+            ST_UB4(src0, src1, src2, src3, dst, dst_stride);
+            ST_UB4(src4, src5, src6, src7, dst + 16, dst_stride);
+            dst += (4 * dst_stride);
+        }
+    } else if (0 == height % 8) {
+        copy_16multx8mult_msa(src, src_stride, dst, dst_stride, height, 32);
+    } else if (0 == height % 4) {
+        for (cnt = (height >> 2); cnt--;) {
+            LD_UB4(src, src_stride, src0, src1, src2, src3);
+            LD_UB4(src + 16, src_stride, src4, src5, src6, src7);
+            src += (4 * src_stride);
+            ST_UB4(src0, src1, src2, src3, dst, dst_stride);
+            ST_UB4(src4, src5, src6, src7, dst + 16, dst_stride);
+            dst += (4 * dst_stride);
+        }
+    }
+}
+
+static void copy_width48_msa(uint8_t *src, int32_t src_stride,
+                             uint8_t *dst, int32_t dst_stride,
+                             int32_t height)
+{
+    copy_16multx8mult_msa(src, src_stride, dst, dst_stride, height, 48);
+}
+
+static void copy_width64_msa(uint8_t *src, int32_t src_stride,
+                             uint8_t *dst, int32_t dst_stride,
+                             int32_t height)
+{
+    copy_16multx8mult_msa(src, src_stride, dst, dst_stride, height, 64);
+}
+
+uint8_t 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 FILT_8TAP_DPADD_S_H(vec0, vec1, vec2, vec3,             \
+                            filt0, filt1, filt2, filt3)         \
+( {                                                             \
+    v8i16 tmp0, tmp1;                                           \
+                                                                \
+    tmp0 = __msa_dotp_s_h((v16i8) vec0, (v16i8) filt0);         \
+    tmp0 = __msa_dpadd_s_h(tmp0, (v16i8) vec1, (v16i8) filt1);  \
+    tmp1 = __msa_dotp_s_h((v16i8) vec2, (v16i8) filt2);         \
+    tmp1 = __msa_dpadd_s_h(tmp1, (v16i8) vec3, (v16i8) filt3);  \
+    tmp0 = __msa_adds_s_h(tmp0, tmp1);                          \
+                                                                \
+    tmp0;                                                       \
+} )
+
+#define HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3,                  \
+                                   mask0, mask1, mask2, mask3,              \
+                                   filt0, filt1, filt2, filt3,              \
+                                   out0, out1)                              \
+{                                                                           \
+    v16i8 vec0_m, vec1_m, vec2_m, vec3_m,  vec4_m, vec5_m, vec6_m, vec7_m;  \
+    v8i16 res0_m, res1_m, res2_m, res3_m;                                   \
+                                                                            \
+    VSHF_B2_SB(src0, src1, src2, src3, mask0, mask0, vec0_m, vec1_m);       \
+    DOTP_SB2_SH(vec0_m, vec1_m, filt0, filt0, res0_m, res1_m);              \
+    VSHF_B2_SB(src0, src1, src2, src3, mask1, mask1, vec2_m, vec3_m);       \
+    DPADD_SB2_SH(vec2_m, vec3_m, filt1, filt1, res0_m, res1_m);             \
+    VSHF_B2_SB(src0, src1, src2, src3, mask2, mask2, vec4_m, vec5_m);       \
+    DOTP_SB2_SH(vec4_m, vec5_m, filt2, filt2, res2_m, res3_m);              \
+    VSHF_B2_SB(src0, src1, src2, src3, mask3, mask3, vec6_m, vec7_m);       \
+    DPADD_SB2_SH(vec6_m, vec7_m, filt3, filt3, res2_m, res3_m);             \
+    ADDS_SH2_SH(res0_m, res2_m, res1_m, res3_m, out0, out1);                \
+}
+
+#define HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3,                    \
+                                   mask0, mask1, mask2, mask3,                \
+                                   filt0, filt1, filt2, filt3,                \
+                                   out0, out1, out2, out3)                    \
+{                                                                             \
+    v16i8 vec0_m, vec1_m, vec2_m, vec3_m, vec4_m, vec5_m, vec6_m, vec7_m;     \
+    v8i16 res0_m, res1_m, res2_m, res3_m, res4_m, res5_m, res6_m, res7_m;     \
+                                                                              \
+    VSHF_B2_SB(src0, src0, src1, src1, mask0, mask0, vec0_m, vec1_m);         \
+    VSHF_B2_SB(src2, src2, src3, src3, mask0, mask0, vec2_m, vec3_m);         \
+    DOTP_SB4_SH(vec0_m, vec1_m, vec2_m, vec3_m, filt0, filt0, filt0, filt0,   \
+                res0_m, res1_m, res2_m, res3_m);                              \
+    VSHF_B2_SB(src0, src0, src1, src1, mask2, mask2, vec0_m, vec1_m);         \
+    VSHF_B2_SB(src2, src2, src3, src3, mask2, mask2, vec2_m, vec3_m);         \
+    DOTP_SB4_SH(vec0_m, vec1_m, vec2_m, vec3_m, filt2, filt2, filt2, filt2,   \
+                res4_m, res5_m, res6_m, res7_m);                              \
+    VSHF_B2_SB(src0, src0, src1, src1, mask1, mask1, vec4_m, vec5_m);         \
+    VSHF_B2_SB(src2, src2, src3, src3, mask1, mask1, vec6_m, vec7_m);         \
+    DPADD_SB4_SH(vec4_m, vec5_m, vec6_m, vec7_m, filt1, filt1, filt1, filt1,  \
+                 res0_m, res1_m, res2_m, res3_m);                             \
+    VSHF_B2_SB(src0, src0, src1, src1, mask3, mask3, vec4_m, vec5_m);         \
+    VSHF_B2_SB(src2, src2, src3, src3, mask3, mask3, vec6_m, vec7_m);         \
+    DPADD_SB4_SH(vec4_m, vec5_m, vec6_m, vec7_m, filt3, filt3, filt3, filt3,  \
+                 res4_m, res5_m, res6_m, res7_m);                             \
+    ADDS_SH4_SH(res0_m, res4_m, res1_m, res5_m, res2_m, res6_m, res3_m,       \
+                res7_m, out0, out1, out2, out3);                              \
+}
+
+static void common_hz_8t_4x4_msa(uint8_t *src, int32_t src_stride,
+                                 uint8_t *dst, int32_t dst_stride,
+                                 const int8_t *filter, uint8_t rnd_val)
+{
+    v16u8 mask0, mask1, mask2, mask3, out;
+    v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
+    v8i16 filt, out0, out1;
+    v8i16 rnd_vec;
+
+    mask0 = LD_UB(&mc_filt_mask_arr[16]);
+    src -= 3;
+    rnd_vec = __msa_fill_h(rnd_val);
+
+    /* rearranging filter */
+    filt = LD_SH(filter);
+    SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    mask1 = mask0 + 2;
+    mask2 = mask0 + 4;
+    mask3 = mask0 + 6;
+
+    LD_SB4(src, src_stride, src0, src1, src2, src3);
+    XORI_B4_128_SB(src0, src1, src2, src3);
+    HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
+                               mask3, filt0, filt1, filt2, filt3, out0, out1);
+    SRAR_H2_SH(out0, out1, rnd_vec);
+    SAT_SH2_SH(out0, out1, 7);
+    out = PCKEV_XORI128_UB(out0, out1);
+    ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
+}
+
+static void common_hz_8t_4x8_msa(uint8_t *src, int32_t src_stride,
+                                 uint8_t *dst, int32_t dst_stride,
+                                 const int8_t *filter, uint8_t rnd_val)
+{
+    v16i8 filt0, filt1, filt2, filt3;
+    v16i8 src0, src1, src2, src3;
+    v16u8 mask0, mask1, mask2, mask3, out;
+    v8i16 filt, out0, out1, out2, out3;
+    v8i16 rnd_vec;
+
+    mask0 = LD_UB(&mc_filt_mask_arr[16]);
+    src -= 3;
+    rnd_vec = __msa_fill_h(rnd_val);
+
+    /* rearranging filter */
+    filt = LD_SH(filter);
+    SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    mask1 = mask0 + 2;
+    mask2 = mask0 + 4;
+    mask3 = mask0 + 6;
+
+    LD_SB4(src, src_stride, src0, src1, src2, src3);
+    XORI_B4_128_SB(src0, src1, src2, src3);
+    src += (4 * src_stride);
+    HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
+                               mask3, filt0, filt1, filt2, filt3, out0, out1);
+    LD_SB4(src, src_stride, src0, src1, src2, src3);
+    XORI_B4_128_SB(src0, src1, src2, src3);
+    HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
+                               mask3, filt0, filt1, filt2, filt3, out2, out3);
+    SRAR_H4_SH(out0, out1, out2, out3, rnd_vec);
+    SAT_SH4_SH(out0, out1, out2, out3, 7);
+    out = PCKEV_XORI128_UB(out0, out1);
+    ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
+    dst += (4 * dst_stride);
+    out = PCKEV_XORI128_UB(out2, out3);
+    ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
+}
+
+static void common_hz_8t_4x16_msa(uint8_t *src, int32_t src_stride,
+                                  uint8_t *dst, int32_t dst_stride,
+                                  const int8_t *filter, uint8_t rnd_val)
+{
+    v16u8 mask0, mask1, mask2, mask3, out;
+    v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
+    v8i16 filt, out0, out1, out2, out3;
+    v8i16 rnd_vec;
+
+    mask0 = LD_UB(&mc_filt_mask_arr[16]);
+    src -= 3;
+    rnd_vec = __msa_fill_h(rnd_val);
+
+    /* rearranging filter */
+    filt = LD_SH(filter);
+    SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    mask1 = mask0 + 2;
+    mask2 = mask0 + 4;
+    mask3 = mask0 + 6;
+
+    LD_SB4(src, src_stride, src0, src1, src2, src3);
+    XORI_B4_128_SB(src0, src1, src2, src3);
+    src += (4 * src_stride);
+    HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
+                               mask3, filt0, filt1, filt2, filt3, out0, out1);
+    LD_SB4(src, src_stride, src0, src1, src2, src3);
+    XORI_B4_128_SB(src0, src1, src2, src3);
+    src += (4 * src_stride);
+    HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
+                               mask3, filt0, filt1, filt2, filt3, out2, out3);
+    SRAR_H4_SH(out0, out1, out2, out3, rnd_vec);
+    SAT_SH4_SH(out0, out1, out2, out3, 7);
+    out = PCKEV_XORI128_UB(out0, out1);
+    ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
+    dst += (4 * dst_stride);
+    out = PCKEV_XORI128_UB(out2, out3);
+    ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
+    dst += (4 * dst_stride);
+
+    LD_SB4(src, src_stride, src0, src1, src2, src3);
+    XORI_B4_128_SB(src0, src1, src2, src3);
+    src += (4 * src_stride);
+    HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
+                               mask3, filt0, filt1, filt2, filt3, out0, out1);
+    LD_SB4(src, src_stride, src0, src1, src2, src3);
+    XORI_B4_128_SB(src0, src1, src2, src3);
+    src += (4 * src_stride);
+    HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
+                               mask3, filt0, filt1, filt2, filt3, out2, out3);
+
+    SRAR_H4_SH(out0, out1, out2, out3, rnd_vec);
+    SAT_SH4_SH(out0, out1, out2, out3, 7);
+    out = PCKEV_XORI128_UB(out0, out1);
+    ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
+    dst += (4 * dst_stride);
+    out = PCKEV_XORI128_UB(out2, out3);
+    ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
+}
+
+static void common_hz_8t_4w_msa(uint8_t *src, int32_t src_stride,
+                                uint8_t *dst, int32_t dst_stride,
+                                const int8_t *filter, int32_t height, uint8_t rnd_val)
+{
+    if (4 == height) {
+        common_hz_8t_4x4_msa(src, src_stride, dst, dst_stride, filter, rnd_val);
+    } else if (8 == height) {
+        common_hz_8t_4x8_msa(src, src_stride, dst, dst_stride, filter, rnd_val);
+    } else if (16 == height) {
+        common_hz_8t_4x16_msa(src, src_stride, dst, dst_stride, filter,
+                              rnd_val);
+    }
+}
+
+static void common_hz_8t_8x4_msa(uint8_t *src, int32_t src_stride,
+                                 uint8_t *dst, int32_t dst_stride,
+                                 const int8_t *filter, uint8_t rnd_val)
+{
+    v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
+    v16u8 mask0, mask1, mask2, mask3, tmp0, tmp1;
+    v8i16 filt, out0, out1, out2, out3;
+    v8i16 rnd_vec;
+
+    mask0 = LD_UB(&mc_filt_mask_arr[0]);
+    src -= 3;
+    rnd_vec = __msa_fill_h(rnd_val);
+
+    /* rearranging filter */
+    filt = LD_SH(filter);
+    SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    mask1 = mask0 + 2;
+    mask2 = mask0 + 4;
+    mask3 = mask0 + 6;
+
+    LD_SB4(src, src_stride, src0, src1, src2, src3);
+    XORI_B4_128_SB(src0, src1, src2, src3);
+    HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
+                               mask3, filt0, filt1, filt2, filt3, out0, out1,
+                               out2, out3);
+    SRAR_H4_SH(out0, out1, out2, out3, rnd_vec);
+    SAT_SH4_SH(out0, out1, out2, out3, 7);
+    tmp0 = PCKEV_XORI128_UB(out0, out1);
+    tmp1 = PCKEV_XORI128_UB(out2, out3);
+    ST8x4_UB(tmp0, tmp1, dst, dst_stride);
+}
+
+static void common_hz_8t_8x8mult_msa(uint8_t *src, int32_t src_stride,
+                                     uint8_t *dst, int32_t dst_stride,
+                                     const int8_t *filter, int32_t height,
+                                     uint8_t rnd_val)
+{
+    uint32_t loop_cnt;
+    v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
+    v16u8 mask0, mask1, mask2, mask3, tmp0, tmp1;
+    v8i16 filt, out0, out1, out2, out3;
+    v8i16 rnd_vec;
+
+    mask0 = LD_UB(&mc_filt_mask_arr[0]);
+    src -= 3;
+    rnd_vec = __msa_fill_h(rnd_val);
+
+    /* rearranging filter */
+    filt = LD_SH(filter);
+    SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    mask1 = mask0 + 2;
+    mask2 = mask0 + 4;
+    mask3 = mask0 + 6;
+
+    for (loop_cnt = (height >> 2); loop_cnt--;) {
+        LD_SB4(src, src_stride, src0, src1, src2, src3);
+        XORI_B4_128_SB(src0, src1, src2, src3);
+        src += (4 * src_stride);
+        HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
+                                   mask3, filt0, filt1, filt2, filt3, out0,
+                                   out1, out2, out3);
+        SRAR_H4_SH(out0, out1, out2, out3, rnd_vec);
+        SAT_SH4_SH(out0, out1, out2, out3, 7);
+        tmp0 = PCKEV_XORI128_UB(out0, out1);
+        tmp1 = PCKEV_XORI128_UB(out2, out3);
+        ST8x4_UB(tmp0, tmp1, dst, dst_stride);
+        dst += (4 * dst_stride);
+    }
+}
+
+static void common_hz_8t_8w_msa(uint8_t *src, int32_t src_stride,
+                                uint8_t *dst, int32_t dst_stride,
+                                const int8_t *filter, int32_t height,
+                                uint8_t rnd_val)
+{
+    if (4 == height) {
+        common_hz_8t_8x4_msa(src, src_stride, dst, dst_stride, filter, rnd_val);
+    } else {
+        common_hz_8t_8x8mult_msa(src, src_stride, dst, dst_stride, filter,
+                                 height, rnd_val);
+    }
+}
+
+static void common_hz_8t_12w_msa(uint8_t *src, int32_t src_stride,
+                                 uint8_t *dst, int32_t dst_stride,
+                                 const int8_t *filter, int32_t height,
+                                 uint8_t rnd_val)
+{
+    uint8_t *src1_ptr, *dst1;
+    uint32_t loop_cnt;
+    v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
+    v8i16 filt, out0, out1, out2, out3;
+    v16u8 mask0, mask1, mask2, mask3, mask4, mask5, mask6, mask00, tmp0, tmp1;
+    v8i16 rnd_vec;
+
+    mask00 = LD_UB(&mc_filt_mask_arr[0]);
+    mask0 = LD_UB(&mc_filt_mask_arr[16]);
+    rnd_vec = __msa_fill_h(rnd_val);
+
+    src1_ptr = src - 3;
+    dst1 = dst;
+
+    dst = dst1 + 8;
+    src = src1_ptr + 8;
+
+    /* rearranging filter */
+    filt = LD_SH(filter);
+    SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    mask1 = mask00 + 2;
+    mask2 = mask00 + 4;
+    mask3 = mask00 + 6;
+    mask4 = mask0 + 2;
+    mask5 = mask0 + 4;
+    mask6 = mask0 + 6;
+
+    for (loop_cnt = (height >> 2); loop_cnt--;) {
+        /* 8 width */
+        LD_SB4(src1_ptr, src_stride, src0, src1, src2, src3);
+        XORI_B4_128_SB(src0, src1, src2, src3);
+        src1_ptr += (4 * src_stride);
+        HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask00, mask1, mask2,
+                                   mask3, filt0, filt1, filt2, filt3, out0,
+                                   out1, out2, out3);
+        SRAR_H4_SH(out0, out1, out2, out3, rnd_vec);
+        SAT_SH4_SH(out0, out1, out2, out3, 7);
+        tmp0 = PCKEV_XORI128_UB(out0, out1);
+        tmp1 = PCKEV_XORI128_UB(out2, out3);
+        ST8x4_UB(tmp0, tmp1, dst1, dst_stride);
+        dst1 += (4 * dst_stride);
+
+        /* 4 width */
+        LD_SB4(src, src_stride, src0, src1, src2, src3);
+        XORI_B4_128_SB(src0, src1, src2, src3);
+        src += (4 * src_stride);
+        HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask4, mask5,
+                                   mask6, filt0, filt1, filt2, filt3, out0,
+                                   out1);
+        SRAR_H2_SH(out0, out1, rnd_vec);
+        SAT_SH2_SH(out0, out1, 7);
+        tmp0 = PCKEV_XORI128_UB(out0, out1);
+        ST4x4_UB(tmp0, tmp0, 0, 1, 2, 3, dst, dst_stride);
+        dst += (4 * dst_stride);
+    }
+}
+
+static void common_hz_8t_16w_msa(uint8_t *src, int32_t src_stride,
+                                 uint8_t *dst, int32_t dst_stride,
+                                 const int8_t *filter, int32_t height,
+                                 uint8_t rnd_val)
+{
+    uint32_t loop_cnt;
+    v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
+    v16u8 mask0, mask1, mask2, mask3, out;
+    v8i16 filt, out0, out1, out2, out3;
+    v8i16 rnd_vec;
+
+    mask0 = LD_UB(&mc_filt_mask_arr[0]);
+    src -= 3;
+    rnd_vec = __msa_fill_h(rnd_val);
+
+    /* rearranging filter */
+    filt = LD_SH(filter);
+    SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    mask1 = mask0 + 2;
+    mask2 = mask0 + 4;
+    mask3 = mask0 + 6;
+
+    for (loop_cnt = (height >> 1); loop_cnt--;) {
+        LD_SB2(src, src_stride, src0, src2);
+        LD_SB2(src + 8, src_stride, src1, src3);
+        XORI_B4_128_SB(src0, src1, src2, src3);
+        src += (2 * src_stride);
+        HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
+                                   mask3, filt0, filt1, filt2, filt3, out0,
+                                   out1, out2, out3);
+        SRAR_H4_SH(out0, out1, out2, out3, rnd_vec);
+        SAT_SH4_SH(out0, out1, out2, out3, 7);
+        out = PCKEV_XORI128_UB(out0, out1);
+        ST_UB(out, dst);
+        dst += dst_stride;
+        out = PCKEV_XORI128_UB(out2, out3);
+        ST_UB(out, dst);
+        dst += dst_stride;
+    }
+}
+
+static void common_hz_8t_24w_msa(uint8_t *src, int32_t src_stride,
+                                 uint8_t *dst, int32_t dst_stride,
+                                 const int8_t *filter, int32_t height,
+                                 uint8_t rnd_val)
+{
+    uint32_t loop_cnt;
+    v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
+    v16u8 mask0, mask1, mask2, mask3, mask4, mask5, mask6, mask7, out;
+    v16i8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, vec8, vec9, vec10;
+    v16i8 vec11;
+    v8i16 out0, out1, out2, out3, out4, out5, out6, out7, out8, out9, out10;
+    v8i16 out11, filt;
+    v8i16 rnd_vec;
+
+    mask0 = LD_UB(&mc_filt_mask_arr[0]);
+    src -= 3;
+    rnd_vec = __msa_fill_h(rnd_val);
+
+    /* rearranging filter */
+    filt = LD_SH(filter);
+    SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    mask1 = mask0 + 2;
+    mask2 = mask0 + 4;
+    mask3 = mask0 + 6;
+    mask4 = mask0 + 8;
+    mask5 = mask0 + 10;
+    mask6 = mask0 + 12;
+    mask7 = mask0 + 14;
+
+    for (loop_cnt = (height >> 1); loop_cnt--;) {
+        LD_SB2(src, src_stride, src0, src2);
+        LD_SB2(src + 16, src_stride, src1, src3);
+        XORI_B4_128_SB(src0, src1, src2, src3);
+        src += (2 * src_stride);
+        VSHF_B2_SB(src0, src0, src1, src1, mask0, mask0, vec0, vec8);
+        VSHF_B2_SB(src2, src2, src3, src3, mask0, mask0, vec2, vec9);
+        VSHF_B2_SB(src0, src1, src2, src3, mask4, mask4, vec1, vec3);
+        DOTP_SB4_SH(vec0, vec8, vec2, vec9, filt0, filt0, filt0, filt0, out0,
+                    out8, out2, out9);
+        DOTP_SB2_SH(vec1, vec3, filt0, filt0, out1, out3);
+        VSHF_B2_SB(src0, src0, src1, src1, mask2, mask2, vec0, vec8);
+        VSHF_B2_SB(src2, src2, src3, src3, mask2, mask2, vec2, vec9);
+        VSHF_B2_SB(src0, src1, src2, src3, mask6, mask6, vec1, vec3);
+        DOTP_SB4_SH(vec0, vec8, vec2, vec9, filt2, filt2, filt2, filt2, out4,
+                    out10, out6, out11);
+        DOTP_SB2_SH(vec1, vec3, filt2, filt2, out5, out7);
+        VSHF_B2_SB(src0, src0, src1, src1, mask1, mask1, vec4, vec10);
+        VSHF_B2_SB(src2, src2, src3, src3, mask1, mask1, vec6, vec11);
+        VSHF_B2_SB(src0, src1, src2, src3, mask5, mask5, vec5, vec7);
+        DPADD_SB4_SH(vec4, vec10, vec6, vec11, filt1, filt1, filt1, filt1,
+                     out0, out8, out2, out9);
+        DPADD_SB2_SH(vec5, vec7, filt1, filt1, out1, out3);
+        VSHF_B2_SB(src0, src0, src1, src1, mask3, mask3, vec4, vec10);
+        VSHF_B2_SB(src2, src2, src3, src3, mask3, mask3, vec6, vec11);
+        VSHF_B2_SB(src0, src1, src2, src3, mask7, mask7, vec5, vec7);
+        DPADD_SB4_SH(vec4, vec10, vec6, vec11, filt3, filt3, filt3, filt3,
+                     out4, out10, out6, out11);
+        DPADD_SB2_SH(vec5, vec7, filt3, filt3, out5, out7);
+        ADDS_SH4_SH(out0, out4, out8, out10, out2, out6, out9, out11, out0,
+                    out8, out2, out9);
+        ADDS_SH2_SH(out1, out5, out3, out7, out1, out3);
+        SRAR_H4_SH(out0, out8, out2, out9, rnd_vec);
+        SRAR_H2_SH(out1, out3, rnd_vec);
+        SAT_SH4_SH(out0, out8, out2, out9, 7);
+        SAT_SH2_SH(out1, out3, 7);
+        out = PCKEV_XORI128_UB(out8, out9);
+        ST8x2_UB(out, dst + 16, dst_stride);
+        out = PCKEV_XORI128_UB(out0, out1);
+        ST_UB(out, dst);
+        dst += dst_stride;
+        out = PCKEV_XORI128_UB(out2, out3);
+        ST_UB(out, dst);
+        dst += dst_stride;
+    }
+}
+
+static void common_hz_8t_32w_msa(uint8_t *src, int32_t src_stride,
+                                 uint8_t *dst, int32_t dst_stride,
+                                 const int8_t *filter, int32_t height,
+                                 uint8_t rnd_val)
+{
+    uint32_t loop_cnt;
+    v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
+    v16u8 mask0, mask1, mask2, mask3, out;
+    v8i16 filt, out0, out1, out2, out3;
+    v8i16 rnd_vec;
+
+    mask0 = LD_UB(&mc_filt_mask_arr[0]);
+    src -= 3;
+    rnd_vec = __msa_fill_h(rnd_val);
+
+    /* rearranging filter */
+    filt = LD_SH(filter);
+    SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    mask1 = mask0 + 2;
+    mask2 = mask0 + 4;
+    mask3 = mask0 + 6;
+
+    for (loop_cnt = (height >> 1); loop_cnt--;) {
+        src0 = LD_SB(src);
+        src2 = LD_SB(src + 16);
+        src3 = LD_SB(src + 24);
+        src1 = __msa_sldi_b(src2, src0, 8);
+        src += src_stride;
+        XORI_B4_128_SB(src0, src1, src2, src3);
+        HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
+                                   mask3, filt0, filt1, filt2, filt3, out0,
+                                   out1, out2, out3);
+        SRAR_H4_SH(out0, out1, out2, out3, rnd_vec);
+        SAT_SH4_SH(out0, out1, out2, out3, 7);
+
+        src0 = LD_SB(src);
+        src2 = LD_SB(src + 16);
+        src3 = LD_SB(src + 24);
+        src1 = __msa_sldi_b(src2, src0, 8);
+        src += src_stride;
+
+        out = PCKEV_XORI128_UB(out0, out1);
+        ST_UB(out, dst);
+        out = PCKEV_XORI128_UB(out2, out3);
+        ST_UB(out, dst + 16);
+        dst += dst_stride;
+
+        XORI_B4_128_SB(src0, src1, src2, src3);
+        HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
+                                   mask3, filt0, filt1, filt2, filt3, out0,
+                                   out1, out2, out3);
+        SRAR_H4_SH(out0, out1, out2, out3, rnd_vec);
+        SAT_SH4_SH(out0, out1, out2, out3, 7);
+        out = PCKEV_XORI128_UB(out0, out1);
+        ST_UB(out, dst);
+        out = PCKEV_XORI128_UB(out2, out3);
+        ST_UB(out, dst + 16);
+        dst += dst_stride;
+    }
+}
+
+static void common_hz_8t_48w_msa(uint8_t *src, int32_t src_stride,
+                                 uint8_t *dst, int32_t dst_stride,
+                                 const int8_t *filter, int32_t height,
+                                 uint8_t rnd_val)
+{
+    uint32_t loop_cnt;
+    v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3, vec0, vec1, vec2;
+    v16u8 mask0, mask1, mask2, mask3, mask4, mask5, mask6, mask7, out;
+    v8i16 filt, out0, out1, out2, out3, out4, out5, out6;
+    v8i16 rnd_vec;
+
+    mask0 = LD_UB(&mc_filt_mask_arr[0]);
+    src -= 3;
+    rnd_vec = __msa_fill_h(rnd_val);
+
+    /* rearranging filter */
+    filt = LD_SH(filter);
+    SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    mask1 = mask0 + 2;
+    mask2 = mask0 + 4;
+    mask3 = mask0 + 6;
+    mask4 = mask0 + 8;
+    mask5 = mask0 + 10;
+    mask6 = mask0 + 12;
+    mask7 = mask0 + 14;
+
+    for (loop_cnt = height; loop_cnt--;) {
+        LD_SB3(src, 16, src0, src2, src3);
+        src1 = __msa_sldi_b(src2, src0, 8);
+
+        XORI_B4_128_SB(src0, src1, src2, src3);
+        VSHF_B3_SB(src0, src0, src1, src1, src2, src2, mask0, mask0, mask0,
+                   vec0, vec1, vec2);
+        DOTP_SB3_SH(vec0, vec1, vec2, filt0, filt0, filt0, out0, out1, out2);
+        VSHF_B3_SB(src0, src0, src1, src1, src2, src2, mask1, mask1, mask1,
+                   vec0, vec1, vec2);
+        DPADD_SB2_SH(vec0, vec1, filt1, filt1, out0, out1);
+        out2 = __msa_dpadd_s_h(out2, vec2, filt1);
+        VSHF_B3_SB(src0, src0, src1, src1, src2, src2, mask2, mask2, mask2,
+                   vec0, vec1, vec2);
+        DOTP_SB3_SH(vec0, vec1, vec2, filt2, filt2, filt2, out3, out4, out5);
+        VSHF_B3_SB(src0, src0, src1, src1, src2, src2, mask3, mask3, mask3,
+                   vec0, vec1, vec2);
+        DPADD_SB2_SH(vec0, vec1, filt3, filt3, out3, out4);
+        out5 = __msa_dpadd_s_h(out5, vec2, filt3);
+        ADDS_SH2_SH(out0, out3, out1, out4, out0, out1);
+        out2 = __msa_adds_s_h(out2, out5);
+        SRAR_H2_SH(out0, out1, rnd_vec);
+        out6 = __msa_srar_h(out2, rnd_vec);
+        SAT_SH3_SH(out0, out1, out6, 7);
+        out = PCKEV_XORI128_UB(out0, out1);
+        ST_UB(out, dst);
+
+        src1 = LD_SB(src + 40);
+        src += src_stride;
+        src1 = (v16i8) __msa_xori_b((v16u8) src1, 128);
+
+        VSHF_B3_SB(src2, src3, src3, src3, src1, src1, mask4, mask0, mask0,
+                   vec0, vec1, vec2);
+        DOTP_SB3_SH(vec0, vec1, vec2, filt0, filt0, filt0, out0, out1, out2);
+        VSHF_B3_SB(src2, src3, src3, src3, src1, src1, mask5, mask1, mask1,
+                   vec0, vec1, vec2);
+        DPADD_SB2_SH(vec0, vec1, filt1, filt1, out0, out1);
+        out2 = __msa_dpadd_s_h(out2, vec2, filt1);
+        VSHF_B3_SB(src2, src3, src3, src3, src1, src1, mask6, mask2, mask2,
+                   vec0, vec1, vec2);
+        DOTP_SB3_SH(vec0, vec1, vec2, filt2, filt2, filt2, out3, out4, out5);
+        VSHF_B3_SB(src2, src3, src3, src3, src1, src1, mask7, mask3, mask3,
+                   vec0, vec1, vec2);
+        DPADD_SB2_SH(vec0, vec1, filt3, filt3, out3, out4);
+        out5 = __msa_dpadd_s_h(out5, vec2, filt3);
+        ADDS_SH2_SH(out0, out3, out1, out4, out3, out4);
+        out5 = __msa_adds_s_h(out2, out5);
+        SRAR_H3_SH(out3, out4, out5, rnd_vec);
+        SAT_SH3_SH(out3, out4, out5, 7);
+        out = PCKEV_XORI128_UB(out6, out3);
+        ST_UB(out, dst + 16);
+        out = PCKEV_XORI128_UB(out4, out5);
+        ST_UB(out, dst + 32);
+        dst += dst_stride;
+    }
+}
+
+static void common_hz_8t_64w_msa(uint8_t *src, int32_t src_stride,
+                                 uint8_t *dst, int32_t dst_stride,
+                                 const int8_t *filter, int32_t height,
+                                 uint8_t rnd_val)
+{
+    int32_t loop_cnt;
+    v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
+    v16u8 mask0, mask1, mask2, mask3, out;
+    v8i16 filt, out0, out1, out2, out3;
+    v8i16 rnd_vec;
+
+    mask0 = LD_UB(&mc_filt_mask_arr[0]);
+    src -= 3;
+    rnd_vec = __msa_fill_h(rnd_val);
+
+    /* rearranging filter */
+    filt = LD_SH(filter);
+    SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    mask1 = mask0 + 2;
+    mask2 = mask0 + 4;
+    mask3 = mask0 + 6;
+
+    for (loop_cnt = height; loop_cnt--;) {
+        src0 = LD_SB(src);
+        src2 = LD_SB(src + 16);
+        src3 = LD_SB(src + 24);
+        src1 = __msa_sldi_b(src2, src0, 8);
+
+        XORI_B4_128_SB(src0, src1, src2, src3);
+        HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1,
+                                   mask2, mask3, filt0, filt1, filt2, filt3,
+                                   out0, out1, out2, out3);
+        SRAR_H4_SH(out0, out1, out2, out3, rnd_vec);
+        SAT_SH4_SH(out0, out1, out2, out3, 7);
+        out = PCKEV_XORI128_UB(out0, out1);
+        ST_UB(out, dst);
+        out = PCKEV_XORI128_UB(out2, out3);
+        ST_UB(out, dst + 16);
+
+        src0 = LD_SB(src + 32);
+        src2 = LD_SB(src + 48);
+        src3 = LD_SB(src + 56);
+        src1 = __msa_sldi_b(src2, src0, 8);
+        src += src_stride;
+
+        XORI_B4_128_SB(src0, src1, src2, src3);
+        HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1,
+                                   mask2, mask3, filt0, filt1, filt2, filt3,
+                                   out0, out1, out2, out3);
+        SRAR_H4_SH(out0, out1, out2, out3, rnd_vec);
+        SAT_SH4_SH(out0, out1, out2, out3, 7);
+        out = PCKEV_XORI128_UB(out0, out1);
+        ST_UB(out, dst + 32);
+        out = PCKEV_XORI128_UB(out2, out3);
+        ST_UB(out, dst + 48);
+        dst += dst_stride;
+    }
+}
+
+static void common_vt_8t_4w_msa(uint8_t *src, int32_t src_stride,
+                                uint8_t *dst, int32_t dst_stride,
+                                const int8_t *filter, int32_t height,
+                                uint8_t rnd_val)
+{
+    uint32_t loop_cnt;
+    v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
+    v16i8 src10_r, src32_r, src54_r, src76_r, src98_r, src21_r, src43_r;
+    v16i8 src65_r, src87_r, src109_r, src2110, src4332, src6554, src8776;
+    v16i8 src10998, filt0, filt1, filt2, filt3;
+    v16u8 out;
+    v8i16 filt, out10, out32;
+    v8i16 rnd_vec;
+
+    src -= (3 * src_stride);
+    rnd_vec = __msa_fill_h(rnd_val);
+
+    filt = LD_SH(filter);
+    SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    LD_SB7(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
+    src += (7 * src_stride);
+
+    ILVR_B4_SB(src1, src0, src3, src2, src5, src4, src2, src1, src10_r, src32_r,
+               src54_r, src21_r);
+    ILVR_B2_SB(src4, src3, src6, src5, src43_r, src65_r);
+    ILVR_D3_SB(src21_r, src10_r, src43_r, src32_r, src65_r, src54_r, src2110,
+               src4332, src6554);
+    XORI_B3_128_SB(src2110, src4332, src6554);
+
+    for (loop_cnt = (height >> 2); loop_cnt--;) {
+        LD_SB4(src, src_stride, src7, src8, src9, src10);
+        src += (4 * src_stride);
+
+        ILVR_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_r,
+                   src87_r, src98_r, src109_r);
+        ILVR_D2_SB(src87_r, src76_r, src109_r, src98_r, src8776, src10998);
+        XORI_B2_128_SB(src8776, src10998);
+        out10 = FILT_8TAP_DPADD_S_H(src2110, src4332, src6554, src8776, filt0,
+                                    filt1, filt2, filt3);
+        out32 = FILT_8TAP_DPADD_S_H(src4332, src6554, src8776, src10998, filt0,
+                                    filt1, filt2, filt3);
+        SRAR_H2_SH(out10, out32, rnd_vec);
+        SAT_SH2_SH(out10, out32, 7);
+        out = PCKEV_XORI128_UB(out10, out32);
+        ST4x4_UB(out, out, 0, 1, 2, 3, dst, dst_stride);
+        dst += (4 * dst_stride);
+
+        src2110 = src6554;
+        src4332 = src8776;
+        src6554 = src10998;
+        src6 = src10;
+    }
+}
+
+static void common_vt_8t_8w_msa(uint8_t *src, int32_t src_stride,
+                                uint8_t *dst, int32_t dst_stride,
+                                const int8_t *filter, int32_t height,
+                                uint8_t rnd_val)
+{
+    uint32_t loop_cnt;
+    v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
+    v16i8 src10_r, src32_r, src54_r, src76_r, src98_r, src21_r, src43_r;
+    v16i8 src65_r, src87_r, src109_r, filt0, filt1, filt2, filt3;
+    v16u8 tmp0, tmp1;
+    v8i16 filt, out0_r, out1_r, out2_r, out3_r;
+    v8i16 rnd_vec;
+
+    src -= (3 * src_stride);
+    rnd_vec = __msa_fill_h(rnd_val);
+
+    filt = LD_SH(filter);
+    SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    LD_SB7(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
+    XORI_B7_128_SB(src0, src1, src2, src3, src4, src5, src6);
+    src += (7 * src_stride);
+    ILVR_B4_SB(src1, src0, src3, src2, src5, src4, src2, src1, src10_r, src32_r,
+               src54_r, src21_r);
+    ILVR_B2_SB(src4, src3, src6, src5, src43_r, src65_r);
+
+    for (loop_cnt = (height >> 2); loop_cnt--;) {
+        LD_SB4(src, src_stride, src7, src8, src9, src10);
+        XORI_B4_128_SB(src7, src8, src9, src10);
+        src += (4 * src_stride);
+
+        ILVR_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_r,
+                   src87_r, src98_r, src109_r);
+        out0_r = FILT_8TAP_DPADD_S_H(src10_r, src32_r, src54_r, src76_r, filt0,
+                                     filt1, filt2, filt3);
+        out1_r = FILT_8TAP_DPADD_S_H(src21_r, src43_r, src65_r, src87_r, filt0,
+                                     filt1, filt2, filt3);
+        out2_r = FILT_8TAP_DPADD_S_H(src32_r, src54_r, src76_r, src98_r, filt0,
+                                     filt1, filt2, filt3);
+        out3_r = FILT_8TAP_DPADD_S_H(src43_r, src65_r, src87_r, src109_r, filt0,
+                                     filt1, filt2, filt3);
+        SRAR_H4_SH(out0_r, out1_r, out2_r, out3_r, rnd_vec);
+        SAT_SH4_SH(out0_r, out1_r, out2_r, out3_r, 7);
+        tmp0 = PCKEV_XORI128_UB(out0_r, out1_r);
+        tmp1 = PCKEV_XORI128_UB(out2_r, out3_r);
+        ST8x4_UB(tmp0, tmp1, dst, dst_stride);
+        dst += (4 * dst_stride);
+
+        src10_r = src54_r;
+        src32_r = src76_r;
+        src54_r = src98_r;
+        src21_r = src65_r;
+        src43_r = src87_r;
+        src65_r = src109_r;
+        src6 = src10;
+    }
+}
+
+static void common_vt_8t_12w_msa(uint8_t *src, int32_t src_stride,
+                                 uint8_t *dst, int32_t dst_stride,
+                                 const int8_t *filter, int32_t height,
+                                 uint8_t rnd_val)
+{
+    int32_t loop_cnt;
+    uint32_t out2, out3;
+    uint64_t out0, out1;
+    v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, res0, res1;
+    v16i8 res2, vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
+    v8i16 vec01, vec23, vec45, vec67, tmp0, tmp1, tmp2;
+    v8i16 filt, filt0, filt1, filt2, filt3;
+    v8i16 rnd_vec;
+    v4i32 mask = { 2, 6, 2, 6 };
+
+    src -= (3 * src_stride);
+    rnd_vec = __msa_fill_h(rnd_val);
+
+    /* rearranging filter_y */
+    filt = LD_SH(filter);
+    SPLATI_H4_SH(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    LD_SB7(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
+    src += (7 * src_stride);
+
+    XORI_B7_128_SB(src0, src1, src2, src3, src4, src5, src6);
+
+    /* 4 width */
+    VSHF_W2_SB(src0, src1, src1, src2, mask, mask, vec0, vec1);
+    VSHF_W2_SB(src2, src3, src3, src4, mask, mask, vec2, vec3);
+    VSHF_W2_SB(src4, src5, src5, src6, mask, mask, vec4, vec5);
+
+    for (loop_cnt = (height >> 1); loop_cnt--;) {
+        LD_SB2(src, src_stride, src7, src8);
+        XORI_B2_128_SB(src7, src8);
+        src += (2 * src_stride);
+
+        ILVR_B4_SH(src1, src0, src3, src2, src5, src4, src7, src6,
+                   vec01, vec23, vec45, vec67);
+        tmp0 = FILT_8TAP_DPADD_S_H(vec01, vec23, vec45, vec67, filt0, filt1,
+                                   filt2, filt3);
+        ILVR_B4_SH(src2, src1, src4, src3, src6, src5, src8, src7, vec01, vec23,
+                   vec45, vec67);
+        tmp1 = FILT_8TAP_DPADD_S_H(vec01, vec23, vec45, vec67, filt0, filt1,
+                                   filt2, filt3);
+
+        /* 4 width */
+        VSHF_W2_SB(src6, src7, src7, src8, mask, mask, vec6, vec7);
+        ILVR_B4_SH(vec1, vec0, vec3, vec2, vec5, vec4, vec7, vec6, vec01, vec23,
+                   vec45, vec67);
+        tmp2 = FILT_8TAP_DPADD_S_H(vec01, vec23, vec45, vec67, filt0, filt1,
+                                   filt2, filt3);
+        SRAR_H3_SH(tmp0, tmp1, tmp2, rnd_vec);
+        SAT_SH3_SH(tmp0, tmp1, tmp2, 7);
+        PCKEV_B3_SB(tmp0, tmp0, tmp1, tmp1, tmp2, tmp2, res0, res1, res2);
+        XORI_B3_128_SB(res0, res1, res2);
+
+        out0 = __msa_copy_u_d((v2i64) res0, 0);
+        out1 = __msa_copy_u_d((v2i64) res1, 0);
+        out2 = __msa_copy_u_w((v4i32) res2, 0);
+        out3 = __msa_copy_u_w((v4i32) res2, 1);
+        SD(out0, dst);
+        SW(out2, (dst + 8));
+        dst += dst_stride;
+        SD(out1, dst);
+        SW(out3, (dst + 8));
+        dst += dst_stride;
+
+        src0 = src2;
+        src1 = src3;
+        src2 = src4;
+        src3 = src5;
+        src4 = src6;
+        src5 = src7;
+        src6 = src8;
+        vec0 = vec2;
+        vec1 = vec3;
+        vec2 = vec4;
+        vec3 = vec5;
+        vec4 = vec6;
+        vec5 = vec7;
+    }
+}
+
+static void common_vt_8t_16w_msa(uint8_t *src, int32_t src_stride,
+                                 uint8_t *dst, int32_t dst_stride,
+                                 const int8_t *filter, int32_t height,
+                                 uint8_t rnd_val)
+{
+    uint32_t loop_cnt;
+    v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
+    v16i8 filt0, filt1, filt2, filt3;
+    v16i8 src10_r, src32_r, src54_r, src76_r, src98_r, src21_r, src43_r;
+    v16i8 src65_r, src87_r, src109_r, src10_l, src32_l, src54_l, src76_l;
+    v16i8 src98_l, src21_l, src43_l, src65_l, src87_l, src109_l;
+    v16u8 tmp0, tmp1, tmp2, tmp3;
+    v8i16 filt, out0_r, out1_r, out2_r, out3_r, out0_l, out1_l, out2_l, out3_l;
+    v8i16 rnd_vec;
+
+    src -= (3 * src_stride);
+    rnd_vec = __msa_fill_h(rnd_val);
+
+    filt = LD_SH(filter);
+    SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    LD_SB7(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
+    XORI_B7_128_SB(src0, src1, src2, src3, src4, src5, src6);
+    src += (7 * src_stride);
+    ILVR_B4_SB(src1, src0, src3, src2, src5, src4, src2, src1, src10_r, src32_r,
+               src54_r, src21_r);
+    ILVR_B2_SB(src4, src3, src6, src5, src43_r, src65_r);
+    ILVL_B4_SB(src1, src0, src3, src2, src5, src4, src2, src1, src10_l, src32_l,
+               src54_l, src21_l);
+    ILVL_B2_SB(src4, src3, src6, src5, src43_l, src65_l);
+
+    for (loop_cnt = (height >> 2); loop_cnt--;) {
+        LD_SB4(src, src_stride, src7, src8, src9, src10);
+        XORI_B4_128_SB(src7, src8, src9, src10);
+        src += (4 * src_stride);
+
+        ILVR_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_r,
+                   src87_r, src98_r, src109_r);
+        ILVL_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_l,
+                   src87_l, src98_l, src109_l);
+        out0_r = FILT_8TAP_DPADD_S_H(src10_r, src32_r, src54_r, src76_r, filt0,
+                                     filt1, filt2, filt3);
+        out1_r = FILT_8TAP_DPADD_S_H(src21_r, src43_r, src65_r, src87_r, filt0,
+                                     filt1, filt2, filt3);
+        out2_r = FILT_8TAP_DPADD_S_H(src32_r, src54_r, src76_r, src98_r, filt0,
+                                     filt1, filt2, filt3);
+        out3_r = FILT_8TAP_DPADD_S_H(src43_r, src65_r, src87_r, src109_r, filt0,
+                                     filt1, filt2, filt3);
+        out0_l = FILT_8TAP_DPADD_S_H(src10_l, src32_l, src54_l, src76_l, filt0,
+                                     filt1, filt2, filt3);
+        out1_l = FILT_8TAP_DPADD_S_H(src21_l, src43_l, src65_l, src87_l, filt0,
+                                     filt1, filt2, filt3);
+        out2_l = FILT_8TAP_DPADD_S_H(src32_l, src54_l, src76_l, src98_l, filt0,
+                                     filt1, filt2, filt3);
+        out3_l = FILT_8TAP_DPADD_S_H(src43_l, src65_l, src87_l, src109_l, filt0,
+                                     filt1, filt2, filt3);
+        SRAR_H4_SH(out0_r, out1_r, out2_r, out3_r, rnd_vec);
+        SRAR_H4_SH(out0_l, out1_l, out2_l, out3_l, rnd_vec);
+        SAT_SH4_SH(out0_r, out1_r, out2_r, out3_r, 7);
+        SAT_SH4_SH(out0_l, out1_l, out2_l, out3_l, 7);
+        PCKEV_B4_UB(out0_l, out0_r, out1_l, out1_r, out2_l, out2_r, out3_l,
+                    out3_r, tmp0, tmp1, tmp2, tmp3);
+        XORI_B4_128_UB(tmp0, tmp1, tmp2, tmp3);
+        ST_UB4(tmp0, tmp1, tmp2, tmp3, dst, dst_stride);
+        dst += (4 * dst_stride);
+
+        src10_r = src54_r;
+        src32_r = src76_r;
+        src54_r = src98_r;
+        src21_r = src65_r;
+        src43_r = src87_r;
+        src65_r = src109_r;
+        src10_l = src54_l;
+        src32_l = src76_l;
+        src54_l = src98_l;
+        src21_l = src65_l;
+        src43_l = src87_l;
+        src65_l = src109_l;
+        src6 = src10;
+    }
+}
+
+static void common_vt_8t_16w_mult_msa(uint8_t *src, int32_t src_stride,
+                                      uint8_t *dst, int32_t dst_stride,
+                                      const int8_t *filter, int32_t height,
+                                      uint8_t rnd_val, int32_t width)
+{
+    uint8_t *src_tmp;
+    uint8_t *dst_tmp;
+    uint32_t loop_cnt, cnt;
+    v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
+    v16i8 filt0, filt1, filt2, filt3;
+    v16i8 src10_r, src32_r, src54_r, src76_r, src98_r, src21_r, src43_r;
+    v16i8 src65_r, src87_r, src109_r, src10_l, src32_l, src54_l, src76_l;
+    v16i8 src98_l, src21_l, src43_l, src65_l, src87_l, src109_l;
+    v16u8 tmp0, tmp1, tmp2, tmp3;
+    v8i16 filt, out0_r, out1_r, out2_r, out3_r, out0_l, out1_l, out2_l, out3_l;
+    v8i16 rnd_vec;
+
+    src -= (3 * src_stride);
+    rnd_vec = __msa_fill_h(rnd_val);
+
+    filt = LD_SH(filter);
+    SPLATI_H4_SB(filt, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    for (cnt = (width >> 4); cnt--;) {
+        src_tmp = src;
+        dst_tmp = dst;
+
+        LD_SB7(src_tmp, src_stride, src0, src1, src2, src3, src4, src5, src6);
+        XORI_B7_128_SB(src0, src1, src2, src3, src4, src5, src6);
+        src_tmp += (7 * src_stride);
+        ILVR_B4_SB(src1, src0, src3, src2, src5, src4, src2, src1, src10_r,
+                   src32_r, src54_r, src21_r);
+        ILVR_B2_SB(src4, src3, src6, src5, src43_r, src65_r);
+        ILVL_B4_SB(src1, src0, src3, src2, src5, src4, src2, src1, src10_l,
+                   src32_l, src54_l, src21_l);
+        ILVL_B2_SB(src4, src3, src6, src5, src43_l, src65_l);
+
+        for (loop_cnt = (height >> 2); loop_cnt--;) {
+            LD_SB4(src_tmp, src_stride, src7, src8, src9, src10);
+            XORI_B4_128_SB(src7, src8, src9, src10);
+            src_tmp += (4 * src_stride);
+            ILVR_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_r,
+                       src87_r, src98_r, src109_r);
+            ILVL_B4_SB(src7, src6, src8, src7, src9, src8, src10, src9, src76_l,
+                       src87_l, src98_l, src109_l);
+            out0_r = FILT_8TAP_DPADD_S_H(src10_r, src32_r, src54_r, src76_r,
+                                         filt0, filt1, filt2, filt3);
+            out1_r = FILT_8TAP_DPADD_S_H(src21_r, src43_r, src65_r, src87_r,
+                                         filt0, filt1, filt2, filt3);
+            out2_r = FILT_8TAP_DPADD_S_H(src32_r, src54_r, src76_r, src98_r,
+                                         filt0, filt1, filt2, filt3);
+            out3_r = FILT_8TAP_DPADD_S_H(src43_r, src65_r, src87_r, src109_r,
+                                         filt0, filt1, filt2, filt3);
+            out0_l = FILT_8TAP_DPADD_S_H(src10_l, src32_l, src54_l, src76_l,
+                                         filt0, filt1, filt2, filt3);
+            out1_l = FILT_8TAP_DPADD_S_H(src21_l, src43_l, src65_l, src87_l,
+                                         filt0, filt1, filt2, filt3);
+            out2_l = FILT_8TAP_DPADD_S_H(src32_l, src54_l, src76_l, src98_l,
+                                         filt0, filt1, filt2, filt3);
+            out3_l = FILT_8TAP_DPADD_S_H(src43_l, src65_l, src87_l, src109_l,
+                                         filt0, filt1, filt2, filt3);
+            SRAR_H4_SH(out0_r, out1_r, out2_r, out3_r, rnd_vec);
+            SRAR_H4_SH(out0_l, out1_l, out2_l, out3_l, rnd_vec);
+            SAT_SH4_SH(out0_r, out1_r, out2_r, out3_r, 7);
+            SAT_SH4_SH(out0_l, out1_l, out2_l, out3_l, 7);
+            PCKEV_B4_UB(out0_l, out0_r, out1_l, out1_r, out2_l, out2_r, out3_l,
+                        out3_r, tmp0, tmp1, tmp2, tmp3);
+            XORI_B4_128_UB(tmp0, tmp1, tmp2, tmp3);
+            ST_UB4(tmp0, tmp1, tmp2, tmp3, dst_tmp, dst_stride);
+            dst_tmp += (4 * dst_stride);
+
+            src10_r = src54_r;
+            src32_r = src76_r;
+            src54_r = src98_r;
+            src21_r = src65_r;
+            src43_r = src87_r;
+            src65_r = src109_r;
+            src10_l = src54_l;
+            src32_l = src76_l;
+            src54_l = src98_l;
+            src21_l = src65_l;
+            src43_l = src87_l;
+            src65_l = src109_l;
+            src6 = src10;
+        }
+
+        src += 16;
+        dst += 16;
+    }
+}
+
+static void common_vt_8t_24w_msa(uint8_t *src, int32_t src_stride,
+                                 uint8_t *dst, int32_t dst_stride,
+                                 const int8_t *filter, int32_t height, uint8_t rnd_val)
+{
+    common_vt_8t_16w_mult_msa(src, src_stride, dst, dst_stride, filter, height,
+                              rnd_val, 16);
+
+    common_vt_8t_8w_msa(src + 16, src_stride, dst + 16, dst_stride, filter,
+                        height, rnd_val);
+}
+
+static void common_vt_8t_32w_msa(uint8_t *src, int32_t src_stride,
+                                 uint8_t *dst, int32_t dst_stride,
+                                 const int8_t *filter, int32_t height, uint8_t rnd_val)
+{
+    common_vt_8t_16w_mult_msa(src, src_stride, dst, dst_stride, filter, height,
+                              rnd_val, 32);
+}
+
+static void common_vt_8t_48w_msa(uint8_t *src, int32_t src_stride,
+                                 uint8_t *dst, int32_t dst_stride,
+                                 const int8_t *filter, int32_t height, uint8_t rnd_val)
+{
+    common_vt_8t_16w_mult_msa(src, src_stride, dst, dst_stride, filter, height,
+                              rnd_val, 48);
+}
+
+static void common_vt_8t_64w_msa(uint8_t *src, int32_t src_stride,
+                                 uint8_t *dst, int32_t dst_stride,
+                                 const int8_t *filter, int32_t height, uint8_t rnd_val)
+{
+    common_vt_8t_16w_mult_msa(src, src_stride, dst, dst_stride, filter, height,
+                              rnd_val, 64);
+}
+
+static void hevc_hv_uni_8t_4w_msa(uint8_t *src,
+                                  int32_t src_stride,
+                                  uint8_t *dst,
+                                  int32_t dst_stride,
+                                  const int8_t *filter_x,
+                                  const int8_t *filter_y,
+                                  int32_t height)
+{
+    uint32_t loop_cnt;
+    v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
+    v8i16 filt0, filt1, filt2, filt3;
+    v4i32 filt_h0, filt_h1, filt_h2, filt_h3;
+    v16i8 mask1, mask2, mask3;
+    v8i16 filter_vec, const_vec;
+    v16i8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
+    v16i8 vec8, vec9, vec10, vec11, vec12, vec13, vec14, vec15;
+    v8i16 dst30, dst41, dst52, dst63, dst66, dst87;
+    v4i32 dst0_r, dst1_r;
+    v8i16 dst10_r, dst32_r, dst54_r, dst76_r;
+    v8i16 dst21_r, dst43_r, dst65_r, dst87_r;
+    v16i8 mask0 = { 0, 1, 1, 2, 2, 3, 3, 4, 16, 17, 17, 18, 18, 19, 19, 20 };
+    v8i16 mask4 = { 0, 4, 1, 5, 2, 6, 3, 7 };
+
+    src -= ((3 * src_stride) + 3);
+    filter_vec = LD_SH(filter_x);
+    SPLATI_H4_SH(filter_vec, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    filter_vec = LD_SH(filter_y);
+    vec0 = __msa_clti_s_b((v16i8) filter_vec, 0);
+    filter_vec = (v8i16) __msa_ilvr_b(vec0, (v16i8) filter_vec);
+
+    SPLATI_W4_SW(filter_vec, filt_h0, filt_h1, filt_h2, filt_h3);
+
+    mask1 = mask0 + 2;
+    mask2 = mask0 + 4;
+    mask3 = mask0 + 6;
+
+    const_vec = __msa_ldi_h(128);
+    const_vec <<= 6;
+
+    LD_SB7(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
+    src += (7 * src_stride);
+    XORI_B7_128_SB(src0, src1, src2, src3, src4, src5, src6);
+
+    VSHF_B4_SB(src0, src3, mask0, mask1, mask2, mask3, vec0, vec1, vec2, vec3);
+    VSHF_B4_SB(src1, src4, mask0, mask1, mask2, mask3, vec4, vec5, vec6, vec7);
+    VSHF_B4_SB(src2, src5, mask0, mask1, mask2, mask3,
+               vec8, vec9, vec10, vec11);
+    VSHF_B4_SB(src3, src6, mask0, mask1, mask2, mask3,
+               vec12, vec13, vec14, vec15);
+
+    dst30 = const_vec;
+    DPADD_SB4_SH(vec0, vec1, vec2, vec3, filt0, filt1, filt2, filt3,
+                 dst30, dst30, dst30, dst30);
+    dst41 = const_vec;
+    DPADD_SB4_SH(vec4, vec5, vec6, vec7, filt0, filt1, filt2, filt3,
+                 dst41, dst41, dst41, dst41);
+    dst52 = const_vec;
+    DPADD_SB4_SH(vec8, vec9, vec10, vec11, filt0, filt1, filt2, filt3,
+                 dst52, dst52, dst52, dst52);
+    dst63 = const_vec;
+    DPADD_SB4_SH(vec12, vec13, vec14, vec15, filt0, filt1, filt2, filt3,
+                 dst63, dst63, dst63, dst63);
+
+    ILVR_H3_SH(dst41, dst30, dst52, dst41, dst63, dst52,
+               dst10_r, dst21_r, dst32_r);
+    dst43_r = __msa_ilvl_h(dst41, dst30);
+    dst54_r = __msa_ilvl_h(dst52, dst41);
+    dst65_r = __msa_ilvl_h(dst63, dst52);
+    dst66 = (v8i16) __msa_splati_d((v2i64) dst63, 1);
+
+    for (loop_cnt = height >> 1; loop_cnt--;) {
+        LD_SB2(src, src_stride, src7, src8);
+        src += 2 * src_stride;
+        XORI_B2_128_SB(src7, src8);
+
+        VSHF_B4_SB(src7, src8, mask0, mask1, mask2, mask3,
+                   vec0, vec1, vec2, vec3);
+        dst87 = const_vec;
+        DPADD_SB4_SH(vec0, vec1, vec2, vec3, filt0, filt1, filt2, filt3,
+                     dst87, dst87, dst87, dst87);
+
+        dst76_r = __msa_ilvr_h(dst87, dst66);
+        dst0_r = HEVC_FILT_8TAP(dst10_r, dst32_r, dst54_r, dst76_r,
+                                filt_h0, filt_h1, filt_h2, filt_h3);
+        dst87_r = __msa_vshf_h(mask4, dst87, dst87);
+        dst1_r = HEVC_FILT_8TAP(dst21_r, dst43_r, dst65_r, dst87_r,
+                                filt_h0, filt_h1, filt_h2, filt_h3);
+
+        dst0_r >>= 6;
+        dst1_r >>= 6;
+        SRARI_W2_SW(dst0_r, dst1_r, 6);
+        dst0_r = CLIP_SW_0_255(dst0_r);
+        dst1_r = CLIP_SW_0_255(dst1_r);
+
+        HEVC_PCK_SW_SB2(dst1_r, dst0_r, dst0_r);
+        ST4x2_UB(dst0_r, dst, dst_stride);
+        dst += (2 * dst_stride);
+
+        dst10_r = dst32_r;
+        dst32_r = dst54_r;
+        dst54_r = dst76_r;
+        dst21_r = dst43_r;
+        dst43_r = dst65_r;
+        dst65_r = dst87_r;
+        dst66 = (v8i16) __msa_splati_d((v2i64) dst87, 1);
+    }
+}
+
+static void hevc_hv_uni_8t_8multx2mult_msa(uint8_t *src,
+                                           int32_t src_stride,
+                                           uint8_t *dst,
+                                           int32_t dst_stride,
+                                           const int8_t *filter_x,
+                                           const int8_t *filter_y,
+                                           int32_t height, int32_t width)
+{
+    uint32_t loop_cnt, cnt;
+    uint8_t *src_tmp;
+    uint8_t *dst_tmp;
+    v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
+    v8i16 filt0, filt1, filt2, filt3;
+    v4i32 filt_h0, filt_h1, filt_h2, filt_h3;
+    v16i8 mask1, mask2, mask3;
+    v8i16 filter_vec, const_vec;
+    v16i8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
+    v16i8 vec8, vec9, vec10, vec11, vec12, vec13, vec14, vec15;
+    v8i16 dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7, dst8;
+    v4i32 dst0_r, dst0_l, dst1_r, dst1_l;
+    v8i16 dst10_r, dst32_r, dst54_r, dst76_r;
+    v8i16 dst10_l, dst32_l, dst54_l, dst76_l;
+    v8i16 dst21_r, dst43_r, dst65_r, dst87_r;
+    v8i16 dst21_l, dst43_l, dst65_l, dst87_l;
+    v16i8 mask0 = { 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8 };
+
+    src -= ((3 * src_stride) + 3);
+    const_vec = __msa_ldi_h(128);
+    const_vec <<= 6;
+
+    filter_vec = LD_SH(filter_x);
+    SPLATI_H4_SH(filter_vec, 0, 1, 2, 3, filt0, filt1, filt2, filt3);
+
+    filter_vec = LD_SH(filter_y);
+    vec0 = __msa_clti_s_b((v16i8) filter_vec, 0);
+    filter_vec = (v8i16) __msa_ilvr_b(vec0, (v16i8) filter_vec);
+
+    SPLATI_W4_SW(filter_vec, filt_h0, filt_h1, filt_h2, filt_h3);
+
+    mask1 = mask0 + 2;
+    mask2 = mask0 + 4;
+    mask3 = mask0 + 6;
+
+    for (cnt = width >> 3; cnt--;) {
+        src_tmp = src;
+        dst_tmp = dst;
+
+        LD_SB7(src_tmp, src_stride, src0, src1, src2, src3, src4, src5, src6);
+        src_tmp += (7 * src_stride);
+        XORI_B7_128_SB(src0, src1, src2, src3, src4, src5, src6);
+
+        /* row 0 row 1 row 2 row 3 */
+        VSHF_B4_SB(src0, src0, mask0, mask1, mask2, mask3,
+                   vec0, vec1, vec2, vec3);
+        VSHF_B4_SB(src1, src1, mask0, mask1, mask2, mask3,
+                   vec4, vec5, vec6, vec7);
+        VSHF_B4_SB(src2, src2, mask0, mask1, mask2, mask3,
+                   vec8, vec9, vec10, vec11);
+        VSHF_B4_SB(src3, src3, mask0, mask1, mask2, mask3,
+                   vec12, vec13, vec14, vec15);
+        dst0 = const_vec;
+        DPADD_SB4_SH(vec0, vec1, vec2, vec3, filt0, filt1, filt2, filt3,
+                     dst0, dst0, dst0, dst0);
+        dst1 = const_vec;
+        DPADD_SB4_SH(vec4, vec5, vec6, vec7, filt0, filt1, filt2, filt3,
+                     dst1, dst1, dst1, dst1);
+        dst2 = const_vec;
+        DPADD_SB4_SH(vec8, vec9, vec10, vec11, filt0, filt1, filt2, filt3,
+                     dst2, dst2, dst2, dst2);
+        dst3 = const_vec;
+        DPADD_SB4_SH(vec12, vec13, vec14, vec15, filt0, filt1, filt2, filt3,
+                     dst3, dst3, dst3, dst3);
+
+        VSHF_B4_SB(src4, src4, mask0, mask1, mask2, mask3,
+                   vec0, vec1, vec2, vec3);
+        VSHF_B4_SB(src5, src5, mask0, mask1, mask2, mask3,
+                   vec4, vec5, vec6, vec7);
+        VSHF_B4_SB(src6, src6, mask0, mask1, mask2, mask3,
+                   vec8, vec9, vec10, vec11);
+        dst4 = const_vec;
+        DPADD_SB4_SH(vec0, vec1, vec2, vec3, filt0, filt1, filt2, filt3,
+                     dst4, dst4, dst4, dst4);
+        dst5 = const_vec;
+        DPADD_SB4_SH(vec4, vec5, vec6, vec7, filt0, filt1, filt2, filt3,
+                     dst5, dst5, dst5, dst5);
+        dst6 = const_vec;
+        DPADD_SB4_SH(vec8, vec9, vec10, vec11, filt0, filt1, filt2, filt3,
+                     dst6, dst6, dst6, dst6);
+
+        ILVR_H4_SH(dst1, dst0, dst3, dst2, dst5, dst4, dst2, dst1,
+                   dst10_r, dst32_r, dst54_r, dst21_r);
+        ILVR_H2_SH(dst4, dst3, dst6, dst5, dst43_r, dst65_r);
+        ILVL_H4_SH(dst1, dst0, dst3, dst2, dst5, dst4, dst2, dst1,
+                   dst10_l, dst32_l, dst54_l, dst21_l);
+        ILVL_H2_SH(dst4, dst3, dst6, dst5, dst43_l, dst65_l);
+
+        for (loop_cnt = height >> 1; loop_cnt--;) {
+            LD_SB2(src_tmp, src_stride, src7, src8);
+            XORI_B2_128_SB(src7, src8);
+            src_tmp += 2 * src_stride;
+
+            VSHF_B4_SB(src7, src7, mask0, mask1, mask2, mask3,
+                       vec0, vec1, vec2, vec3);
+            dst7 = const_vec;
+            DPADD_SB4_SH(vec0, vec1, vec2, vec3, filt0, filt1, filt2, filt3,
+                         dst7, dst7, dst7, dst7);
+
+            ILVRL_H2_SH(dst7, dst6, dst76_r, dst76_l);
+            dst0_r = HEVC_FILT_8TAP(dst10_r, dst32_r, dst54_r, dst76_r,
+                                    filt_h0, filt_h1, filt_h2, filt_h3);
+            dst0_l = HEVC_FILT_8TAP(dst10_l, dst32_l, dst54_l, dst76_l,
+                                    filt_h0, filt_h1, filt_h2, filt_h3);
+            dst0_r >>= 6;
+            dst0_l >>= 6;
+
+            VSHF_B4_SB(src8, src8, mask0, mask1, mask2, mask3,
+                       vec0, vec1, vec2, vec3);
+            dst8 = const_vec;
+            DPADD_SB4_SH(vec0, vec1, vec2, vec3, filt0, filt1, filt2, filt3,
+                         dst8, dst8, dst8, dst8);
+
+            ILVRL_H2_SH(dst8, dst7, dst87_r, dst87_l);
+            dst1_r = HEVC_FILT_8TAP(dst21_r, dst43_r, dst65_r, dst87_r,
+                                    filt_h0, filt_h1, filt_h2, filt_h3);
+            dst1_l = HEVC_FILT_8TAP(dst21_l, dst43_l, dst65_l, dst87_l,
+                                    filt_h0, filt_h1, filt_h2, filt_h3);
+            dst1_r >>= 6;
+            dst1_l >>= 6;
+            SRARI_W4_SW(dst0_r, dst0_l, dst1_r, dst1_l, 6);
+            dst0_r = CLIP_SW_0_255(dst0_r);
+            dst0_l = CLIP_SW_0_255(dst0_l);
+            dst1_r = CLIP_SW_0_255(dst1_r);
+            dst1_l = CLIP_SW_0_255(dst1_l);
+
+            HEVC_PCK_SW_SB4(dst0_l, dst0_r, dst1_l, dst1_r, dst0_r);
+            ST8x2_UB(dst0_r, dst_tmp, dst_stride);
+            dst_tmp += (2 * dst_stride);
+
+            dst10_r = dst32_r;
+            dst32_r = dst54_r;
+            dst54_r = dst76_r;
+            dst10_l = dst32_l;
+            dst32_l = dst54_l;
+            dst54_l = dst76_l;
+            dst21_r = dst43_r;
+            dst43_r = dst65_r;
+            dst65_r = dst87_r;
+            dst21_l = dst43_l;
+            dst43_l = dst65_l;
+            dst65_l = dst87_l;
+            dst6 = dst8;
+        }
+
+        src += 8;
+        dst += 8;
+    }
+}
+
+static void hevc_hv_uni_8t_8w_msa(uint8_t *src,
+                                  int32_t src_stride,
+                                  uint8_t *dst,
+                                  int32_t dst_stride,
+                                  const int8_t *filter_x,
+                                  const int8_t *filter_y,
+                                  int32_t height)
+{
+    hevc_hv_uni_8t_8multx2mult_msa(src, src_stride, dst, dst_stride,
+                                   filter_x, filter_y, height, 8);
+}
+
+static void hevc_hv_uni_8t_12w_msa(uint8_t *src,
+                                   int32_t src_stride,
+                                   uint8_t *dst,
+                                   int32_t dst_stride,
+                                   const int8_t *filter_x,
+                                   const int8_t *filter_y,
+                                   int32_t height)
+{
+    hevc_hv_uni_8t_8multx2mult_msa(src, src_stride, dst, dst_stride,
+                                   filter_x, filter_y, height, 8);
+
+    hevc_hv_uni_8t_4w_msa(src + 8, src_stride, dst + 8, dst_stride,
+                          filter_x, filter_y, height);
+}
+
+static void hevc_hv_uni_8t_16w_msa(uint8_t *src,
+                                   int32_t src_stride,
+                                   uint8_t *dst,
+                                   int32_t dst_stride,
+                                   const int8_t *filter_x,
+                                   const int8_t *filter_y,
+                                   int32_t height)
+{
+    hevc_hv_uni_8t_8multx2mult_msa(src, src_stride, dst, dst_stride,
+                                   filter_x, filter_y, height, 16);
+}
+
+static void hevc_hv_uni_8t_24w_msa(uint8_t *src,
+                                   int32_t src_stride,
+                                   uint8_t *dst,
+                                   int32_t dst_stride,
+                                   const int8_t *filter_x,
+                                   const int8_t *filter_y,
+                                   int32_t height)
+{
+    hevc_hv_uni_8t_8multx2mult_msa(src, src_stride, dst, dst_stride,
+                                   filter_x, filter_y, height, 24);
+}
+
+static void hevc_hv_uni_8t_32w_msa(uint8_t *src,
+                                   int32_t src_stride,
+                                   uint8_t *dst,
+                                   int32_t dst_stride,
+                                   const int8_t *filter_x,
+                                   const int8_t *filter_y,
+                                   int32_t height)
+{
+    hevc_hv_uni_8t_8multx2mult_msa(src, src_stride, dst, dst_stride,
+                                   filter_x, filter_y, height, 32);
+}
+
+static void hevc_hv_uni_8t_48w_msa(uint8_t *src,
+                                   int32_t src_stride,
+                                   uint8_t *dst,
+                                   int32_t dst_stride,
+                                   const int8_t *filter_x,
+                                   const int8_t *filter_y,
+                                   int32_t height)
+{
+    hevc_hv_uni_8t_8multx2mult_msa(src, src_stride, dst, dst_stride,
+                                   filter_x, filter_y, height, 48);
+}
+
+static void hevc_hv_uni_8t_64w_msa(uint8_t *src,
+                                   int32_t src_stride,
+                                   uint8_t *dst,
+                                   int32_t dst_stride,
+                                   const int8_t *filter_x,
+                                   const int8_t *filter_y,
+                                   int32_t height)
+{
+    hevc_hv_uni_8t_8multx2mult_msa(src, src_stride, dst, dst_stride,
+                                   filter_x, filter_y, height, 64);
+}
+
+#define UNI_MC_COPY(WIDTH)                                                 \
+void ff_hevc_put_hevc_uni_pel_pixels##WIDTH##_8_msa(uint8_t *dst,          \
+                                                    ptrdiff_t dst_stride,  \
+                                                    uint8_t *src,          \
+                                                    ptrdiff_t src_stride,  \
+                                                    int height,            \
+                                                    intptr_t mx,           \
+                                                    intptr_t my,           \
+                                                    int width)             \
+{                                                                          \
+    copy_width##WIDTH##_msa(src, src_stride, dst, dst_stride, height);     \
+}
+
+UNI_MC_COPY(8);
+UNI_MC_COPY(12);
+UNI_MC_COPY(16);
+UNI_MC_COPY(24);
+UNI_MC_COPY(32);
+UNI_MC_COPY(48);
+UNI_MC_COPY(64);
+
+#undef UNI_MC_COPY
+
+#define UNI_MC(PEL, DIR, WIDTH, TAP, DIR1, FILT_DIR)                           \
+void ff_hevc_put_hevc_uni_##PEL##_##DIR####WIDTH##_8_msa(uint8_t *dst,         \
+                                                         ptrdiff_t             \
+                                                         dst_stride,           \
+                                                         uint8_t *src,         \
+                                                         ptrdiff_t             \
+                                                         src_stride,           \
+                                                         int height,           \
+                                                         intptr_t mx,          \
+                                                         intptr_t my,          \
+                                                         int width)            \
+{                                                                              \
+    const int8_t *filter = ff_hevc_##PEL##_filters[FILT_DIR - 1];              \
+                                                                               \
+    common_##DIR1##_##TAP##t_##WIDTH##w_msa(src, src_stride, dst, dst_stride,  \
+                                            filter, height, 6);                \
+}
+
+UNI_MC(qpel, h, 4, 8, hz, mx);
+UNI_MC(qpel, h, 8, 8, hz, mx);
+UNI_MC(qpel, h, 12, 8, hz, mx);
+UNI_MC(qpel, h, 16, 8, hz, mx);
+UNI_MC(qpel, h, 24, 8, hz, mx);
+UNI_MC(qpel, h, 32, 8, hz, mx);
+UNI_MC(qpel, h, 48, 8, hz, mx);
+UNI_MC(qpel, h, 64, 8, hz, mx);
+
+UNI_MC(qpel, v, 4, 8, vt, my);
+UNI_MC(qpel, v, 8, 8, vt, my);
+UNI_MC(qpel, v, 12, 8, vt, my);
+UNI_MC(qpel, v, 16, 8, vt, my);
+UNI_MC(qpel, v, 24, 8, vt, my);
+UNI_MC(qpel, v, 32, 8, vt, my);
+UNI_MC(qpel, v, 48, 8, vt, my);
+UNI_MC(qpel, v, 64, 8, vt, my);
+
+#undef UNI_MC
+
+#define UNI_MC_HV(PEL, DIR, WIDTH, TAP, DIR1)                           \
+void ff_hevc_put_hevc_uni_##PEL##_##DIR####WIDTH##_8_msa(uint8_t *dst,  \
+                                                         ptrdiff_t      \
+                                                         dst_stride,    \
+                                                         uint8_t *src,  \
+                                                         ptrdiff_t      \
+                                                         src_stride,    \
+                                                         int height,    \
+                                                         intptr_t mx,   \
+                                                         intptr_t my,   \
+                                                         int width)     \
+{                                                                       \
+    const int8_t *filter_x = ff_hevc_##PEL##_filters[mx - 1];           \
+    const int8_t *filter_y = ff_hevc_##PEL##_filters[my - 1];           \
+                                                                        \
+    hevc_##DIR1##_uni_##TAP##t_##WIDTH##w_msa(src, src_stride, dst,     \
+                                              dst_stride, filter_x,     \
+                                              filter_y, height);        \
+}
+
+UNI_MC_HV(qpel, hv, 4, 8, hv);
+UNI_MC_HV(qpel, hv, 8, 8, hv);
+UNI_MC_HV(qpel, hv, 12, 8, hv);
+UNI_MC_HV(qpel, hv, 16, 8, hv);
+UNI_MC_HV(qpel, hv, 24, 8, hv);
+UNI_MC_HV(qpel, hv, 32, 8, hv);
+UNI_MC_HV(qpel, hv, 48, 8, hv);
+UNI_MC_HV(qpel, hv, 64, 8, hv);
+
+#undef UNI_MC_HV
diff --git a/libavcodec/mips/hevcdsp_msa.c b/libavcodec/mips/hevcdsp_msa.c
index 781264d..fcc344b 100644
--- a/libavcodec/mips/hevcdsp_msa.c
+++ b/libavcodec/mips/hevcdsp_msa.c
@@ -46,24 +46,6 @@
     out;                                                                       \
 } )
 
-#define HEVC_RND_W_CLIP_UNSIGNED_CHAR_W_VEC2(vec0_r, vec0_l,           \
-                                             vec1_r, vec1_l,           \
-                                             out0, out1)               \
-{                                                                      \
-    (vec0_r) = __msa_srari_w((vec0_r), 6);                             \
-    (vec0_l) = __msa_srari_w((vec0_l), 6);                             \
-    (vec1_r) = __msa_srari_w((vec1_r), 6);                             \
-    (vec1_l) = __msa_srari_w((vec1_l), 6);                             \
-                                                                       \
-    (vec0_r) = CLIP_UNSIGNED_CHAR_W((vec0_r));                         \
-    (vec0_l) = CLIP_UNSIGNED_CHAR_W((vec0_l));                         \
-    (vec1_r) = CLIP_UNSIGNED_CHAR_W((vec1_r));                         \
-    (vec1_l) = CLIP_UNSIGNED_CHAR_W((vec1_l));                         \
-                                                                       \
-    out0 = (v4i32) __msa_pckev_h((v8i16) (vec0_l), (v8i16) (vec0_r));  \
-    out1 = (v4i32) __msa_pckev_h((v8i16) (vec1_l), (v8i16) (vec1_r));  \
-}
-
 static void hevc_copy_4w_msa(uint8_t * __restrict src, int32_t src_stride,
                              int16_t * __restrict dst, int32_t dst_stride,
                              int32_t height)
@@ -2288,2404 +2270,88 @@ static void hevc_hv_8t_64w_msa(uint8_t * __restrict src, int32_t src_stride,
                                filter_x, filter_y, height, 64);
 }
 
-static void hevc_hv_uni_8t_4w_msa(uint8_t * __restrict src,
-                                  int32_t src_stride,
-                                  uint8_t * __restrict dst,
-                                  int32_t dst_stride,
-                                  const int8_t * __restrict filter_x,
-                                  const int8_t * __restrict filter_y,
-                                  int32_t height)
-{
-    uint32_t loop_cnt;
-    uint32_t out0, out1;
-    v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
-    v8i16 filt0, filt1, filt2, filt3, filter_vec;
-    v4i32 filt_h0, filt_h1, filt_h2, filt_h3;
-    v16i8 mask1, mask2, mask3;
-    v8u16 const_vec;
-    v16i8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
-    v16i8 vec8, vec9, vec10, vec11, vec12, vec13, vec14, vec15;
-    v8i16 dst30, dst41, dst52, dst63, dst66, dst87;
-    v4i32 dst0_r, dst1_r;
-    v8i16 dst10_r, dst32_r, dst54_r, dst76_r;
-    v8i16 dst21_r, dst43_r, dst65_r, dst87_r;
-    v16i8 tmp;
-    v16i8 mask0 = { 0, 1, 1, 2, 2, 3, 3, 4, 16, 17, 17, 18, 18, 19, 19, 20 };
-    v8i16 mask4 = { 0, 4, 1, 5, 2, 6, 3, 7 };
-
-    src -= ((3 * src_stride) + 3);
-
-    filter_vec = LOAD_SH(filter_x);
-    filt0 = __msa_splati_h(filter_vec, 0);
-    filt1 = __msa_splati_h(filter_vec, 1);
-    filt2 = __msa_splati_h(filter_vec, 2);
-    filt3 = __msa_splati_h(filter_vec, 3);
-
-    filter_vec = LOAD_SH(filter_y);
-    tmp = __msa_clti_s_b((v16i8) filter_vec, 0);
-    filter_vec = (v8i16) __msa_ilvr_b(tmp, (v16i8) filter_vec);
-
-    filt_h0 = __msa_splati_w((v4i32) filter_vec, 0);
-    filt_h1 = __msa_splati_w((v4i32) filter_vec, 1);
-    filt_h2 = __msa_splati_w((v4i32) filter_vec, 2);
-    filt_h3 = __msa_splati_w((v4i32) filter_vec, 3);
-
-    mask1 = mask0 + 2;
-    mask2 = mask0 + 4;
-    mask3 = mask0 + 6;
-
-    const_vec = (v8u16) __msa_ldi_h(128);
-    const_vec <<= 6;
-
-    LOAD_7VECS_SB(src, src_stride,
-                  src0, src1, src2, src3, src4, src5, src6);
-    src += (7 * src_stride);
-
-    XORI_B_7VECS_SB(src0, src1, src2, src3, src4, src5, src6,
-                    src0, src1, src2, src3, src4, src5, src6, 128);
-
-    /* Row 0 Row 1 Row 2 Row 3 */
-    vec0 = __msa_vshf_b(mask0, src3, src0);
-    vec1 = __msa_vshf_b(mask1, src3, src0);
-    vec2 = __msa_vshf_b(mask2, src3, src0);
-    vec3 = __msa_vshf_b(mask3, src3, src0);
-
-    vec4 = __msa_vshf_b(mask0, src4, src1);
-    vec5 = __msa_vshf_b(mask1, src4, src1);
-    vec6 = __msa_vshf_b(mask2, src4, src1);
-    vec7 = __msa_vshf_b(mask3, src4, src1);
-
-    vec8 = __msa_vshf_b(mask0, src5, src2);
-    vec9 = __msa_vshf_b(mask1, src5, src2);
-    vec10 = __msa_vshf_b(mask2, src5, src2);
-    vec11 = __msa_vshf_b(mask3, src5, src2);
-
-    vec12 = __msa_vshf_b(mask0, src6, src3);
-    vec13 = __msa_vshf_b(mask1, src6, src3);
-    vec14 = __msa_vshf_b(mask2, src6, src3);
-    vec15 = __msa_vshf_b(mask3, src6, src3);
-
-    dst30 = HEVC_FILT_8TAP_DPADD_H(vec0, vec1, vec2, vec3,
-                                   filt0, filt1, filt2, filt3, const_vec);
-
-    dst41 = HEVC_FILT_8TAP_DPADD_H(vec4, vec5, vec6, vec7,
-                                   filt0, filt1, filt2, filt3, const_vec);
-
-    dst52 = HEVC_FILT_8TAP_DPADD_H(vec8, vec9, vec10, vec11,
-                                   filt0, filt1, filt2, filt3, const_vec);
-
-    dst63 = HEVC_FILT_8TAP_DPADD_H(vec12, vec13, vec14, vec15,
-                                   filt0, filt1, filt2, filt3, const_vec);
-
-    dst10_r = __msa_ilvr_h(dst41, dst30);
-    dst21_r = __msa_ilvr_h(dst52, dst41);
-    dst32_r = __msa_ilvr_h(dst63, dst52);
-
-    dst43_r = __msa_ilvl_h(dst41, dst30);
-    dst54_r = __msa_ilvl_h(dst52, dst41);
-    dst65_r = __msa_ilvl_h(dst63, dst52);
-
-    dst66 = (v8i16) __msa_splati_d((v2i64) dst63, 1);
-
-    for (loop_cnt = height >> 1; loop_cnt--;) {
-        src7 = LOAD_SB(src);
-        src += src_stride;
-        src8 = LOAD_SB(src);
-        src += src_stride;
-
-        XORI_B_2VECS_SB(src7, src8, src7, src8, 128);
-
-        vec0 = __msa_vshf_b(mask0, src8, src7);
-        vec1 = __msa_vshf_b(mask1, src8, src7);
-        vec2 = __msa_vshf_b(mask2, src8, src7);
-        vec3 = __msa_vshf_b(mask3, src8, src7);
-
-        dst87 = HEVC_FILT_8TAP_DPADD_H(vec0, vec1, vec2, vec3,
-                                       filt0, filt1, filt2, filt3, const_vec);
-
-        dst76_r = __msa_ilvr_h(dst87, dst66);
-
-        dst0_r = HEVC_FILT_8TAP_DPADD_W(dst10_r, dst32_r, dst54_r, dst76_r,
-                                        filt_h0, filt_h1, filt_h2, filt_h3);
-
-        dst87_r = __msa_vshf_h(mask4, dst87, dst87);
-
-        dst1_r = HEVC_FILT_8TAP_DPADD_W(dst21_r, dst43_r, dst65_r, dst87_r,
-                                        filt_h0, filt_h1, filt_h2, filt_h3);
-
-        dst0_r >>= 6;
-        dst1_r >>= 6;
-
-        dst0_r = __msa_srari_w(dst0_r, 6);
-        dst1_r = __msa_srari_w(dst1_r, 6);
-
-        dst0_r = CLIP_UNSIGNED_CHAR_W(dst0_r);
-        dst1_r = CLIP_UNSIGNED_CHAR_W(dst1_r);
-
-        dst0_r = (v4i32) __msa_pckev_h((v8i16) dst1_r, (v8i16) dst0_r);
-        dst0_r = (v4i32) __msa_pckev_b((v16i8) dst0_r, (v16i8) dst0_r);
-
-        out0 = __msa_copy_u_w(dst0_r, 0);
-        out1 = __msa_copy_u_w(dst0_r, 1);
-
-        STORE_WORD(dst, out0);
-        dst += dst_stride;
-        STORE_WORD(dst, out1);
-        dst += dst_stride;
-
-        dst10_r = dst32_r;
-        dst32_r = dst54_r;
-        dst54_r = dst76_r;
-
-        dst21_r = dst43_r;
-        dst43_r = dst65_r;
-        dst65_r = dst87_r;
-
-        dst66 = (v8i16) __msa_splati_d((v2i64) dst87, 1);
-    }
+#define MC_COPY(WIDTH)                                                    \
+void ff_hevc_put_hevc_pel_pixels##WIDTH##_8_msa(int16_t *dst,             \
+                                                uint8_t *src,             \
+                                                ptrdiff_t src_stride,     \
+                                                int height,               \
+                                                intptr_t mx,              \
+                                                intptr_t my,              \
+                                                int width)                \
+{                                                                         \
+    hevc_copy_##WIDTH##w_msa(src, src_stride, dst, MAX_PB_SIZE, height);  \
 }
 
-static void hevc_hv_uni_8t_8multx2mult_msa(uint8_t * __restrict src,
-                                           int32_t src_stride,
-                                           uint8_t * __restrict dst,
-                                           int32_t dst_stride,
-                                           const int8_t * __restrict filter_x,
-                                           const int8_t * __restrict filter_y,
-                                           int32_t height, int32_t width)
-{
-    uint32_t loop_cnt, cnt;
-    uint64_t out0, out1;
-    uint8_t *src_tmp;
-    uint8_t *dst_tmp;
-    v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
-    v8i16 filt0, filt1, filt2, filt3, filter_vec;
-    v4i32 filt_h0, filt_h1, filt_h2, filt_h3;
-    v16i8 mask1, mask2, mask3;
-    v8u16 const_vec;
-    v16i8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
-    v16i8 vec8, vec9, vec10, vec11, vec12, vec13, vec14, vec15;
-    v8i16 dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7, dst8;
-    v4i32 dst0_r, dst0_l, dst1_r, dst1_l;
-    v8i16 dst10_r, dst32_r, dst54_r, dst76_r;
-    v8i16 dst10_l, dst32_l, dst54_l, dst76_l;
-    v8i16 dst21_r, dst43_r, dst65_r, dst87_r;
-    v8i16 dst21_l, dst43_l, dst65_l, dst87_l;
-    v16i8 tmp;
-    v16i8 mask0 = { 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8 };
-
-    src -= ((3 * src_stride) + 3);
-
-    const_vec = (v8u16) __msa_ldi_h(128);
-    const_vec <<= 6;
+MC_COPY(4);
+MC_COPY(6);
+MC_COPY(8);
+MC_COPY(12);
+MC_COPY(16);
+MC_COPY(24);
+MC_COPY(32);
+MC_COPY(48);
+MC_COPY(64);
 
-    filter_vec = LOAD_SH(filter_x);
-    filt0 = __msa_splati_h(filter_vec, 0);
-    filt1 = __msa_splati_h(filter_vec, 1);
-    filt2 = __msa_splati_h(filter_vec, 2);
-    filt3 = __msa_splati_h(filter_vec, 3);
+#undef MC_COPY
 
-    filter_vec = LOAD_SH(filter_y);
-    tmp = __msa_clti_s_b((v16i8) filter_vec, 0);
-    filter_vec = (v8i16) __msa_ilvr_b(tmp, (v16i8) filter_vec);
+#define MC(PEL, DIR, WIDTH, TAP, DIR1, FILT_DIR)                            \
+void ff_hevc_put_hevc_##PEL##_##DIR####WIDTH##_8_msa(int16_t *dst,          \
+                                                     uint8_t *src,          \
+                                                     ptrdiff_t src_stride,  \
+                                                     int height,            \
+                                                     intptr_t mx,           \
+                                                     intptr_t my,           \
+                                                     int width)             \
+{                                                                           \
+    const int8_t *filter = ff_hevc_##PEL##_filters[FILT_DIR - 1];           \
+                                                                            \
+    hevc_##DIR1##_##TAP##t_##WIDTH##w_msa(src, src_stride, dst,             \
+                                          MAX_PB_SIZE, filter, height);     \
+}
 
-    filt_h0 = __msa_splati_w((v4i32) filter_vec, 0);
-    filt_h1 = __msa_splati_w((v4i32) filter_vec, 1);
-    filt_h2 = __msa_splati_w((v4i32) filter_vec, 2);
-    filt_h3 = __msa_splati_w((v4i32) filter_vec, 3);
+MC(qpel, h, 4, 8, hz, mx);
+MC(qpel, h, 8, 8, hz, mx);
+MC(qpel, h, 12, 8, hz, mx);
+MC(qpel, h, 16, 8, hz, mx);
+MC(qpel, h, 24, 8, hz, mx);
+MC(qpel, h, 32, 8, hz, mx);
+MC(qpel, h, 48, 8, hz, mx);
+MC(qpel, h, 64, 8, hz, mx);
 
-    mask1 = mask0 + 2;
-    mask2 = mask0 + 4;
-    mask3 = mask0 + 6;
+MC(qpel, v, 4, 8, vt, my);
+MC(qpel, v, 8, 8, vt, my);
+MC(qpel, v, 12, 8, vt, my);
+MC(qpel, v, 16, 8, vt, my);
+MC(qpel, v, 24, 8, vt, my);
+MC(qpel, v, 32, 8, vt, my);
+MC(qpel, v, 48, 8, vt, my);
+MC(qpel, v, 64, 8, vt, my);
 
-    for (cnt = width >> 3; cnt--;) {
-        src_tmp = src;
-        dst_tmp = dst;
+#undef MC
 
-        LOAD_7VECS_SB(src_tmp, src_stride,
-                      src0, src1, src2, src3, src4, src5, src6);
-        src_tmp += (7 * src_stride);
+#define MC_HV(PEL, DIR, WIDTH, TAP, DIR1)                                     \
+void ff_hevc_put_hevc_##PEL##_##DIR####WIDTH##_8_msa(int16_t *dst,            \
+                                                     uint8_t *src,            \
+                                                     ptrdiff_t src_stride,    \
+                                                     int height,              \
+                                                     intptr_t mx,             \
+                                                     intptr_t my,             \
+                                                     int width)               \
+{                                                                             \
+    const int8_t *filter_x = ff_hevc_##PEL##_filters[mx - 1];                 \
+    const int8_t *filter_y = ff_hevc_##PEL##_filters[my - 1];                 \
+                                                                              \
+    hevc_##DIR1##_##TAP##t_##WIDTH##w_msa(src, src_stride, dst, MAX_PB_SIZE,  \
+                                          filter_x, filter_y, height);        \
+}
 
-        XORI_B_7VECS_SB(src0, src1, src2, src3, src4, src5, src6,
-                        src0, src1, src2, src3, src4, src5, src6, 128);
-
-        /* Row 0 Row 1 Row 2 Row 3 */
-        vec0 = __msa_vshf_b(mask0, src0, src0);
-        vec1 = __msa_vshf_b(mask1, src0, src0);
-        vec2 = __msa_vshf_b(mask2, src0, src0);
-        vec3 = __msa_vshf_b(mask3, src0, src0);
-
-        vec4 = __msa_vshf_b(mask0, src1, src1);
-        vec5 = __msa_vshf_b(mask1, src1, src1);
-        vec6 = __msa_vshf_b(mask2, src1, src1);
-        vec7 = __msa_vshf_b(mask3, src1, src1);
-
-        vec8 = __msa_vshf_b(mask0, src2, src2);
-        vec9 = __msa_vshf_b(mask1, src2, src2);
-        vec10 = __msa_vshf_b(mask2, src2, src2);
-        vec11 = __msa_vshf_b(mask3, src2, src2);
-
-        vec12 = __msa_vshf_b(mask0, src3, src3);
-        vec13 = __msa_vshf_b(mask1, src3, src3);
-        vec14 = __msa_vshf_b(mask2, src3, src3);
-        vec15 = __msa_vshf_b(mask3, src3, src3);
-
-        dst0 = HEVC_FILT_8TAP_DPADD_H(vec0, vec1, vec2, vec3,
-                                      filt0, filt1, filt2, filt3, const_vec);
-
-        dst1 = HEVC_FILT_8TAP_DPADD_H(vec4, vec5, vec6, vec7,
-                                      filt0, filt1, filt2, filt3, const_vec);
-
-        dst2 = HEVC_FILT_8TAP_DPADD_H(vec8, vec9, vec10, vec11,
-                                      filt0, filt1, filt2, filt3, const_vec);
-
-        dst3 = HEVC_FILT_8TAP_DPADD_H(vec12, vec13, vec14, vec15,
-                                      filt0, filt1, filt2, filt3, const_vec);
-
-        vec0 = __msa_vshf_b(mask0, src4, src4);
-        vec1 = __msa_vshf_b(mask1, src4, src4);
-        vec2 = __msa_vshf_b(mask2, src4, src4);
-        vec3 = __msa_vshf_b(mask3, src4, src4);
-
-        vec4 = __msa_vshf_b(mask0, src5, src5);
-        vec5 = __msa_vshf_b(mask1, src5, src5);
-        vec6 = __msa_vshf_b(mask2, src5, src5);
-        vec7 = __msa_vshf_b(mask3, src5, src5);
-
-        vec8 = __msa_vshf_b(mask0, src6, src6);
-        vec9 = __msa_vshf_b(mask1, src6, src6);
-        vec10 = __msa_vshf_b(mask2, src6, src6);
-        vec11 = __msa_vshf_b(mask3, src6, src6);
-
-        dst4 = HEVC_FILT_8TAP_DPADD_H(vec0, vec1, vec2, vec3,
-                                      filt0, filt1, filt2, filt3, const_vec);
-
-        dst5 = HEVC_FILT_8TAP_DPADD_H(vec4, vec5, vec6, vec7,
-                                      filt0, filt1, filt2, filt3, const_vec);
-
-        dst6 = HEVC_FILT_8TAP_DPADD_H(vec8, vec9, vec10, vec11,
-                                      filt0, filt1, filt2, filt3, const_vec);
-
-        ILVR_H_6VECS_SH(dst0, dst2, dst4, dst1, dst3, dst5,
-                        dst1, dst3, dst5, dst2, dst4, dst6,
-                        dst10_r, dst32_r, dst54_r, dst21_r, dst43_r, dst65_r);
-
-        ILVL_H_6VECS_SH(dst0, dst2, dst4, dst1, dst3, dst5,
-                        dst1, dst3, dst5, dst2, dst4, dst6,
-                        dst10_l, dst32_l, dst54_l, dst21_l, dst43_l, dst65_l);
-
-        for (loop_cnt = height >> 1; loop_cnt--;) {
-            src7 = LOAD_SB(src_tmp);
-            src_tmp += src_stride;
-
-            src7 = (v16i8) __msa_xori_b((v16u8) src7, 128);
-
-            vec0 = __msa_vshf_b(mask0, src7, src7);
-            vec1 = __msa_vshf_b(mask1, src7, src7);
-            vec2 = __msa_vshf_b(mask2, src7, src7);
-            vec3 = __msa_vshf_b(mask3, src7, src7);
-
-            dst7 = HEVC_FILT_8TAP_DPADD_H(vec0, vec1, vec2, vec3,
-                                          filt0, filt1, filt2, filt3,
-                                          const_vec);
-
-            dst76_r = __msa_ilvr_h(dst7, dst6);
-            dst76_l = __msa_ilvl_h(dst7, dst6);
-
-            dst0_r = HEVC_FILT_8TAP_DPADD_W(dst10_r, dst32_r, dst54_r, dst76_r,
-                                            filt_h0, filt_h1, filt_h2, filt_h3);
-
-            dst0_l = HEVC_FILT_8TAP_DPADD_W(dst10_l, dst32_l, dst54_l, dst76_l,
-                                            filt_h0, filt_h1, filt_h2, filt_h3);
-
-            dst0_r >>= 6;
-            dst0_l >>= 6;
-
-            /* Row 8 */
-            src8 = LOAD_SB(src_tmp);
-            src_tmp += src_stride;
-
-            src8 = (v16i8) __msa_xori_b((v16u8) src8, 128);
-
-            vec0 = __msa_vshf_b(mask0, src8, src8);
-            vec1 = __msa_vshf_b(mask1, src8, src8);
-            vec2 = __msa_vshf_b(mask2, src8, src8);
-            vec3 = __msa_vshf_b(mask3, src8, src8);
-
-            dst8 = HEVC_FILT_8TAP_DPADD_H(vec0, vec1, vec2, vec3,
-                                          filt0, filt1, filt2, filt3,
-                                          const_vec);
-
-            dst87_r = __msa_ilvr_h(dst8, dst7);
-            dst87_l = __msa_ilvl_h(dst8, dst7);
-
-            dst1_r = HEVC_FILT_8TAP_DPADD_W(dst21_r, dst43_r, dst65_r, dst87_r,
-                                            filt_h0, filt_h1, filt_h2, filt_h3);
-
-            dst1_l = HEVC_FILT_8TAP_DPADD_W(dst21_l, dst43_l, dst65_l, dst87_l,
-                                            filt_h0, filt_h1, filt_h2, filt_h3);
-
-            dst1_r >>= 6;
-            dst1_l >>= 6;
-
-            HEVC_RND_W_CLIP_UNSIGNED_CHAR_W_VEC2(dst0_r, dst0_l, dst1_r, dst1_l,
-                                                 dst0_r, dst1_r);
-
-            dst0_r = (v4i32) __msa_pckev_b((v16i8) dst1_r, (v16i8) dst0_r);
-
-            out0 = __msa_copy_u_d((v2i64) dst0_r, 0);
-            out1 = __msa_copy_u_d((v2i64) dst0_r, 1);
-
-            STORE_DWORD(dst_tmp, out0);
-            dst_tmp += dst_stride;
-            STORE_DWORD(dst_tmp, out1);
-            dst_tmp += dst_stride;
-
-            dst10_r = dst32_r;
-            dst32_r = dst54_r;
-            dst54_r = dst76_r;
-
-            dst10_l = dst32_l;
-            dst32_l = dst54_l;
-            dst54_l = dst76_l;
-
-            dst21_r = dst43_r;
-            dst43_r = dst65_r;
-            dst65_r = dst87_r;
-
-            dst21_l = dst43_l;
-            dst43_l = dst65_l;
-            dst65_l = dst87_l;
-
-            dst6 = dst8;
-        }
-
-        src += 8;
-        dst += 8;
-    }
-}
-
-static void hevc_hv_uni_8t_8w_msa(uint8_t * __restrict src,
-                                  int32_t src_stride,
-                                  uint8_t * __restrict dst,
-                                  int32_t dst_stride,
-                                  const int8_t * __restrict filter_x,
-                                  const int8_t * __restrict filter_y,
-                                  int32_t height)
-{
-    hevc_hv_uni_8t_8multx2mult_msa(src, src_stride, dst, dst_stride,
-                                   filter_x, filter_y, height, 8);
-}
-
-static void hevc_hv_uni_8t_12w_msa(uint8_t * __restrict src,
-                                   int32_t src_stride,
-                                   uint8_t * __restrict dst,
-                                   int32_t dst_stride,
-                                   const int8_t * __restrict filter_x,
-                                   const int8_t * __restrict filter_y,
-                                   int32_t height)
-{
-    hevc_hv_uni_8t_8multx2mult_msa(src, src_stride, dst, dst_stride,
-                                   filter_x, filter_y, height, 8);
-
-    hevc_hv_uni_8t_4w_msa(src + 8, src_stride, dst + 8, dst_stride,
-                          filter_x, filter_y, height);
-}
-
-static void hevc_hv_uni_8t_16w_msa(uint8_t * __restrict src,
-                                   int32_t src_stride,
-                                   uint8_t * __restrict dst,
-                                   int32_t dst_stride,
-                                   const int8_t * __restrict filter_x,
-                                   const int8_t * __restrict filter_y,
-                                   int32_t height)
-{
-    hevc_hv_uni_8t_8multx2mult_msa(src, src_stride, dst, dst_stride,
-                                   filter_x, filter_y, height, 16);
-}
-
-static void hevc_hv_uni_8t_24w_msa(uint8_t * __restrict src,
-                                   int32_t src_stride,
-                                   uint8_t * __restrict dst,
-                                   int32_t dst_stride,
-                                   const int8_t * __restrict filter_x,
-                                   const int8_t * __restrict filter_y,
-                                   int32_t height)
-{
-    hevc_hv_uni_8t_8multx2mult_msa(src, src_stride, dst, dst_stride,
-                                   filter_x, filter_y, height, 24);
-}
-
-static void hevc_hv_uni_8t_32w_msa(uint8_t * __restrict src,
-                                   int32_t src_stride,
-                                   uint8_t * __restrict dst,
-                                   int32_t dst_stride,
-                                   const int8_t * __restrict filter_x,
-                                   const int8_t * __restrict filter_y,
-                                   int32_t height)
-{
-    hevc_hv_uni_8t_8multx2mult_msa(src, src_stride, dst, dst_stride,
-                                   filter_x, filter_y, height, 32);
-}
-
-static void hevc_hv_uni_8t_48w_msa(uint8_t * __restrict src,
-                                   int32_t src_stride,
-                                   uint8_t * __restrict dst,
-                                   int32_t dst_stride,
-                                   const int8_t * __restrict filter_x,
-                                   const int8_t * __restrict filter_y,
-                                   int32_t height)
-{
-    hevc_hv_uni_8t_8multx2mult_msa(src, src_stride, dst, dst_stride,
-                                   filter_x, filter_y, height, 48);
-}
-
-static void hevc_hv_uni_8t_64w_msa(uint8_t * __restrict src,
-                                   int32_t src_stride,
-                                   uint8_t * __restrict dst,
-                                   int32_t dst_stride,
-                                   const int8_t * __restrict filter_x,
-                                   const int8_t * __restrict filter_y,
-                                   int32_t height)
-{
-    hevc_hv_uni_8t_8multx2mult_msa(src, src_stride, dst, dst_stride,
-                                   filter_x, filter_y, height, 64);
-}
-
-static uint8_t 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 FILT_8TAP_DPADD_S_H(vec0, vec1, vec2, vec3,                 \
-                            filt0, filt1, filt2, filt3)             \
-( {                                                                 \
-    v8i16 tmp0, tmp1;                                               \
-                                                                    \
-    tmp0 = __msa_dotp_s_h((v16i8) (vec0), (v16i8) (filt0));         \
-    tmp0 = __msa_dpadd_s_h(tmp0, (v16i8) (vec1), (v16i8) (filt1));  \
-    tmp1 = __msa_dotp_s_h((v16i8) (vec2), (v16i8) (filt2));         \
-    tmp1 = __msa_dpadd_s_h(tmp1, (v16i8) (vec3), (v16i8) (filt3));  \
-    tmp0 = __msa_adds_s_h(tmp0, tmp1);                              \
-                                                                    \
-    tmp0;                                                           \
-} )
-
-#define HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3,                   \
-                                   mask0, mask1, mask2, mask3,               \
-                                   filt0, filt1, filt2, filt3,               \
-                                   out0, out1)                               \
-{                                                                            \
-    v16i8 vec0_m, vec1_m, vec2_m, vec3_m,  vec4_m, vec5_m, vec6_m, vec7_m;   \
-    v8i16 res0_m, res1_m, res2_m, res3_m;                                    \
-                                                                             \
-    vec0_m = __msa_vshf_b((v16i8) (mask0), (v16i8) (src1), (v16i8) (src0));  \
-    vec1_m = __msa_vshf_b((v16i8) (mask0), (v16i8) (src3), (v16i8) (src2));  \
-                                                                             \
-    res0_m = __msa_dotp_s_h(vec0_m, (v16i8) (filt0));                        \
-    res1_m = __msa_dotp_s_h(vec1_m, (v16i8) (filt0));                        \
-                                                                             \
-    vec2_m = __msa_vshf_b((v16i8) (mask1), (v16i8) (src1), (v16i8) (src0));  \
-    vec3_m = __msa_vshf_b((v16i8) (mask1), (v16i8) (src3), (v16i8) (src2));  \
-                                                                             \
-    res0_m = __msa_dpadd_s_h(res0_m, (filt1), vec2_m);                       \
-    res1_m = __msa_dpadd_s_h(res1_m, (filt1), vec3_m);                       \
-                                                                             \
-    vec4_m = __msa_vshf_b((v16i8) (mask2), (v16i8) (src1), (v16i8) (src0));  \
-    vec5_m = __msa_vshf_b((v16i8) (mask2), (v16i8) (src3), (v16i8) (src2));  \
-                                                                             \
-    res2_m = __msa_dotp_s_h((v16i8) (filt2), vec4_m);                        \
-    res3_m = __msa_dotp_s_h((v16i8) (filt2), vec5_m);                        \
-                                                                             \
-    vec6_m = __msa_vshf_b((v16i8) (mask3), (v16i8) (src1), (v16i8) (src0));  \
-    vec7_m = __msa_vshf_b((v16i8) (mask3), (v16i8) (src3), (v16i8) (src2));  \
-                                                                             \
-    res2_m = __msa_dpadd_s_h(res2_m, (v16i8) (filt3), vec6_m);               \
-    res3_m = __msa_dpadd_s_h(res3_m, (v16i8) (filt3), vec7_m);               \
-                                                                             \
-    out0 = __msa_adds_s_h(res0_m, res2_m);                                   \
-    out1 = __msa_adds_s_h(res1_m, res3_m);                                   \
-}
-
-#define HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3,                   \
-                                   mask0, mask1, mask2, mask3,               \
-                                   filt0, filt1, filt2, filt3,               \
-                                   out0, out1, out2, out3)                   \
-{                                                                            \
-    v16i8 vec0_m, vec1_m, vec2_m, vec3_m, vec4_m, vec5_m, vec6_m, vec7_m;    \
-    v8i16 res0_m, res1_m, res2_m, res3_m, res4_m, res5_m, res6_m, res7_m;    \
-                                                                             \
-    vec0_m = __msa_vshf_b((v16i8) (mask0), (v16i8) (src0), (v16i8) (src0));  \
-    vec1_m = __msa_vshf_b((v16i8) (mask0), (v16i8) (src1), (v16i8) (src1));  \
-    vec2_m = __msa_vshf_b((v16i8) (mask0), (v16i8) (src2), (v16i8) (src2));  \
-    vec3_m = __msa_vshf_b((v16i8) (mask0), (v16i8) (src3), (v16i8) (src3));  \
-                                                                             \
-    res0_m = __msa_dotp_s_h(vec0_m, (v16i8) (filt0));                        \
-    res1_m = __msa_dotp_s_h(vec1_m, (v16i8) (filt0));                        \
-    res2_m = __msa_dotp_s_h(vec2_m, (v16i8) (filt0));                        \
-    res3_m = __msa_dotp_s_h(vec3_m, (v16i8) (filt0));                        \
-                                                                             \
-    vec0_m = __msa_vshf_b((v16i8) (mask2), (v16i8) (src0), (v16i8) (src0));  \
-    vec1_m = __msa_vshf_b((v16i8) (mask2), (v16i8) (src1), (v16i8) (src1));  \
-    vec2_m = __msa_vshf_b((v16i8) (mask2), (v16i8) (src2), (v16i8) (src2));  \
-    vec3_m = __msa_vshf_b((v16i8) (mask2), (v16i8) (src3), (v16i8) (src3));  \
-                                                                             \
-    res4_m = __msa_dotp_s_h(vec0_m, (v16i8) (filt2));                        \
-    res5_m = __msa_dotp_s_h(vec1_m, (v16i8) (filt2));                        \
-    res6_m = __msa_dotp_s_h(vec2_m, (v16i8) (filt2));                        \
-    res7_m = __msa_dotp_s_h(vec3_m, (v16i8) (filt2));                        \
-                                                                             \
-    vec4_m = __msa_vshf_b((v16i8) (mask1), (v16i8) (src0), (v16i8) (src0));  \
-    vec5_m = __msa_vshf_b((v16i8) (mask1), (v16i8) (src1), (v16i8) (src1));  \
-    vec6_m = __msa_vshf_b((v16i8) (mask1), (v16i8) (src2), (v16i8) (src2));  \
-    vec7_m = __msa_vshf_b((v16i8) (mask1), (v16i8) (src3), (v16i8) (src3));  \
-                                                                             \
-    res0_m = __msa_dpadd_s_h(res0_m, (v16i8) (filt1), vec4_m);               \
-    res1_m = __msa_dpadd_s_h(res1_m, (v16i8) (filt1), vec5_m);               \
-    res2_m = __msa_dpadd_s_h(res2_m, (v16i8) (filt1), vec6_m);               \
-    res3_m = __msa_dpadd_s_h(res3_m, (v16i8) (filt1), vec7_m);               \
-                                                                             \
-    vec4_m = __msa_vshf_b((v16i8) (mask3), (v16i8) (src0), (v16i8) (src0));  \
-    vec5_m = __msa_vshf_b((v16i8) (mask3), (v16i8) (src1), (v16i8) (src1));  \
-    vec6_m = __msa_vshf_b((v16i8) (mask3), (v16i8) (src2), (v16i8) (src2));  \
-    vec7_m = __msa_vshf_b((v16i8) (mask3), (v16i8) (src3), (v16i8) (src3));  \
-                                                                             \
-    res4_m = __msa_dpadd_s_h(res4_m, (v16i8) (filt3), vec4_m);               \
-    res5_m = __msa_dpadd_s_h(res5_m, (v16i8) (filt3), vec5_m);               \
-    res6_m = __msa_dpadd_s_h(res6_m, (v16i8) (filt3), vec6_m);               \
-    res7_m = __msa_dpadd_s_h(res7_m, (v16i8) (filt3), vec7_m);               \
-                                                                             \
-    out0 = __msa_adds_s_h(res0_m, res4_m);                                   \
-    out1 = __msa_adds_s_h(res1_m, res5_m);                                   \
-    out2 = __msa_adds_s_h(res2_m, res6_m);                                   \
-    out3 = __msa_adds_s_h(res3_m, res7_m);                                   \
-}
-
-static void common_hz_8t_4x4_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
-                                 const int8_t *filter, uint8_t rnd_val)
-{
-    v16i8 filt0, filt1, filt2, filt3;
-    v16i8 src0, src1, src2, src3;
-    v16u8 mask0, mask1, mask2, mask3;
-    v8i16 filt, out0, out1;
-    v8u16 rnd_vec;
-
-    mask0 = LOAD_UB(&mc_filt_mask_arr[16]);
-
-    src -= 3;
-
-    /* rearranging filter */
-    filt = LOAD_SH(filter);
-    filt0 = (v16i8) __msa_splati_h(filt, 0);
-    filt1 = (v16i8) __msa_splati_h(filt, 1);
-    filt2 = (v16i8) __msa_splati_h(filt, 2);
-    filt3 = (v16i8) __msa_splati_h(filt, 3);
-
-    mask1 = mask0 + 2;
-    mask2 = mask0 + 4;
-    mask3 = mask0 + 6;
-
-    LOAD_4VECS_SB(src, src_stride, src0, src1, src2, src3);
-
-    XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-    HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
-                               mask3, filt0, filt1, filt2, filt3, out0, out1);
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-    out0 = SRAR_SATURATE_SIGNED_H(out0, rnd_vec, 7);
-    out1 = SRAR_SATURATE_SIGNED_H(out1, rnd_vec, 7);
-
-    PCKEV_2B_XORI128_STORE_4_BYTES_4(out0, out1, dst, dst_stride);
-}
-
-static void common_hz_8t_4x8_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
-                                 const int8_t *filter, uint8_t rnd_val)
-{
-    v16i8 filt0, filt1, filt2, filt3;
-    v16i8 src0, src1, src2, src3;
-    v16u8 mask0, mask1, mask2, mask3;
-    v8i16 filt, out0, out1, out2, out3;
-    v8u16 rnd_vec;
-
-    mask0 = LOAD_UB(&mc_filt_mask_arr[16]);
-
-    src -= 3;
-
-    /* rearranging filter */
-    filt = LOAD_SH(filter);
-    filt0 = (v16i8) __msa_splati_h(filt, 0);
-    filt1 = (v16i8) __msa_splati_h(filt, 1);
-    filt2 = (v16i8) __msa_splati_h(filt, 2);
-    filt3 = (v16i8) __msa_splati_h(filt, 3);
-
-    mask1 = mask0 + 2;
-    mask2 = mask0 + 4;
-    mask3 = mask0 + 6;
-
-    LOAD_4VECS_SB(src, src_stride, src0, src1, src2, src3);
-    src += (4 * src_stride);
-
-    XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-    HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
-                               mask3, filt0, filt1, filt2, filt3, out0, out1);
-
-    LOAD_4VECS_SB(src, src_stride, src0, src1, src2, src3);
-
-    XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-    HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
-                               mask3, filt0, filt1, filt2, filt3, out2, out3);
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-    out0 = SRAR_SATURATE_SIGNED_H(out0, rnd_vec, 7);
-    out1 = SRAR_SATURATE_SIGNED_H(out1, rnd_vec, 7);
-    out2 = SRAR_SATURATE_SIGNED_H(out2, rnd_vec, 7);
-    out3 = SRAR_SATURATE_SIGNED_H(out3, rnd_vec, 7);
-
-    PCKEV_2B_XORI128_STORE_4_BYTES_4(out0, out1, dst, dst_stride);
-    dst += (4 * dst_stride);
-    PCKEV_2B_XORI128_STORE_4_BYTES_4(out2, out3, dst, dst_stride);
-}
-
-static void common_hz_8t_4x16_msa(uint8_t *src, int32_t src_stride,
-                                  uint8_t *dst, int32_t dst_stride,
-                                  const int8_t *filter, uint8_t rnd_val)
-{
-    v16i8 filt0, filt1, filt2, filt3;
-    v16i8 src0, src1, src2, src3;
-    v16u8 mask0, mask1, mask2, mask3;
-    v8i16 filt, out0, out1, out2, out3;
-    v8u16 rnd_vec;
-
-    mask0 = LOAD_UB(&mc_filt_mask_arr[16]);
-
-    src -= 3;
-
-    /* rearranging filter */
-    filt = LOAD_SH(filter);
-    filt0 = (v16i8) __msa_splati_h(filt, 0);
-    filt1 = (v16i8) __msa_splati_h(filt, 1);
-    filt2 = (v16i8) __msa_splati_h(filt, 2);
-    filt3 = (v16i8) __msa_splati_h(filt, 3);
-
-    mask1 = mask0 + 2;
-    mask2 = mask0 + 4;
-    mask3 = mask0 + 6;
-
-    LOAD_4VECS_SB(src, src_stride, src0, src1, src2, src3);
-    src += (4 * src_stride);
-
-    XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-    HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
-                               mask3, filt0, filt1, filt2, filt3, out0, out1);
-
-    LOAD_4VECS_SB(src, src_stride, src0, src1, src2, src3);
-    src += (4 * src_stride);
-
-    XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-    HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
-                               mask3, filt0, filt1, filt2, filt3, out2, out3);
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-    out0 = SRAR_SATURATE_SIGNED_H(out0, rnd_vec, 7);
-    out1 = SRAR_SATURATE_SIGNED_H(out1, rnd_vec, 7);
-    out2 = SRAR_SATURATE_SIGNED_H(out2, rnd_vec, 7);
-    out3 = SRAR_SATURATE_SIGNED_H(out3, rnd_vec, 7);
-
-    PCKEV_2B_XORI128_STORE_4_BYTES_4(out0, out1, dst, dst_stride);
-    dst += (4 * dst_stride);
-    PCKEV_2B_XORI128_STORE_4_BYTES_4(out2, out3, dst, dst_stride);
-    dst += (4 * dst_stride);
-
-    LOAD_4VECS_SB(src, src_stride, src0, src1, src2, src3);
-    src += (4 * src_stride);
-
-    XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-    HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
-                               mask3, filt0, filt1, filt2, filt3, out0, out1);
-
-    LOAD_4VECS_SB(src, src_stride, src0, src1, src2, src3);
-    src += (4 * src_stride);
-
-    XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-    HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
-                               mask3, filt0, filt1, filt2, filt3, out2, out3);
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-    out0 = SRAR_SATURATE_SIGNED_H(out0, rnd_vec, 7);
-    out1 = SRAR_SATURATE_SIGNED_H(out1, rnd_vec, 7);
-    out2 = SRAR_SATURATE_SIGNED_H(out2, rnd_vec, 7);
-    out3 = SRAR_SATURATE_SIGNED_H(out3, rnd_vec, 7);
-
-    PCKEV_2B_XORI128_STORE_4_BYTES_4(out0, out1, dst, dst_stride);
-    dst += (4 * dst_stride);
-    PCKEV_2B_XORI128_STORE_4_BYTES_4(out2, out3, dst, dst_stride);
-}
-
-static void common_hz_8t_4w_msa(uint8_t *src, int32_t src_stride,
-                                uint8_t *dst, int32_t dst_stride,
-                                const int8_t *filter, int32_t height,
-                                uint8_t rnd_val)
-{
-    if (4 == height) {
-        common_hz_8t_4x4_msa(src, src_stride, dst, dst_stride, filter, rnd_val);
-    } else if (8 == height) {
-        common_hz_8t_4x8_msa(src, src_stride, dst, dst_stride, filter, rnd_val);
-    } else if (16 == height) {
-        common_hz_8t_4x16_msa(src, src_stride, dst, dst_stride, filter,
-                              rnd_val);
-    }
-}
-
-static void common_hz_8t_8x4_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
-                                 const int8_t *filter, uint8_t rnd_val)
-{
-    v16i8 filt0, filt1, filt2, filt3;
-    v16i8 src0, src1, src2, src3;
-    v16u8 mask0, mask1, mask2, mask3;
-    v8i16 filt, out0, out1, out2, out3;
-    v8u16 rnd_vec;
-
-    mask0 = LOAD_UB(&mc_filt_mask_arr[0]);
-
-    src -= 3;
-
-    /* rearranging filter */
-    filt = LOAD_SH(filter);
-    filt0 = (v16i8) __msa_splati_h(filt, 0);
-    filt1 = (v16i8) __msa_splati_h(filt, 1);
-    filt2 = (v16i8) __msa_splati_h(filt, 2);
-    filt3 = (v16i8) __msa_splati_h(filt, 3);
-
-    mask1 = mask0 + 2;
-    mask2 = mask0 + 4;
-    mask3 = mask0 + 6;
-
-    LOAD_4VECS_SB(src, src_stride, src0, src1, src2, src3);
-
-    XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-    HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
-                               mask3, filt0, filt1, filt2, filt3, out0, out1,
-                               out2, out3);
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-    out0 = SRAR_SATURATE_SIGNED_H(out0, rnd_vec, 7);
-    out1 = SRAR_SATURATE_SIGNED_H(out1, rnd_vec, 7);
-    out2 = SRAR_SATURATE_SIGNED_H(out2, rnd_vec, 7);
-    out3 = SRAR_SATURATE_SIGNED_H(out3, rnd_vec, 7);
-
-    PCKEV_B_4_XORI128_STORE_8_BYTES_4(out0, out1, out2, out3, dst, dst_stride);
-}
-
-static void common_hz_8t_8x8mult_msa(uint8_t *src, int32_t src_stride,
-                                     uint8_t *dst, int32_t dst_stride,
-                                     const int8_t *filter, int32_t height,
-                                     uint8_t rnd_val)
-{
-    uint32_t loop_cnt;
-    v16i8 filt0, filt1, filt2, filt3;
-    v16i8 src0, src1, src2, src3;
-    v16u8 mask0, mask1, mask2, mask3;
-    v8i16 filt, out0, out1, out2, out3;
-    v8u16 rnd_vec;
-
-    mask0 = LOAD_UB(&mc_filt_mask_arr[0]);
-
-    src -= 3;
-
-    /* rearranging filter */
-    filt = LOAD_SH(filter);
-    filt0 = (v16i8) __msa_splati_h(filt, 0);
-    filt1 = (v16i8) __msa_splati_h(filt, 1);
-    filt2 = (v16i8) __msa_splati_h(filt, 2);
-    filt3 = (v16i8) __msa_splati_h(filt, 3);
-
-    mask1 = mask0 + 2;
-    mask2 = mask0 + 4;
-    mask3 = mask0 + 6;
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-
-    for (loop_cnt = (height >> 2); loop_cnt--;) {
-        LOAD_4VECS_SB(src, src_stride, src0, src1, src2, src3);
-        src += (4 * src_stride);
-
-        XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-        HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
-                                   mask3, filt0, filt1, filt2, filt3, out0,
-                                   out1, out2, out3);
-
-        out0 = SRAR_SATURATE_SIGNED_H(out0, rnd_vec, 7);
-        out1 = SRAR_SATURATE_SIGNED_H(out1, rnd_vec, 7);
-        out2 = SRAR_SATURATE_SIGNED_H(out2, rnd_vec, 7);
-        out3 = SRAR_SATURATE_SIGNED_H(out3, rnd_vec, 7);
-
-        PCKEV_B_4_XORI128_STORE_8_BYTES_4(out0, out1, out2, out3,
-                                          dst, dst_stride);
-        dst += (4 * dst_stride);
-    }
-}
-
-static void common_hz_8t_8w_msa(uint8_t *src, int32_t src_stride,
-                                uint8_t *dst, int32_t dst_stride,
-                                const int8_t *filter, int32_t height,
-                                uint8_t rnd_val)
-{
-    if (4 == height) {
-        common_hz_8t_8x4_msa(src, src_stride, dst, dst_stride, filter, rnd_val);
-    } else {
-        common_hz_8t_8x8mult_msa(src, src_stride, dst, dst_stride,
-                                 filter, height, rnd_val);
-    }
-}
-
-static void common_hz_8t_12w_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
-                                 const int8_t *filter, int32_t height,
-                                 uint8_t rnd_val)
-{
-    uint8_t *src1_ptr, *dst1;
-    uint32_t loop_cnt;
-    v16i8 filt0, filt1, filt2, filt3;
-    v16i8 src0, src1, src2, src3;
-    v8u16 rnd_vec;
-    v16u8 mask0, mask1, mask2, mask3, mask4, mask5, mask6, mask00;
-    v8i16 filt, out0, out1, out2, out3;
-
-    mask00 = LOAD_UB(&mc_filt_mask_arr[0]);
-    mask0 = LOAD_UB(&mc_filt_mask_arr[16]);
-
-    src1_ptr = src - 3;
-    dst1 = dst;
-
-    dst = dst1 + 8;
-    src = src1_ptr + 8;
-
-    /* rearranging filter */
-    filt = LOAD_SH(filter);
-    filt0 = (v16i8) __msa_splati_h(filt, 0);
-    filt1 = (v16i8) __msa_splati_h(filt, 1);
-    filt2 = (v16i8) __msa_splati_h(filt, 2);
-    filt3 = (v16i8) __msa_splati_h(filt, 3);
-
-    mask1 = mask00 + 2;
-    mask2 = mask00 + 4;
-    mask3 = mask00 + 6;
-    mask4 = mask0 + 2;
-    mask5 = mask0 + 4;
-    mask6 = mask0 + 6;
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-
-    for (loop_cnt = (height >> 2); loop_cnt--;) {
-        /* 8 width */
-        LOAD_4VECS_SB(src1_ptr, src_stride, src0, src1, src2, src3);
-        src1_ptr += (4 * src_stride);
-
-        XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-        HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask00, mask1, mask2,
-                                   mask3, filt0, filt1, filt2, filt3, out0,
-                                   out1, out2, out3);
-
-        out0 = SRAR_SATURATE_SIGNED_H(out0, rnd_vec, 7);
-        out1 = SRAR_SATURATE_SIGNED_H(out1, rnd_vec, 7);
-        out2 = SRAR_SATURATE_SIGNED_H(out2, rnd_vec, 7);
-        out3 = SRAR_SATURATE_SIGNED_H(out3, rnd_vec, 7);
-
-        PCKEV_B_4_XORI128_STORE_8_BYTES_4(out0, out1, out2, out3,
-                                          dst1, dst_stride);
-        dst1 += (4 * dst_stride);
-
-        /* 4 width */
-        LOAD_4VECS_SB(src, src_stride, src0, src1, src2, src3);
-        src += (4 * src_stride);
-
-        XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-        HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask4, mask5,
-                                   mask6, filt0, filt1, filt2, filt3, out0,
-                                   out1);
-
-        out0 = SRAR_SATURATE_SIGNED_H(out0, rnd_vec, 7);
-        out1 = SRAR_SATURATE_SIGNED_H(out1, rnd_vec, 7);
-
-        PCKEV_2B_XORI128_STORE_4_BYTES_4(out0, out1, dst, dst_stride);
-        dst += (4 * dst_stride);
-    }
-}
-
-static void common_hz_8t_16w_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
-                                 const int8_t *filter, int32_t height,
-                                 uint8_t rnd_val)
-{
-    uint32_t loop_cnt;
-    v16i8 src0, src1, src2, src3;
-    v16i8 filt0, filt1, filt2, filt3;
-    v16u8 mask0, mask1, mask2, mask3;
-    v8i16 filt, out0, out1, out2, out3;
-    v8u16 rnd_vec;
-
-    mask0 = LOAD_UB(&mc_filt_mask_arr[0]);
-
-    src -= 3;
-
-    /* rearranging filter */
-    filt = LOAD_SH(filter);
-    filt0 = (v16i8) __msa_splati_h(filt, 0);
-    filt1 = (v16i8) __msa_splati_h(filt, 1);
-    filt2 = (v16i8) __msa_splati_h(filt, 2);
-    filt3 = (v16i8) __msa_splati_h(filt, 3);
-
-    mask1 = mask0 + 2;
-    mask2 = mask0 + 4;
-    mask3 = mask0 + 6;
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-
-    for (loop_cnt = (height >> 1); loop_cnt--;) {
-        src0 = LOAD_SB(src);
-        src1 = LOAD_SB(src + 8);
-        src += src_stride;
-        src2 = LOAD_SB(src);
-        src3 = LOAD_SB(src + 8);
-        src += src_stride;
-
-        XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-        HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
-                                   mask3, filt0, filt1, filt2, filt3, out0,
-                                   out1, out2, out3);
-
-        out0 = SRAR_SATURATE_SIGNED_H(out0, rnd_vec, 7);
-        out1 = SRAR_SATURATE_SIGNED_H(out1, rnd_vec, 7);
-        out2 = SRAR_SATURATE_SIGNED_H(out2, rnd_vec, 7);
-        out3 = SRAR_SATURATE_SIGNED_H(out3, rnd_vec, 7);
-
-        PCKEV_B_XORI128_STORE_VEC(out1, out0, dst);
-        dst += dst_stride;
-        PCKEV_B_XORI128_STORE_VEC(out3, out2, dst);
-        dst += dst_stride;
-    }
-}
-
-static void common_hz_8t_24w_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
-                                 const int8_t *filter, int32_t height,
-                                 uint8_t rnd_val)
-{
-    uint32_t loop_cnt;
-    v16i8 src0, src1, src2, src3;
-    v16i8 filt0, filt1, filt2, filt3;
-    v16i8 mask0, mask1, mask2, mask3, mask4, mask5, mask6, mask7;
-    v16i8 vec0, vec1, vec2, vec3, vec4, vec5;
-    v16i8 vec6, vec7, vec8, vec9, vec10, vec11;
-    v8i16 out0, out1, out2, out3, out4, out5;
-    v8i16 out6, out7, out8, out9, out10, out11;
-    v8i16 filt;
-    v8u16 rnd_vec;
-
-    mask0 = LOAD_SB(&mc_filt_mask_arr[0]);
-
-    src -= 3;
-
-    /* rearranging filter */
-    filt = LOAD_SH(filter);
-    filt0 = (v16i8) __msa_splati_h(filt, 0);
-    filt1 = (v16i8) __msa_splati_h(filt, 1);
-    filt2 = (v16i8) __msa_splati_h(filt, 2);
-    filt3 = (v16i8) __msa_splati_h(filt, 3);
-
-    mask1 = mask0 + 2;
-    mask2 = mask0 + 4;
-    mask3 = mask0 + 6;
-    mask4 = mask0 + 8;
-    mask5 = mask0 + 10;
-    mask6 = mask0 + 12;
-    mask7 = mask0 + 14;
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-
-    for (loop_cnt = (height >> 1); loop_cnt--;) {
-        src0 = LOAD_SB(src);
-        src1 = LOAD_SB(src + 16);
-        src += src_stride;
-        src2 = LOAD_SB(src);
-        src3 = LOAD_SB(src + 16);
-        src += src_stride;
-
-        XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-        vec0 = __msa_vshf_b(mask0, src0, src0);
-        vec8 = __msa_vshf_b(mask0, src1, src1);
-        vec2 = __msa_vshf_b(mask0, src2, src2);
-        vec9 = __msa_vshf_b(mask0, src3, src3);
-        vec1 = __msa_vshf_b(mask4, src1, src0);
-        vec3 = __msa_vshf_b(mask4, src3, src2);
-
-        out0 = __msa_dotp_s_h(vec0, filt0);
-        out8 = __msa_dotp_s_h(vec8, filt0);
-        out2 = __msa_dotp_s_h(vec2, filt0);
-        out9 = __msa_dotp_s_h(vec9, filt0);
-        out1 = __msa_dotp_s_h(vec1, filt0);
-        out3 = __msa_dotp_s_h(vec3, filt0);
-
-        vec0 = __msa_vshf_b(mask2, src0, src0);
-        vec8 = __msa_vshf_b(mask2, src1, src1);
-        vec2 = __msa_vshf_b(mask2, src2, src2);
-        vec9 = __msa_vshf_b(mask2, src3, src3);
-        vec1 = __msa_vshf_b(mask6, src1, src0);
-        vec3 = __msa_vshf_b(mask6, src3, src2);
-
-        out4 = __msa_dotp_s_h(vec0, filt2);
-        out10 = __msa_dotp_s_h(vec8, filt2);
-        out6 = __msa_dotp_s_h(vec2, filt2);
-        out11 = __msa_dotp_s_h(vec9, filt2);
-        out5 = __msa_dotp_s_h(vec1, filt2);
-        out7 = __msa_dotp_s_h(vec3, filt2);
-
-        vec4 = __msa_vshf_b(mask1, src0, src0);
-        vec10 = __msa_vshf_b(mask1, src1, src1);
-        vec6 = __msa_vshf_b(mask1, src2, src2);
-        vec11 = __msa_vshf_b(mask1, src3, src3);
-        vec5 = __msa_vshf_b(mask5, src1, src0);
-        vec7 = __msa_vshf_b(mask5, src3, src2);
-
-        out0 = __msa_dpadd_s_h(out0, vec4, filt1);
-        out8 = __msa_dpadd_s_h(out8, vec10, filt1);
-        out2 = __msa_dpadd_s_h(out2, vec6, filt1);
-        out9 = __msa_dpadd_s_h(out9, vec11, filt1);
-        out1 = __msa_dpadd_s_h(out1, vec5, filt1);
-        out3 = __msa_dpadd_s_h(out3, vec7, filt1);
-
-        vec4 = __msa_vshf_b(mask3, src0, src0);
-        vec10 = __msa_vshf_b(mask3, src1, src1);
-        vec6 = __msa_vshf_b(mask3, src2, src2);
-        vec11 = __msa_vshf_b(mask3, src3, src3);
-        vec5 = __msa_vshf_b(mask7, src1, src0);
-        vec7 = __msa_vshf_b(mask7, src3, src2);
-
-        out4 = __msa_dpadd_s_h(out4, vec4, filt3);
-        out10 = __msa_dpadd_s_h(out10, vec10, filt3);
-        out6 = __msa_dpadd_s_h(out6, vec6, filt3);
-        out11 = __msa_dpadd_s_h(out11, vec11, filt3);
-        out5 = __msa_dpadd_s_h(out5, vec5, filt3);
-        out7 = __msa_dpadd_s_h(out7, vec7, filt3);
-
-        out0 = __msa_adds_s_h(out0, out4);
-        out8 = __msa_adds_s_h(out8, out10);
-        out2 = __msa_adds_s_h(out2, out6);
-        out9 = __msa_adds_s_h(out9, out11);
-        out1 = __msa_adds_s_h(out1, out5);
-        out3 = __msa_adds_s_h(out3, out7);
-
-        out0 = SRAR_SATURATE_SIGNED_H(out0, rnd_vec, 7);
-        out8 = SRAR_SATURATE_SIGNED_H(out8, rnd_vec, 7);
-        out2 = SRAR_SATURATE_SIGNED_H(out2, rnd_vec, 7);
-        out9 = SRAR_SATURATE_SIGNED_H(out9, rnd_vec, 7);
-        out1 = SRAR_SATURATE_SIGNED_H(out1, rnd_vec, 7);
-        out3 = SRAR_SATURATE_SIGNED_H(out3, rnd_vec, 7);
-
-        PCKEV_B_XORI128_STORE_8_BYTES_2(out8, out9, dst + 16, dst_stride);
-
-        PCKEV_B_XORI128_STORE_VEC(out1, out0, dst);
-        dst += dst_stride;
-        PCKEV_B_XORI128_STORE_VEC(out3, out2, dst);
-        dst += dst_stride;
-    }
-}
-
-static void common_hz_8t_32w_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
-                                 const int8_t *filter, int32_t height,
-                                 uint8_t rnd_val)
-{
-    uint32_t loop_cnt;
-    v16i8 src0, src1, src2, src3;
-    v16i8 filt0, filt1, filt2, filt3;
-    v16u8 mask0, mask1, mask2, mask3;
-    v8i16 filt, out0, out1, out2, out3;
-    v8u16 rnd_vec;
-
-    mask0 = LOAD_UB(&mc_filt_mask_arr[0]);
-
-    src -= 3;
-
-    /* rearranging filter */
-    filt = LOAD_SH(filter);
-    filt0 = (v16i8) __msa_splati_h(filt, 0);
-    filt1 = (v16i8) __msa_splati_h(filt, 1);
-    filt2 = (v16i8) __msa_splati_h(filt, 2);
-    filt3 = (v16i8) __msa_splati_h(filt, 3);
-
-    mask1 = mask0 + 2;
-    mask2 = mask0 + 4;
-    mask3 = mask0 + 6;
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-
-    for (loop_cnt = (height >> 1); loop_cnt--;) {
-        src0 = LOAD_SB(src);
-        src2 = LOAD_SB(src + 16);
-        src3 = LOAD_SB(src + 24);
-        src1 = __msa_sld_b(src2, src0, 8);
-        src += src_stride;
-
-        XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-        HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
-                                   mask3, filt0, filt1, filt2, filt3, out0,
-                                   out1, out2, out3);
-
-        out0 = SRAR_SATURATE_SIGNED_H(out0, rnd_vec, 7);
-        out1 = SRAR_SATURATE_SIGNED_H(out1, rnd_vec, 7);
-        out2 = SRAR_SATURATE_SIGNED_H(out2, rnd_vec, 7);
-        out3 = SRAR_SATURATE_SIGNED_H(out3, rnd_vec, 7);
-
-        src0 = LOAD_SB(src);
-        src2 = LOAD_SB(src + 16);
-        src3 = LOAD_SB(src + 24);
-        src1 = __msa_sld_b(src2, src0, 8);
-        src += src_stride;
-
-        PCKEV_B_XORI128_STORE_VEC(out1, out0, dst);
-        PCKEV_B_XORI128_STORE_VEC(out3, out2, (dst + 16));
-        dst += dst_stride;
-
-        XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-        HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1, mask2,
-                                   mask3, filt0, filt1, filt2, filt3, out0,
-                                   out1, out2, out3);
-
-        out0 = SRAR_SATURATE_SIGNED_H(out0, rnd_vec, 7);
-        out1 = SRAR_SATURATE_SIGNED_H(out1, rnd_vec, 7);
-        out2 = SRAR_SATURATE_SIGNED_H(out2, rnd_vec, 7);
-        out3 = SRAR_SATURATE_SIGNED_H(out3, rnd_vec, 7);
-
-        PCKEV_B_XORI128_STORE_VEC(out1, out0, dst);
-        PCKEV_B_XORI128_STORE_VEC(out3, out2, (dst + 16));
-        dst += dst_stride;
-    }
-}
-
-static void common_hz_8t_48w_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
-                                 const int8_t *filter, int32_t height,
-                                 uint8_t rnd_val)
-{
-    uint32_t loop_cnt;
-    v16i8 src0, src1, src2, src3;
-    v16i8 vec0, vec1, vec2;
-    v16i8 filt0, filt1, filt2, filt3;
-    v16i8 mask0, mask1, mask2, mask3, mask4, mask5, mask6, mask7;
-    v8i16 filt, out0, out1, out2, out3, out4, out5, out6;
-    v8u16 rnd_vec;
-
-    mask0 = LOAD_SB(&mc_filt_mask_arr[0]);
-
-    /* rearranging filter */
-    filt = LOAD_SH(filter);
-    filt0 = (v16i8) __msa_splati_h(filt, 0);
-    filt1 = (v16i8) __msa_splati_h(filt, 1);
-    filt2 = (v16i8) __msa_splati_h(filt, 2);
-    filt3 = (v16i8) __msa_splati_h(filt, 3);
-
-    mask1 = mask0 + 2;
-    mask2 = mask0 + 4;
-    mask3 = mask0 + 6;
-    mask4 = mask0 + 8;
-    mask5 = mask0 + 10;
-    mask6 = mask0 + 12;
-    mask7 = mask0 + 14;
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-
-    src -= 3;
-
-    for (loop_cnt = height; loop_cnt--;) {
-        src0 = LOAD_SB(src);
-        src2 = LOAD_SB(src + 16);
-        src3 = LOAD_SB(src + 32);
-        src1 = __msa_sld_b(src2, src0, 8);
-
-        XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-        vec0 = __msa_vshf_b(mask0, src0, src0);
-        vec1 = __msa_vshf_b(mask0, src1, src1);
-        vec2 = __msa_vshf_b(mask0, src2, src2);
-        out0 = __msa_dotp_s_h(vec0, filt0);
-        out1 = __msa_dotp_s_h(vec1, filt0);
-        out2 = __msa_dotp_s_h(vec2, filt0);
-
-        vec0 = __msa_vshf_b(mask1, src0, src0);
-        vec1 = __msa_vshf_b(mask1, src1, src1);
-        vec2 = __msa_vshf_b(mask1, src2, src2);
-        out0 = __msa_dpadd_s_h(out0, vec0, filt1);
-        out1 = __msa_dpadd_s_h(out1, vec1, filt1);
-        out2 = __msa_dpadd_s_h(out2, vec2, filt1);
-
-        vec0 = __msa_vshf_b(mask2, src0, src0);
-        vec1 = __msa_vshf_b(mask2, src1, src1);
-        vec2 = __msa_vshf_b(mask2, src2, src2);
-        out3 = __msa_dotp_s_h(vec0, filt2);
-        out4 = __msa_dotp_s_h(vec1, filt2);
-        out5 = __msa_dotp_s_h(vec2, filt2);
-
-        vec0 = __msa_vshf_b(mask3, src0, src0);
-        vec1 = __msa_vshf_b(mask3, src1, src1);
-        vec2 = __msa_vshf_b(mask3, src2, src2);
-        out3 = __msa_dpadd_s_h(out3, vec0, filt3);
-        out4 = __msa_dpadd_s_h(out4, vec1, filt3);
-        out5 = __msa_dpadd_s_h(out5, vec2, filt3);
-
-        out0 = __msa_adds_s_h(out0, out3);
-        out1 = __msa_adds_s_h(out1, out4);
-        out2 = __msa_adds_s_h(out2, out5);
-
-        out0 = SRAR_SATURATE_SIGNED_H(out0, rnd_vec, 7);
-        out1 = SRAR_SATURATE_SIGNED_H(out1, rnd_vec, 7);
-        out6 = SRAR_SATURATE_SIGNED_H(out2, rnd_vec, 7);
-
-        PCKEV_B_XORI128_STORE_VEC(out1, out0, dst);
-
-        src1 = LOAD_SB(src + 40);
-        src1 = (v16i8) __msa_xori_b((v16u8) src1, 128);
-
-        vec0 = __msa_vshf_b(mask4, src3, src2);
-        vec1 = __msa_vshf_b(mask0, src3, src3);
-        vec2 = __msa_vshf_b(mask0, src1, src1);
-        out0 = __msa_dotp_s_h(vec0, filt0);
-        out1 = __msa_dotp_s_h(vec1, filt0);
-        out2 = __msa_dotp_s_h(vec2, filt0);
-
-        vec0 = __msa_vshf_b(mask5, src3, src2);
-        vec1 = __msa_vshf_b(mask1, src3, src3);
-        vec2 = __msa_vshf_b(mask1, src1, src1);
-        out0 = __msa_dpadd_s_h(out0, vec0, filt1);
-        out1 = __msa_dpadd_s_h(out1, vec1, filt1);
-        out2 = __msa_dpadd_s_h(out2, vec2, filt1);
-
-        vec0 = __msa_vshf_b(mask6, src3, src2);
-        vec1 = __msa_vshf_b(mask2, src3, src3);
-        vec2 = __msa_vshf_b(mask2, src1, src1);
-        out3 = __msa_dotp_s_h(vec0, filt2);
-        out4 = __msa_dotp_s_h(vec1, filt2);
-        out5 = __msa_dotp_s_h(vec2, filt2);
-
-        vec0 = __msa_vshf_b(mask7, src3, src2);
-        vec1 = __msa_vshf_b(mask3, src3, src3);
-        vec2 = __msa_vshf_b(mask3, src1, src1);
-        out3 = __msa_dpadd_s_h(out3, vec0, filt3);
-        out4 = __msa_dpadd_s_h(out4, vec1, filt3);
-        out5 = __msa_dpadd_s_h(out5, vec2, filt3);
-
-        out3 = __msa_adds_s_h(out0, out3);
-        out4 = __msa_adds_s_h(out1, out4);
-        out5 = __msa_adds_s_h(out2, out5);
-
-        out3 = SRAR_SATURATE_SIGNED_H(out3, rnd_vec, 7);
-        out4 = SRAR_SATURATE_SIGNED_H(out4, rnd_vec, 7);
-        out5 = SRAR_SATURATE_SIGNED_H(out5, rnd_vec, 7);
-
-        PCKEV_B_XORI128_STORE_VEC(out3, out6, (dst + 16));
-        PCKEV_B_XORI128_STORE_VEC(out5, out4, (dst + 32));
-
-        src += src_stride;
-        dst += dst_stride;
-    }
-}
-
-static void common_hz_8t_64w_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
-                                 const int8_t *filter, int32_t height,
-                                 uint8_t rnd_val)
-{
-    int32_t loop_cnt;
-    v16i8 src0, src1, src2, src3;
-    v16i8 filt0, filt1, filt2, filt3;
-    v16u8 mask0, mask1, mask2, mask3;
-    v8i16 filt, out0, out1, out2, out3;
-    v8u16 rnd_vec;
-
-    mask0 = LOAD_UB(&mc_filt_mask_arr[0]);
-
-    src -= 3;
-
-    /* rearranging filter */
-    filt = LOAD_SH(filter);
-    filt0 = (v16i8) __msa_splati_h(filt, 0);
-    filt1 = (v16i8) __msa_splati_h(filt, 1);
-    filt2 = (v16i8) __msa_splati_h(filt, 2);
-    filt3 = (v16i8) __msa_splati_h(filt, 3);
-
-    mask1 = mask0 + 2;
-    mask2 = mask0 + 4;
-    mask3 = mask0 + 6;
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-
-    for (loop_cnt = height; loop_cnt--;) {
-        src0 = LOAD_SB(src);
-        src2 = LOAD_SB(src + 16);
-        src3 = LOAD_SB(src + 24);
-        src1 = __msa_sld_b(src2, src0, 8);
-
-        XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-        HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1,
-                                   mask2, mask3, filt0, filt1, filt2, filt3,
-                                   out0, out1, out2, out3);
-
-        out0 = SRAR_SATURATE_SIGNED_H(out0, rnd_vec, 7);
-        out1 = SRAR_SATURATE_SIGNED_H(out1, rnd_vec, 7);
-        out2 = SRAR_SATURATE_SIGNED_H(out2, rnd_vec, 7);
-        out3 = SRAR_SATURATE_SIGNED_H(out3, rnd_vec, 7);
-
-        PCKEV_B_XORI128_STORE_VEC(out1, out0, dst);
-        PCKEV_B_XORI128_STORE_VEC(out3, out2, dst + 16);
-
-        src0 = LOAD_SB(src + 32);
-        src2 = LOAD_SB(src + 48);
-        src3 = LOAD_SB(src + 56);
-        src1 = __msa_sld_b(src2, src0, 8);
-
-        XORI_B_4VECS_SB(src0, src1, src2, src3, src0, src1, src2, src3, 128);
-
-        HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, mask0, mask1,
-                                   mask2, mask3, filt0, filt1, filt2, filt3,
-                                   out0, out1, out2, out3);
-
-        out0 = SRAR_SATURATE_SIGNED_H(out0, rnd_vec, 7);
-        out1 = SRAR_SATURATE_SIGNED_H(out1, rnd_vec, 7);
-        out2 = SRAR_SATURATE_SIGNED_H(out2, rnd_vec, 7);
-        out3 = SRAR_SATURATE_SIGNED_H(out3, rnd_vec, 7);
-
-        PCKEV_B_XORI128_STORE_VEC(out1, out0, dst + 32);
-        PCKEV_B_XORI128_STORE_VEC(out3, out2, dst + 48);
-
-        src += src_stride;
-        dst += dst_stride;
-    }
-}
-
-static void common_vt_8t_4w_msa(uint8_t *src, int32_t src_stride,
-                                uint8_t *dst, int32_t dst_stride,
-                                const int8_t *filter, int32_t height,
-                                uint8_t rnd_val)
-{
-    uint32_t loop_cnt;
-    v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
-    v16i8 src10_r, src32_r, src54_r, src76_r, src98_r;
-    v16i8 src21_r, src43_r, src65_r, src87_r, src109_r;
-    v16i8 src2110, src4332, src6554, src8776, src10998;
-    v16i8 filt0, filt1, filt2, filt3;
-    v8i16 filt, out10, out32;
-    v8u16 rnd_vec;
-
-    src -= (3 * src_stride);
-
-    filt = LOAD_SH(filter);
-    filt0 = (v16i8) __msa_splati_h(filt, 0);
-    filt1 = (v16i8) __msa_splati_h(filt, 1);
-    filt2 = (v16i8) __msa_splati_h(filt, 2);
-    filt3 = (v16i8) __msa_splati_h(filt, 3);
-
-    LOAD_7VECS_SB(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
-    src += (7 * src_stride);
-
-    ILVR_B_6VECS_SB(src0, src2, src4, src1, src3, src5,
-                    src1, src3, src5, src2, src4, src6,
-                    src10_r, src32_r, src54_r, src21_r, src43_r, src65_r);
-
-    ILVR_D_3VECS_SB(src2110, src21_r, src10_r, src4332, src43_r, src32_r,
-                    src6554, src65_r, src54_r);
-
-    XORI_B_3VECS_SB(src2110, src4332, src6554, src2110, src4332, src6554, 128);
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-
-    for (loop_cnt = (height >> 2); loop_cnt--;) {
-        LOAD_4VECS_SB(src, src_stride, src7, src8, src9, src10);
-        src += (4 * src_stride);
-
-        ILVR_B_4VECS_SB(src6, src7, src8, src9, src7, src8, src9, src10,
-                        src76_r, src87_r, src98_r, src109_r);
-
-        ILVR_D_2VECS_SB(src8776, src87_r, src76_r, src10998, src109_r, src98_r);
-
-        XORI_B_2VECS_SB(src8776, src10998, src8776, src10998, 128);
-
-        out10 = FILT_8TAP_DPADD_S_H(src2110, src4332, src6554, src8776,
-                                    filt0, filt1, filt2, filt3);
-        out32 = FILT_8TAP_DPADD_S_H(src4332, src6554, src8776, src10998,
-                                    filt0, filt1, filt2, filt3);
-
-        out10 = SRAR_SATURATE_SIGNED_H(out10, rnd_vec, 7);
-        out32 = SRAR_SATURATE_SIGNED_H(out32, rnd_vec, 7);
-
-        PCKEV_2B_XORI128_STORE_4_BYTES_4(out10, out32, dst, dst_stride);
-        dst += (4 * dst_stride);
-
-        src2110 = src6554;
-        src4332 = src8776;
-        src6554 = src10998;
-
-        src6 = src10;
-    }
-}
-
-static void common_vt_8t_8w_msa(uint8_t *src, int32_t src_stride,
-                                uint8_t *dst, int32_t dst_stride,
-                                const int8_t *filter, int32_t height,
-                                uint8_t rnd_val)
-{
-    uint32_t loop_cnt;
-    v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
-    v16i8 src10_r, src32_r, src54_r, src76_r, src98_r;
-    v16i8 src21_r, src43_r, src65_r, src87_r, src109_r;
-    v16i8 filt0, filt1, filt2, filt3;
-    v8i16 filt, out0_r, out1_r, out2_r, out3_r;
-    v8u16 rnd_vec;
-
-    src -= (3 * src_stride);
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-
-    filt = LOAD_SH(filter);
-    filt0 = (v16i8) __msa_splati_h(filt, 0);
-    filt1 = (v16i8) __msa_splati_h(filt, 1);
-    filt2 = (v16i8) __msa_splati_h(filt, 2);
-    filt3 = (v16i8) __msa_splati_h(filt, 3);
-
-    LOAD_7VECS_SB(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
-    src += (7 * src_stride);
-
-    XORI_B_7VECS_SB(src0, src1, src2, src3, src4, src5, src6,
-                    src0, src1, src2, src3, src4, src5, src6, 128);
-
-    ILVR_B_6VECS_SB(src0, src2, src4, src1, src3, src5,
-                    src1, src3, src5, src2, src4, src6,
-                    src10_r, src32_r, src54_r, src21_r, src43_r, src65_r);
-
-    for (loop_cnt = (height >> 2); loop_cnt--;) {
-        LOAD_4VECS_SB(src, src_stride, src7, src8, src9, src10);
-        src += (4 * src_stride);
-
-        XORI_B_4VECS_SB(src7, src8, src9, src10, src7, src8, src9, src10, 128);
-
-        ILVR_B_4VECS_SB(src6, src7, src8, src9, src7, src8, src9, src10,
-                        src76_r, src87_r, src98_r, src109_r);
-
-        out0_r = FILT_8TAP_DPADD_S_H(src10_r, src32_r, src54_r, src76_r,
-                                     filt0, filt1, filt2, filt3);
-        out1_r = FILT_8TAP_DPADD_S_H(src21_r, src43_r, src65_r, src87_r,
-                                     filt0, filt1, filt2, filt3);
-        out2_r = FILT_8TAP_DPADD_S_H(src32_r, src54_r, src76_r, src98_r,
-                                     filt0, filt1, filt2, filt3);
-        out3_r = FILT_8TAP_DPADD_S_H(src43_r, src65_r, src87_r, src109_r,
-                                     filt0, filt1, filt2, filt3);
-
-        out0_r = SRAR_SATURATE_SIGNED_H(out0_r, rnd_vec, 7);
-        out1_r = SRAR_SATURATE_SIGNED_H(out1_r, rnd_vec, 7);
-        out2_r = SRAR_SATURATE_SIGNED_H(out2_r, rnd_vec, 7);
-        out3_r = SRAR_SATURATE_SIGNED_H(out3_r, rnd_vec, 7);
-
-        PCKEV_B_4_XORI128_STORE_8_BYTES_4(out0_r, out1_r, out2_r, out3_r,
-                                          dst, dst_stride);
-        dst += (4 * dst_stride);
-
-        src10_r = src54_r;
-        src32_r = src76_r;
-        src54_r = src98_r;
-        src21_r = src65_r;
-        src43_r = src87_r;
-        src65_r = src109_r;
-
-        src6 = src10;
-    }
-}
-
-static void common_vt_8t_12w_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
-                                 const int8_t *filter, int32_t height,
-                                 uint8_t rnd_val)
-{
-    int32_t loop_cnt;
-    uint32_t out2, out3;
-    uint64_t out0, out1;
-    v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
-    v16i8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
-    v16i8 vec8, vec9, vec10, vec11, vec12, vec13, vec14, vec15, vec16;
-    v16i8 res0, res1, res2;
-    v8i16 vec01, vec23, vec45, vec67;
-    v8i16 tmp0, tmp1, tmp2;
-    v8i16 filt, filt0, filt1, filt2, filt3;
-    v8u16 rnd_vec;
-    v4i32 mask = { 2, 6, 2, 6 };
-
-    src -= (3 * src_stride);
-
-    LOAD_7VECS_UB(src, src_stride, src0, src1, src2, src3, src4, src5, src6);
-    src += (7 * src_stride);
-
-    XORI_B_4VECS_SB(src0, src1, src2, src3, vec0, vec1, vec2, vec3, 128);
-    vec4 = (v16i8) __msa_xori_b((v16u8) src4, 128);
-    vec5 = (v16i8) __msa_xori_b((v16u8) src5, 128);
-    vec6 = (v16i8) __msa_xori_b((v16u8) src6, 128);
-
-    /* 4 width */
-    vec9 = (v16i8) __msa_vshf_w(mask, (v4i32) vec1, (v4i32) vec0);
-    vec10 = (v16i8) __msa_vshf_w(mask, (v4i32) vec2, (v4i32) vec1);
-    vec11 = (v16i8) __msa_vshf_w(mask, (v4i32) vec3, (v4i32) vec2);
-    vec12 = (v16i8) __msa_vshf_w(mask, (v4i32) vec4, (v4i32) vec3);
-    vec13 = (v16i8) __msa_vshf_w(mask, (v4i32) vec5, (v4i32) vec4);
-    vec14 = (v16i8) __msa_vshf_w(mask, (v4i32) vec6, (v4i32) vec5);
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-
-    /* rearranging filter_y */
-    filt = LOAD_SH(filter);
-    filt0 = (v8i16) __msa_splati_h(filt, 0);
-    filt1 = (v8i16) __msa_splati_h(filt, 1);
-    filt2 = (v8i16) __msa_splati_h(filt, 2);
-    filt3 = (v8i16) __msa_splati_h(filt, 3);
-
-    for (loop_cnt = (height >> 1); loop_cnt--;) {
-        LOAD_2VECS_UB(src, src_stride, src7, src8);
-        src += (2 * src_stride);
-
-        XORI_B_2VECS_SB(src7, src8, vec7, vec8, 128);
-
-        ILVR_B_4VECS_SH(vec0, vec2, vec4, vec6, vec1, vec3, vec5, vec7,
-                        vec01, vec23, vec45, vec67);
-
-        tmp0 = FILT_8TAP_DPADD_S_H(vec01, vec23, vec45, vec67, filt0, filt1,
-                                   filt2, filt3);
-
-        ILVR_B_4VECS_SH(vec1, vec3, vec5, vec7, vec2, vec4, vec6, vec8,
-                        vec01, vec23, vec45, vec67);
-
-        tmp1 = FILT_8TAP_DPADD_S_H(vec01, vec23, vec45, vec67, filt0, filt1,
-                                   filt2, filt3);
-
-        /* 4 width */
-        vec15 = (v16i8) __msa_vshf_w(mask, (v4i32) vec7, (v4i32) vec6);
-        vec16 = (v16i8) __msa_vshf_w(mask, (v4i32) vec8, (v4i32) vec7);
-
-        ILVR_B_4VECS_SH(vec9, vec11, vec13, vec15, vec10, vec12, vec14, vec16,
-                        vec01, vec23, vec45, vec67);
-
-        tmp2 = FILT_8TAP_DPADD_S_H(vec01, vec23, vec45, vec67, filt0, filt1,
-                                   filt2, filt3);
-
-        tmp0 = SRAR_SATURATE_SIGNED_H(tmp0, rnd_vec, 7);
-        tmp1 = SRAR_SATURATE_SIGNED_H(tmp1, rnd_vec, 7);
-        tmp2 = SRAR_SATURATE_SIGNED_H(tmp2, rnd_vec, 7);
-
-        res0 = __msa_pckev_b((v16i8) tmp0, (v16i8) tmp0);
-        res1 = __msa_pckev_b((v16i8) tmp1, (v16i8) tmp1);
-        res2 = __msa_pckev_b((v16i8) tmp2, (v16i8) tmp2);
-
-        XORI_B_3VECS_SB(res0, res1, res2, res0, res1, res2, 128);
-
-        out0 = __msa_copy_u_d((v2i64) res0, 0);
-        out1 = __msa_copy_u_d((v2i64) res1, 0);
-        out2 = __msa_copy_u_w((v4i32) res2, 0);
-        out3 = __msa_copy_u_w((v4i32) res2, 1);
-
-        STORE_DWORD(dst, out0);
-        STORE_WORD((dst + 8), out2);
-        dst += dst_stride;
-        STORE_DWORD(dst, out1);
-        STORE_WORD((dst + 8), out3);
-        dst += dst_stride;
-
-        vec0 = vec2;
-        vec1 = vec3;
-        vec2 = vec4;
-        vec3 = vec5;
-        vec4 = vec6;
-        vec5 = vec7;
-        vec6 = vec8;
-        vec9 = vec11;
-        vec10 = vec12;
-        vec11 = vec13;
-        vec12 = vec14;
-        vec13 = vec15;
-        vec14 = vec16;
-    }
-}
-
-static void common_vt_8t_16w_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
-                                 const int8_t *filter, int32_t height,
-                                 uint8_t rnd_val)
-{
-    uint32_t loop_cnt;
-    v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
-    v16i8 filt0, filt1, filt2, filt3;
-    v16i8 src10_r, src32_r, src54_r, src76_r, src98_r;
-    v16i8 src21_r, src43_r, src65_r, src87_r, src109_r;
-    v16i8 src10_l, src32_l, src54_l, src76_l, src98_l;
-    v16i8 src21_l, src43_l, src65_l, src87_l, src109_l;
-    v8i16 out0_r, out1_r, out2_r, out3_r, out0_l, out1_l, out2_l, out3_l;
-    v8i16 filt;
-    v8u16 rnd_vec;
-    v16u8 tmp0, tmp1, tmp2, tmp3;
-
-    src -= (3 * src_stride);
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-
-    filt = LOAD_SH(filter);
-    filt0 = (v16i8) __msa_splati_h(filt, 0);
-    filt1 = (v16i8) __msa_splati_h(filt, 1);
-    filt2 = (v16i8) __msa_splati_h(filt, 2);
-    filt3 = (v16i8) __msa_splati_h(filt, 3);
-
-    LOAD_7VECS_SB(src, src_stride,
-                  src0, src1, src2, src3, src4, src5, src6);
-    src += (7 * src_stride);
-
-    XORI_B_7VECS_SB(src0, src1, src2, src3, src4, src5, src6,
-                    src0, src1, src2, src3, src4, src5, src6, 128);
-
-    ILVR_B_6VECS_SB(src0, src2, src4, src1, src3, src5,
-                    src1, src3, src5, src2, src4, src6,
-                    src10_r, src32_r, src54_r, src21_r, src43_r, src65_r);
-
-    ILVL_B_6VECS_SB(src0, src2, src4, src1, src3, src5,
-                    src1, src3, src5, src2, src4, src6,
-                    src10_l, src32_l, src54_l, src21_l, src43_l, src65_l);
-
-    for (loop_cnt = (height >> 2); loop_cnt--;) {
-        LOAD_4VECS_SB(src, src_stride, src7, src8, src9, src10);
-        src += (4 * src_stride);
-
-        XORI_B_4VECS_SB(src7, src8, src9, src10,
-                        src7, src8, src9, src10, 128);
-
-        ILVR_B_4VECS_SB(src6, src7, src8, src9, src7, src8, src9, src10,
-                        src76_r, src87_r, src98_r, src109_r);
-
-        ILVL_B_4VECS_SB(src6, src7, src8, src9, src7, src8, src9, src10,
-                        src76_l, src87_l, src98_l, src109_l);
-
-        out0_r = FILT_8TAP_DPADD_S_H(src10_r, src32_r, src54_r, src76_r,
-                                     filt0, filt1, filt2, filt3);
-        out1_r = FILT_8TAP_DPADD_S_H(src21_r, src43_r, src65_r, src87_r,
-                                     filt0, filt1, filt2, filt3);
-        out2_r = FILT_8TAP_DPADD_S_H(src32_r, src54_r, src76_r, src98_r,
-                                     filt0, filt1, filt2, filt3);
-        out3_r = FILT_8TAP_DPADD_S_H(src43_r, src65_r, src87_r, src109_r,
-                                     filt0, filt1, filt2, filt3);
-
-        out0_l = FILT_8TAP_DPADD_S_H(src10_l, src32_l, src54_l, src76_l,
-                                     filt0, filt1, filt2, filt3);
-        out1_l = FILT_8TAP_DPADD_S_H(src21_l, src43_l, src65_l, src87_l,
-                                     filt0, filt1, filt2, filt3);
-        out2_l = FILT_8TAP_DPADD_S_H(src32_l, src54_l, src76_l, src98_l,
-                                     filt0, filt1, filt2, filt3);
-        out3_l = FILT_8TAP_DPADD_S_H(src43_l, src65_l, src87_l, src109_l,
-                                     filt0, filt1, filt2, filt3);
-
-        out0_r = SRAR_SATURATE_SIGNED_H(out0_r, rnd_vec, 7);
-        out1_r = SRAR_SATURATE_SIGNED_H(out1_r, rnd_vec, 7);
-        out2_r = SRAR_SATURATE_SIGNED_H(out2_r, rnd_vec, 7);
-        out3_r = SRAR_SATURATE_SIGNED_H(out3_r, rnd_vec, 7);
-        out0_l = SRAR_SATURATE_SIGNED_H(out0_l, rnd_vec, 7);
-        out1_l = SRAR_SATURATE_SIGNED_H(out1_l, rnd_vec, 7);
-        out2_l = SRAR_SATURATE_SIGNED_H(out2_l, rnd_vec, 7);
-        out3_l = SRAR_SATURATE_SIGNED_H(out3_l, rnd_vec, 7);
-
-        PCKEV_B_4VECS_UB(out0_l, out1_l, out2_l, out3_l, out0_r, out1_r, out2_r,
-                         out3_r, tmp0, tmp1, tmp2, tmp3);
-
-        XORI_B_4VECS_UB(tmp0, tmp1, tmp2, tmp3, tmp0, tmp1, tmp2, tmp3, 128);
-
-        STORE_4VECS_UB(dst, dst_stride, tmp0, tmp1, tmp2, tmp3);
-        dst += (4 * dst_stride);
-
-        src10_r = src54_r;
-        src32_r = src76_r;
-        src54_r = src98_r;
-        src21_r = src65_r;
-        src43_r = src87_r;
-        src65_r = src109_r;
-        src10_l = src54_l;
-        src32_l = src76_l;
-        src54_l = src98_l;
-        src21_l = src65_l;
-        src43_l = src87_l;
-        src65_l = src109_l;
-
-        src6 = src10;
-    }
-}
-
-static void common_vt_8t_16w_mult_msa(uint8_t *src, int32_t src_stride,
-                                      uint8_t *dst, int32_t dst_stride,
-                                      const int8_t *filter, int32_t height,
-                                      uint8_t rnd_val, int32_t width)
-{
-    uint8_t *src_tmp;
-    uint8_t *dst_tmp;
-    uint32_t loop_cnt, cnt;
-    v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
-    v16i8 filt0, filt1, filt2, filt3;
-    v16i8 src10_r, src32_r, src54_r, src76_r, src98_r;
-    v16i8 src21_r, src43_r, src65_r, src87_r, src109_r;
-    v16i8 src10_l, src32_l, src54_l, src76_l, src98_l;
-    v16i8 src21_l, src43_l, src65_l, src87_l, src109_l;
-    v8i16 filt, out0_r, out1_r, out2_r, out3_r, out0_l, out1_l, out2_l, out3_l;
-    v16u8 tmp0, tmp1, tmp2, tmp3;
-    v8u16 rnd_vec;
-
-    src -= (3 * src_stride);
-
-    rnd_vec = (v8u16) __msa_fill_h(rnd_val);
-
-    filt = LOAD_SH(filter);
-    filt0 = (v16i8) __msa_splati_h(filt, 0);
-    filt1 = (v16i8) __msa_splati_h(filt, 1);
-    filt2 = (v16i8) __msa_splati_h(filt, 2);
-    filt3 = (v16i8) __msa_splati_h(filt, 3);
-
-    for (cnt = (width >> 4); cnt--;) {
-        src_tmp = src;
-        dst_tmp = dst;
-
-        LOAD_7VECS_SB(src_tmp, src_stride,
-                      src0, src1, src2, src3, src4, src5, src6);
-        src_tmp += (7 * src_stride);
-
-        XORI_B_7VECS_SB(src0, src1, src2, src3, src4, src5, src6,
-                        src0, src1, src2, src3, src4, src5, src6, 128);
-
-        ILVR_B_6VECS_SB(src0, src2, src4, src1, src3, src5,
-                        src1, src3, src5, src2, src4, src6,
-                        src10_r, src32_r, src54_r, src21_r, src43_r, src65_r);
-
-        ILVL_B_6VECS_SB(src0, src2, src4, src1, src3, src5,
-                        src1, src3, src5, src2, src4, src6,
-                        src10_l, src32_l, src54_l, src21_l, src43_l, src65_l);
-
-        for (loop_cnt = (height >> 2); loop_cnt--;) {
-            LOAD_4VECS_SB(src_tmp, src_stride, src7, src8, src9, src10);
-            src_tmp += (4 * src_stride);
-
-            XORI_B_4VECS_SB(src7, src8, src9, src10,
-                            src7, src8, src9, src10, 128);
-
-            ILVR_B_4VECS_SB(src6, src7, src8, src9, src7, src8, src9, src10,
-                            src76_r, src87_r, src98_r, src109_r);
-
-            ILVL_B_4VECS_SB(src6, src7, src8, src9, src7, src8, src9, src10,
-                            src76_l, src87_l, src98_l, src109_l);
-
-            out0_r = FILT_8TAP_DPADD_S_H(src10_r, src32_r, src54_r, src76_r,
-                                         filt0, filt1, filt2, filt3);
-            out1_r = FILT_8TAP_DPADD_S_H(src21_r, src43_r, src65_r, src87_r,
-                                         filt0, filt1, filt2, filt3);
-            out2_r = FILT_8TAP_DPADD_S_H(src32_r, src54_r, src76_r, src98_r,
-                                         filt0, filt1, filt2, filt3);
-            out3_r = FILT_8TAP_DPADD_S_H(src43_r, src65_r, src87_r, src109_r,
-                                         filt0, filt1, filt2, filt3);
-
-            out0_l = FILT_8TAP_DPADD_S_H(src10_l, src32_l, src54_l, src76_l,
-                                         filt0, filt1, filt2, filt3);
-            out1_l = FILT_8TAP_DPADD_S_H(src21_l, src43_l, src65_l, src87_l,
-                                         filt0, filt1, filt2, filt3);
-            out2_l = FILT_8TAP_DPADD_S_H(src32_l, src54_l, src76_l, src98_l,
-                                         filt0, filt1, filt2, filt3);
-            out3_l = FILT_8TAP_DPADD_S_H(src43_l, src65_l, src87_l, src109_l,
-                                         filt0, filt1, filt2, filt3);
-
-            out0_r = SRAR_SATURATE_SIGNED_H(out0_r, rnd_vec, 7);
-            out1_r = SRAR_SATURATE_SIGNED_H(out1_r, rnd_vec, 7);
-            out2_r = SRAR_SATURATE_SIGNED_H(out2_r, rnd_vec, 7);
-            out3_r = SRAR_SATURATE_SIGNED_H(out3_r, rnd_vec, 7);
-            out0_l = SRAR_SATURATE_SIGNED_H(out0_l, rnd_vec, 7);
-            out1_l = SRAR_SATURATE_SIGNED_H(out1_l, rnd_vec, 7);
-            out2_l = SRAR_SATURATE_SIGNED_H(out2_l, rnd_vec, 7);
-            out3_l = SRAR_SATURATE_SIGNED_H(out3_l, rnd_vec, 7);
-
-            PCKEV_B_4VECS_UB(out0_l, out1_l, out2_l, out3_l, out0_r, out1_r,
-                             out2_r, out3_r, tmp0, tmp1, tmp2, tmp3);
-
-            XORI_B_4VECS_UB(tmp0, tmp1, tmp2, tmp3,
-                            tmp0, tmp1, tmp2, tmp3, 128);
-
-            STORE_4VECS_UB(dst_tmp, dst_stride, tmp0, tmp1, tmp2, tmp3);
-            dst_tmp += (4 * dst_stride);
-
-            src10_r = src54_r;
-            src32_r = src76_r;
-            src54_r = src98_r;
-            src21_r = src65_r;
-            src43_r = src87_r;
-            src65_r = src109_r;
-            src10_l = src54_l;
-            src32_l = src76_l;
-            src54_l = src98_l;
-            src21_l = src65_l;
-            src43_l = src87_l;
-            src65_l = src109_l;
-
-            src6 = src10;
-        }
-
-        src += 16;
-        dst += 16;
-    }
-}
-
-static void common_vt_8t_24w_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
-                                 const int8_t *filter, int32_t height,
-                                 uint8_t rnd_val)
-{
-    common_vt_8t_16w_mult_msa(src, src_stride, dst, dst_stride, filter,
-                              height, rnd_val, 16);
-
-    common_vt_8t_8w_msa(src + 16, src_stride, dst + 16, dst_stride,
-                        filter, height, rnd_val);
-}
-
-static void common_vt_8t_32w_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
-                                 const int8_t *filter, int32_t height,
-                                 uint8_t rnd_val)
-{
-    common_vt_8t_16w_mult_msa(src, src_stride, dst, dst_stride,
-                              filter, height, rnd_val, 32);
-}
-
-static void common_vt_8t_48w_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
-                                 const int8_t *filter, int32_t height,
-                                 uint8_t rnd_val)
-{
-    common_vt_8t_16w_mult_msa(src, src_stride, dst, dst_stride,
-                              filter, height, rnd_val, 48);
-}
-
-static void common_vt_8t_64w_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
-                                 const int8_t *filter, int32_t height,
-                                 uint8_t rnd_val)
-{
-    common_vt_8t_16w_mult_msa(src, src_stride, dst, dst_stride,
-                              filter, height, rnd_val, 64);
-}
-
-static void copy_width8_msa(uint8_t *src, int32_t src_stride,
-                            uint8_t *dst, int32_t dst_stride,
-                            int32_t height)
-{
-    int32_t cnt;
-    uint64_t out0, out1, out2, out3, out4, out5, out6, out7;
-    v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
-
-    if (0 == height % 12) {
-        for (cnt = (height / 12); cnt--;) {
-            LOAD_8VECS_UB(src, src_stride,
-                          src0, src1, src2, src3, src4, src5, src6, src7);
-            src += (8 * src_stride);
-
-            out0 = __msa_copy_u_d((v2i64) src0, 0);
-            out1 = __msa_copy_u_d((v2i64) src1, 0);
-            out2 = __msa_copy_u_d((v2i64) src2, 0);
-            out3 = __msa_copy_u_d((v2i64) src3, 0);
-            out4 = __msa_copy_u_d((v2i64) src4, 0);
-            out5 = __msa_copy_u_d((v2i64) src5, 0);
-            out6 = __msa_copy_u_d((v2i64) src6, 0);
-            out7 = __msa_copy_u_d((v2i64) src7, 0);
-
-            STORE_DWORD(dst, out0);
-            dst += dst_stride;
-            STORE_DWORD(dst, out1);
-            dst += dst_stride;
-            STORE_DWORD(dst, out2);
-            dst += dst_stride;
-            STORE_DWORD(dst, out3);
-            dst += dst_stride;
-            STORE_DWORD(dst, out4);
-            dst += dst_stride;
-            STORE_DWORD(dst, out5);
-            dst += dst_stride;
-            STORE_DWORD(dst, out6);
-            dst += dst_stride;
-            STORE_DWORD(dst, out7);
-            dst += dst_stride;
-
-            LOAD_4VECS_UB(src, src_stride, src0, src1, src2, src3);
-            src += (4 * src_stride);
-
-            out0 = __msa_copy_u_d((v2i64) src0, 0);
-            out1 = __msa_copy_u_d((v2i64) src1, 0);
-            out2 = __msa_copy_u_d((v2i64) src2, 0);
-            out3 = __msa_copy_u_d((v2i64) src3, 0);
-
-            STORE_DWORD(dst, out0);
-            dst += dst_stride;
-            STORE_DWORD(dst, out1);
-            dst += dst_stride;
-            STORE_DWORD(dst, out2);
-            dst += dst_stride;
-            STORE_DWORD(dst, out3);
-            dst += dst_stride;
-        }
-    } else if (0 == height % 8) {
-        for (cnt = height >> 3; cnt--;) {
-            LOAD_8VECS_UB(src, src_stride,
-                          src0, src1, src2, src3, src4, src5, src6, src7);
-            src += (8 * src_stride);
-
-            out0 = __msa_copy_u_d((v2i64) src0, 0);
-            out1 = __msa_copy_u_d((v2i64) src1, 0);
-            out2 = __msa_copy_u_d((v2i64) src2, 0);
-            out3 = __msa_copy_u_d((v2i64) src3, 0);
-            out4 = __msa_copy_u_d((v2i64) src4, 0);
-            out5 = __msa_copy_u_d((v2i64) src5, 0);
-            out6 = __msa_copy_u_d((v2i64) src6, 0);
-            out7 = __msa_copy_u_d((v2i64) src7, 0);
-
-            STORE_DWORD(dst, out0);
-            dst += dst_stride;
-            STORE_DWORD(dst, out1);
-            dst += dst_stride;
-            STORE_DWORD(dst, out2);
-            dst += dst_stride;
-            STORE_DWORD(dst, out3);
-            dst += dst_stride;
-            STORE_DWORD(dst, out4);
-            dst += dst_stride;
-            STORE_DWORD(dst, out5);
-            dst += dst_stride;
-            STORE_DWORD(dst, out6);
-            dst += dst_stride;
-            STORE_DWORD(dst, out7);
-            dst += dst_stride;
-        }
-    } else if (0 == height % 4) {
-        for (cnt = (height / 4); cnt--;) {
-            LOAD_4VECS_UB(src, src_stride, src0, src1, src2, src3);
-            src += (4 * src_stride);
-
-            out0 = __msa_copy_u_d((v2i64) src0, 0);
-            out1 = __msa_copy_u_d((v2i64) src1, 0);
-            out2 = __msa_copy_u_d((v2i64) src2, 0);
-            out3 = __msa_copy_u_d((v2i64) src3, 0);
-
-            STORE_DWORD(dst, out0);
-            dst += dst_stride;
-            STORE_DWORD(dst, out1);
-            dst += dst_stride;
-            STORE_DWORD(dst, out2);
-            dst += dst_stride;
-            STORE_DWORD(dst, out3);
-            dst += dst_stride;
-        }
-    } else if (0 == height % 2) {
-        for (cnt = (height / 2); cnt--;) {
-            LOAD_2VECS_UB(src, src_stride, src0, src1);
-            src += (2 * src_stride);
-
-            out0 = __msa_copy_u_d((v2i64) src0, 0);
-            out1 = __msa_copy_u_d((v2i64) src1, 0);
-
-            STORE_DWORD(dst, out0);
-            dst += dst_stride;
-            STORE_DWORD(dst, out1);
-            dst += dst_stride;
-        }
-    }
-}
-
-static void copy_width12_msa(uint8_t *src, int32_t src_stride,
-                             uint8_t *dst, int32_t dst_stride,
-                             int32_t height)
-{
-    int32_t cnt;
-    uint64_t out0, out1, out2, out3, out4, out5, out6, out7;
-    uint32_t out8, out9, out10, out11, out12, out13, out14, out15;
-    v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
-
-    for (cnt = 2; cnt--;) {
-        LOAD_8VECS_UB(src, src_stride,
-                      src0, src1, src2, src3, src4, src5, src6, src7);
-        src += (8 * src_stride);
-
-        out0 = __msa_copy_u_d((v2i64) src0, 0);
-        out1 = __msa_copy_u_d((v2i64) src1, 0);
-        out2 = __msa_copy_u_d((v2i64) src2, 0);
-        out3 = __msa_copy_u_d((v2i64) src3, 0);
-        out4 = __msa_copy_u_d((v2i64) src4, 0);
-        out5 = __msa_copy_u_d((v2i64) src5, 0);
-        out6 = __msa_copy_u_d((v2i64) src6, 0);
-        out7 = __msa_copy_u_d((v2i64) src7, 0);
-
-        out8 = __msa_copy_u_w((v4i32) src0, 2);
-        out9 = __msa_copy_u_w((v4i32) src1, 2);
-        out10 = __msa_copy_u_w((v4i32) src2, 2);
-        out11 = __msa_copy_u_w((v4i32) src3, 2);
-        out12 = __msa_copy_u_w((v4i32) src4, 2);
-        out13 = __msa_copy_u_w((v4i32) src5, 2);
-        out14 = __msa_copy_u_w((v4i32) src6, 2);
-        out15 = __msa_copy_u_w((v4i32) src7, 2);
-
-        STORE_DWORD(dst, out0);
-        STORE_WORD(dst + 8, out8);
-        dst += dst_stride;
-        STORE_DWORD(dst, out1);
-        STORE_WORD(dst + 8, out9);
-        dst += dst_stride;
-        STORE_DWORD(dst, out2);
-        STORE_WORD(dst + 8, out10);
-        dst += dst_stride;
-        STORE_DWORD(dst, out3);
-        STORE_WORD(dst + 8, out11);
-        dst += dst_stride;
-        STORE_DWORD(dst, out4);
-        STORE_WORD(dst + 8, out12);
-        dst += dst_stride;
-        STORE_DWORD(dst, out5);
-        STORE_WORD(dst + 8, out13);
-        dst += dst_stride;
-        STORE_DWORD(dst, out6);
-        STORE_WORD(dst + 8, out14);
-        dst += dst_stride;
-        STORE_DWORD(dst, out7);
-        STORE_WORD(dst + 8, out15);
-        dst += dst_stride;
-    }
-}
-
-static void copy_16multx8mult_msa(uint8_t *src, int32_t src_stride,
-                                  uint8_t *dst, int32_t dst_stride,
-                                  int32_t height, int32_t width)
-{
-    int32_t cnt, loop_cnt;
-    uint8_t *src_tmp, *dst_tmp;
-    v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
-
-    for (cnt = (width >> 4); cnt--;) {
-        src_tmp = src;
-        dst_tmp = dst;
-
-        for (loop_cnt = (height >> 3); loop_cnt--;) {
-            LOAD_8VECS_UB(src_tmp, src_stride,
-                          src0, src1, src2, src3, src4, src5, src6, src7);
-            src_tmp += (8 * src_stride);
-
-            STORE_8VECS_UB(dst_tmp, dst_stride,
-                           src0, src1, src2, src3, src4, src5, src6, src7);
-            dst_tmp += (8 * dst_stride);
-        }
-
-        src += 16;
-        dst += 16;
-    }
-}
-
-static void copy_width16_msa(uint8_t *src, int32_t src_stride,
-                             uint8_t *dst, int32_t dst_stride,
-                             int32_t height)
-{
-    int32_t cnt;
-    v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
-
-    if (0 == height % 12) {
-        for (cnt = (height / 12); cnt--;) {
-            LOAD_8VECS_UB(src, src_stride,
-                          src0, src1, src2, src3, src4, src5, src6, src7);
-            src += (8 * src_stride);
-
-            STORE_8VECS_UB(dst, dst_stride,
-                           src0, src1, src2, src3, src4, src5, src6, src7);
-            dst += (8 * dst_stride);
-
-            LOAD_4VECS_UB(src, src_stride, src0, src1, src2, src3);
-            src += (4 * src_stride);
-
-            STORE_4VECS_UB(dst, dst_stride, src0, src1, src2, src3);
-            dst += (4 * dst_stride);
-        }
-    } else if (0 == height % 8) {
-        copy_16multx8mult_msa(src, src_stride, dst, dst_stride, height, 16);
-    } else if (0 == height % 4) {
-        for (cnt = (height >> 2); cnt--;) {
-            LOAD_4VECS_UB(src, src_stride, src0, src1, src2, src3);
-            src += (4 * src_stride);
-
-            STORE_4VECS_UB(dst, dst_stride, src0, src1, src2, src3);
-            dst += (4 * dst_stride);
-        }
-    }
-}
-
-static void copy_width24_msa(uint8_t *src, int32_t src_stride,
-                             uint8_t *dst, int32_t dst_stride,
-                             int32_t height)
-{
-    copy_16multx8mult_msa(src, src_stride, dst, dst_stride, height, 16);
-
-    copy_width8_msa(src + 16, src_stride, dst + 16, dst_stride, height);
-}
-
-static void copy_width32_msa(uint8_t *src, int32_t src_stride,
-                             uint8_t *dst, int32_t dst_stride,
-                             int32_t height)
-{
-    int32_t cnt;
-    v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
-
-    if (0 == height % 12) {
-        for (cnt = (height / 12); cnt--;) {
-            LOAD_4VECS_UB(src, src_stride, src0, src1, src2, src3);
-            LOAD_4VECS_UB(src + 16, src_stride, src4, src5, src6, src7);
-            src += (4 * src_stride);
-
-            STORE_4VECS_UB(dst, dst_stride, src0, src1, src2, src3);
-            STORE_4VECS_UB(dst + 16, dst_stride, src4, src5, src6, src7);
-            dst += (4 * dst_stride);
-
-            LOAD_4VECS_UB(src, src_stride, src0, src1, src2, src3);
-            LOAD_4VECS_UB(src + 16, src_stride, src4, src5, src6, src7);
-            src += (4 * src_stride);
-
-            STORE_4VECS_UB(dst, dst_stride, src0, src1, src2, src3);
-            STORE_4VECS_UB(dst + 16, dst_stride, src4, src5, src6, src7);
-            dst += (4 * dst_stride);
-
-            LOAD_4VECS_UB(src, src_stride, src0, src1, src2, src3);
-            LOAD_4VECS_UB(src + 16, src_stride, src4, src5, src6, src7);
-            src += (4 * src_stride);
-
-            STORE_4VECS_UB(dst, dst_stride, src0, src1, src2, src3);
-            STORE_4VECS_UB(dst + 16, dst_stride, src4, src5, src6, src7);
-            dst += (4 * dst_stride);
-        }
-    } else if (0 == height % 8) {
-        copy_16multx8mult_msa(src, src_stride, dst, dst_stride, height, 32);
-    } else if (0 == height % 4) {
-        for (cnt = (height >> 2); cnt--;) {
-            LOAD_4VECS_UB(src, src_stride, src0, src1, src2, src3);
-            LOAD_4VECS_UB(src + 16, src_stride, src4, src5, src6, src7);
-            src += (4 * src_stride);
-
-            STORE_4VECS_UB(dst, dst_stride, src0, src1, src2, src3);
-            STORE_4VECS_UB(dst + 16, dst_stride, src4, src5, src6, src7);
-            dst += (4 * dst_stride);
-        }
-    }
-}
-
-static void copy_width48_msa(uint8_t *src, int32_t src_stride,
-                             uint8_t *dst, int32_t dst_stride,
-                             int32_t height)
-{
-    copy_16multx8mult_msa(src, src_stride, dst, dst_stride, height, 48);
-}
-
-static void copy_width64_msa(uint8_t *src, int32_t src_stride,
-                             uint8_t *dst, int32_t dst_stride,
-                             int32_t height)
-{
-    copy_16multx8mult_msa(src, src_stride, dst, dst_stride, height, 64);
-}
-
-#define MC_COPY(WIDTH)                                                    \
-void ff_hevc_put_hevc_pel_pixels##WIDTH##_8_msa(int16_t *dst,             \
-                                                uint8_t *src,             \
-                                                ptrdiff_t src_stride,     \
-                                                int height,               \
-                                                intptr_t mx,              \
-                                                intptr_t my,              \
-                                                int width)                \
-{                                                                         \
-    hevc_copy_##WIDTH##w_msa(src, src_stride, dst, MAX_PB_SIZE, height);  \
-}
-
-MC_COPY(4);
-MC_COPY(6);
-MC_COPY(8);
-MC_COPY(12);
-MC_COPY(16);
-MC_COPY(24);
-MC_COPY(32);
-MC_COPY(48);
-MC_COPY(64);
-
-#undef MC_COPY
-
-#define MC(PEL, DIR, WIDTH, TAP, DIR1, FILT_DIR)                            \
-void ff_hevc_put_hevc_##PEL##_##DIR####WIDTH##_8_msa(int16_t *dst,          \
-                                                     uint8_t *src,          \
-                                                     ptrdiff_t src_stride,  \
-                                                     int height,            \
-                                                     intptr_t mx,           \
-                                                     intptr_t my,           \
-                                                     int width)             \
-{                                                                           \
-    const int8_t *filter = ff_hevc_##PEL##_filters[FILT_DIR - 1];           \
-                                                                            \
-    hevc_##DIR1##_##TAP##t_##WIDTH##w_msa(src, src_stride, dst,             \
-                                          MAX_PB_SIZE, filter, height);     \
-}
-
-MC(qpel, h, 4, 8, hz, mx);
-MC(qpel, h, 8, 8, hz, mx);
-MC(qpel, h, 12, 8, hz, mx);
-MC(qpel, h, 16, 8, hz, mx);
-MC(qpel, h, 24, 8, hz, mx);
-MC(qpel, h, 32, 8, hz, mx);
-MC(qpel, h, 48, 8, hz, mx);
-MC(qpel, h, 64, 8, hz, mx);
-
-MC(qpel, v, 4, 8, vt, my);
-MC(qpel, v, 8, 8, vt, my);
-MC(qpel, v, 12, 8, vt, my);
-MC(qpel, v, 16, 8, vt, my);
-MC(qpel, v, 24, 8, vt, my);
-MC(qpel, v, 32, 8, vt, my);
-MC(qpel, v, 48, 8, vt, my);
-MC(qpel, v, 64, 8, vt, my);
-
-#undef MC
-
-#define MC_HV(PEL, DIR, WIDTH, TAP, DIR1)                                     \
-void ff_hevc_put_hevc_##PEL##_##DIR####WIDTH##_8_msa(int16_t *dst,            \
-                                                     uint8_t *src,            \
-                                                     ptrdiff_t src_stride,    \
-                                                     int height,              \
-                                                     intptr_t mx,             \
-                                                     intptr_t my,             \
-                                                     int width)               \
-{                                                                             \
-    const int8_t *filter_x = ff_hevc_##PEL##_filters[mx - 1];                 \
-    const int8_t *filter_y = ff_hevc_##PEL##_filters[my - 1];                 \
-                                                                              \
-    hevc_##DIR1##_##TAP##t_##WIDTH##w_msa(src, src_stride, dst, MAX_PB_SIZE,  \
-                                          filter_x, filter_y, height);        \
-}
-
-MC_HV(qpel, hv, 4, 8, hv);
-MC_HV(qpel, hv, 8, 8, hv);
-MC_HV(qpel, hv, 12, 8, hv);
-MC_HV(qpel, hv, 16, 8, hv);
-MC_HV(qpel, hv, 24, 8, hv);
-MC_HV(qpel, hv, 32, 8, hv);
-MC_HV(qpel, hv, 48, 8, hv);
-MC_HV(qpel, hv, 64, 8, hv);
+MC_HV(qpel, hv, 4, 8, hv);
+MC_HV(qpel, hv, 8, 8, hv);
+MC_HV(qpel, hv, 12, 8, hv);
+MC_HV(qpel, hv, 16, 8, hv);
+MC_HV(qpel, hv, 24, 8, hv);
+MC_HV(qpel, hv, 32, 8, hv);
+MC_HV(qpel, hv, 48, 8, hv);
+MC_HV(qpel, hv, 64, 8, hv);
 
 #undef MC_HV
-
-#define UNI_MC_COPY(WIDTH)                                                 \
-void ff_hevc_put_hevc_uni_pel_pixels##WIDTH##_8_msa(uint8_t *dst,          \
-                                                    ptrdiff_t dst_stride,  \
-                                                    uint8_t *src,          \
-                                                    ptrdiff_t src_stride,  \
-                                                    int height,            \
-                                                    intptr_t mx,           \
-                                                    intptr_t my,           \
-                                                    int width)             \
-{                                                                          \
-    copy_width##WIDTH##_msa(src, src_stride, dst, dst_stride, height);     \
-}
-
-UNI_MC_COPY(8);
-UNI_MC_COPY(12);
-UNI_MC_COPY(16);
-UNI_MC_COPY(24);
-UNI_MC_COPY(32);
-UNI_MC_COPY(48);
-UNI_MC_COPY(64);
-
-#undef UNI_MC_COPY
-
-#define UNI_MC(PEL, DIR, WIDTH, TAP, DIR1, FILT_DIR)                           \
-void ff_hevc_put_hevc_uni_##PEL##_##DIR####WIDTH##_8_msa(uint8_t *dst,         \
-                                                         ptrdiff_t             \
-                                                         dst_stride,           \
-                                                         uint8_t *src,         \
-                                                         ptrdiff_t             \
-                                                         src_stride,           \
-                                                         int height,           \
-                                                         intptr_t mx,          \
-                                                         intptr_t my,          \
-                                                         int width)            \
-{                                                                              \
-    const int8_t *filter = ff_hevc_##PEL##_filters[FILT_DIR - 1];              \
-                                                                               \
-    common_##DIR1##_##TAP##t_##WIDTH##w_msa(src, src_stride, dst, dst_stride,  \
-                                            filter, height, 6);                \
-}
-
-UNI_MC(qpel, h, 4, 8, hz, mx);
-UNI_MC(qpel, h, 8, 8, hz, mx);
-UNI_MC(qpel, h, 12, 8, hz, mx);
-UNI_MC(qpel, h, 16, 8, hz, mx);
-UNI_MC(qpel, h, 24, 8, hz, mx);
-UNI_MC(qpel, h, 32, 8, hz, mx);
-UNI_MC(qpel, h, 48, 8, hz, mx);
-UNI_MC(qpel, h, 64, 8, hz, mx);
-
-UNI_MC(qpel, v, 4, 8, vt, my);
-UNI_MC(qpel, v, 8, 8, vt, my);
-UNI_MC(qpel, v, 12, 8, vt, my);
-UNI_MC(qpel, v, 16, 8, vt, my);
-UNI_MC(qpel, v, 24, 8, vt, my);
-UNI_MC(qpel, v, 32, 8, vt, my);
-UNI_MC(qpel, v, 48, 8, vt, my);
-UNI_MC(qpel, v, 64, 8, vt, my);
-
-#undef UNI_MC
-
-#define UNI_MC_HV(PEL, DIR, WIDTH, TAP, DIR1)                           \
-void ff_hevc_put_hevc_uni_##PEL##_##DIR####WIDTH##_8_msa(uint8_t *dst,  \
-                                                         ptrdiff_t      \
-                                                         dst_stride,    \
-                                                         uint8_t *src,  \
-                                                         ptrdiff_t      \
-                                                         src_stride,    \
-                                                         int height,    \
-                                                         intptr_t mx,   \
-                                                         intptr_t my,   \
-                                                         int width)     \
-{                                                                       \
-    const int8_t *filter_x = ff_hevc_##PEL##_filters[mx - 1];           \
-    const int8_t *filter_y = ff_hevc_##PEL##_filters[my - 1];           \
-                                                                        \
-    hevc_##DIR1##_uni_##TAP##t_##WIDTH##w_msa(src, src_stride, dst,     \
-                                              dst_stride, filter_x,     \
-                                              filter_y, height);        \
-}
-
-UNI_MC_HV(qpel, hv, 4, 8, hv);
-UNI_MC_HV(qpel, hv, 8, 8, hv);
-UNI_MC_HV(qpel, hv, 12, 8, hv);
-UNI_MC_HV(qpel, hv, 16, 8, hv);
-UNI_MC_HV(qpel, hv, 24, 8, hv);
-UNI_MC_HV(qpel, hv, 32, 8, hv);
-UNI_MC_HV(qpel, hv, 48, 8, hv);
-UNI_MC_HV(qpel, hv, 64, 8, hv);
-
-#undef UNI_MC_HV
diff --git a/libavutil/mips/generic_macros_msa.h b/libavutil/mips/generic_macros_msa.h
index fbe7abf..e2c9de2 100644
--- a/libavutil/mips/generic_macros_msa.h
+++ b/libavutil/mips/generic_macros_msa.h
@@ -278,6 +278,7 @@
     out0 = LD_B(RTYPE, (psrc));                 \
     out1 = LD_B(RTYPE, (psrc) + stride);        \
 }
+#define LD_UB2(...) LD_B2(v16u8, __VA_ARGS__)
 #define LD_SB2(...) LD_B2(v16i8, __VA_ARGS__)
 
 #define LD_B3(RTYPE, psrc, stride, out0, out1, out2)  \
@@ -349,6 +350,14 @@
 #define ST_UB4(...) ST_B4(v16u8, __VA_ARGS__)
 #define ST_SB4(...) ST_B4(v16i8, __VA_ARGS__)
 
+#define ST_B8(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7,        \
+              pdst, stride)                                         \
+{                                                                   \
+    ST_B4(RTYPE, in0, in1, in2, in3, pdst, stride);                 \
+    ST_B4(RTYPE, in4, in5, in6, in7, (pdst) + 4 * stride, stride);  \
+}
+#define ST_UB8(...) ST_B8(v16u8, __VA_ARGS__)
+
 /* Description : Store vectors of 8 halfword elements with stride
    Arguments   : Inputs  - in0, in1, stride
                  Outputs - pdst    (destination pointer to store to)
@@ -425,6 +434,26 @@
     SH(out3_m, pblk_2x4_m + 3 * stride);               \
 }
 
+/* Description : Store as 4x2 byte block to destination memory from input vector
+   Arguments   : Inputs  - in, pdst, stride
+                 Return Type - unsigned byte
+   Details     : Index 0 word element from input vector is copied and stored
+                 on first line
+                 Index 1 word element from input vector is copied and stored
+                 on second line
+*/
+#define ST4x2_UB(in, pdst, stride)             \
+{                                              \
+    uint32_t out0_m, out1_m;                   \
+    uint8_t *pblk_4x2_m = (uint8_t *) (pdst);  \
+                                               \
+    out0_m = __msa_copy_u_w((v4i32) in, 0);    \
+    out1_m = __msa_copy_u_w((v4i32) in, 1);    \
+                                               \
+    SW(out0_m, pblk_4x2_m);                    \
+    SW(out1_m, pblk_4x2_m + stride);           \
+}
+
 /* Description : Store as 4x4 byte block to destination memory from input vector
    Arguments   : Inputs  - in0, in1, pdst, stride
                  Return Type - unsigned byte
@@ -598,7 +627,18 @@
     out0 = (RTYPE) __msa_vshf_b((v16i8) mask0, (v16i8) in1, (v16i8) in0);  \
     out1 = (RTYPE) __msa_vshf_b((v16i8) mask1, (v16i8) in3, (v16i8) in2);  \
 }
+#define VSHF_B2_UB(...) VSHF_B2(v16u8, __VA_ARGS__)
 #define VSHF_B2_SB(...) VSHF_B2(v16i8, __VA_ARGS__)
+#define VSHF_B2_UH(...) VSHF_B2(v8u16, __VA_ARGS__)
+#define VSHF_B2_SH(...) VSHF_B2(v8i16, __VA_ARGS__)
+
+#define VSHF_B3(RTYPE, in0, in1, in2, in3, in4, in5, mask0, mask1, mask2,  \
+                out0, out1, out2)                                          \
+{                                                                          \
+    VSHF_B2(RTYPE, in0, in1, in2, in3, mask0, mask1, out0, out1);          \
+    out2 = (RTYPE) __msa_vshf_b((v16i8) mask2, (v16i8) in5, (v16i8) in4);  \
+}
+#define VSHF_B3_SB(...) VSHF_B3(v16i8, __VA_ARGS__)
 
 #define VSHF_B4(RTYPE, in0, in1, mask0, mask1, mask2, mask3,       \
                 out0, out1, out2, out3)                            \
@@ -608,6 +648,57 @@
 }
 #define VSHF_B4_SB(...) VSHF_B4(v16i8, __VA_ARGS__)
 
+/* Description : Shuffle byte vector elements as per mask vector
+   Arguments   : Inputs  - in0, in1, in2, in3, mask0, mask1
+                 Outputs - out0, out1
+                 Return Type - as per RTYPE
+   Details     : Selective byte elements from in0 & in1 are copied to out0 as
+                 per control vector mask0
+                 Selective byte elements from in2 & in3 are copied to out1 as
+                 per control vector mask1
+*/
+#define VSHF_W2(RTYPE, in0, in1, in2, in3, mask0, mask1, out0, out1)      \
+{                                                                         \
+    out0 = (RTYPE) __msa_vshf_w((v4i32) mask0, (v4i32) in1, (v4i32) in0); \
+    out1 = (RTYPE) __msa_vshf_w((v4i32) mask1, (v4i32) in3, (v4i32) in2); \
+}
+#define VSHF_W2_SB(...) VSHF_W2(v16i8, __VA_ARGS__)
+
+/* Description : Dot product of byte vector elements
+   Arguments   : Inputs  - mult0, mult1
+                           cnst0, cnst1
+                 Outputs - out0, out1
+                 Return Type - signed halfword
+   Details     : Signed byte elements from mult0 are multiplied with
+                 signed byte elements from cnst0 producing a result
+                 twice the size of input i.e. signed halfword.
+                 Then this multiplication results of adjacent odd-even elements
+                 are added together and stored to the out vector
+                 (2 signed halfword results)
+*/
+#define DOTP_SB2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1)   \
+{                                                                 \
+    out0 = (RTYPE) __msa_dotp_s_h((v16i8) mult0, (v16i8) cnst0);  \
+    out1 = (RTYPE) __msa_dotp_s_h((v16i8) mult1, (v16i8) cnst1);  \
+}
+#define DOTP_SB2_SH(...) DOTP_SB2(v8i16, __VA_ARGS__)
+
+#define DOTP_SB3(RTYPE, mult0, mult1, mult2, cnst0, cnst1, cnst2,  \
+                 out0, out1, out2)                                 \
+{                                                                  \
+    DOTP_SB2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1);       \
+    out2 = (RTYPE) __msa_dotp_s_h((v16i8) mult2, (v16i8) cnst2);   \
+}
+#define DOTP_SB3_SH(...) DOTP_SB3(v8i16, __VA_ARGS__)
+
+#define DOTP_SB4(RTYPE, mult0, mult1, mult2, mult3,                   \
+                 cnst0, cnst1, cnst2, cnst3, out0, out1, out2, out3)  \
+{                                                                     \
+    DOTP_SB2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1);          \
+    DOTP_SB2(RTYPE, mult2, mult3, cnst2, cnst3, out2, out3);          \
+}
+#define DOTP_SB4_SH(...) DOTP_SB4(v8i16, __VA_ARGS__)
+
 /* Description : Dot product & addition of byte vector elements
    Arguments   : Inputs  - mult0, mult1
                            cnst0, cnst1
@@ -701,6 +792,22 @@
     CLIP_SH2_0_255(in2, in3);               \
 }
 
+/* Description : Clips all signed word elements of input vector
+                 between 0 & 255
+   Arguments   : Inputs  - in       (input vector)
+                 Outputs - out_m    (output vector with clipped elements)
+                 Return Type - signed word
+*/
+#define CLIP_SW_0_255(in)                                 \
+( {                                                       \
+    v4i32 max_m = __msa_ldi_w(255);                       \
+    v4i32 out_m;                                          \
+                                                          \
+    out_m = __msa_maxi_s_w((v4i32) in, 0);                \
+    out_m = __msa_min_s_w((v4i32) max_m, (v4i32) out_m);  \
+    out_m;                                                \
+} )
+
 /* Description : Horizontal subtraction of unsigned byte vector elements
    Arguments   : Inputs  - in0, in1
                  Outputs - out0, out1
@@ -1021,6 +1128,37 @@
 }
 #define SAT_UH4_UH(...) SAT_UH4(v8u16, __VA_ARGS__)
 
+/* Description : Saturate the halfword element values to the max
+                 unsigned value of (sat_val+1 bits)
+                 The element data width remains unchanged
+   Arguments   : Inputs  - in0, in1, in2, in3, sat_val
+                 Outputs - in0, in1, in2, in3 (in place)
+                 Return Type - unsigned halfword
+   Details     : Each unsigned halfword element from 'in0' is saturated to the
+                 value generated with (sat_val+1) bit range
+                 Results are in placed to original vectors
+*/
+#define SAT_SH2(RTYPE, in0, in1, sat_val)               \
+{                                                       \
+    in0 = (RTYPE) __msa_sat_s_h((v8i16) in0, sat_val);  \
+    in1 = (RTYPE) __msa_sat_s_h((v8i16) in1, sat_val);  \
+}
+#define SAT_SH2_SH(...) SAT_SH2(v8i16, __VA_ARGS__)
+
+#define SAT_SH3(RTYPE, in0, in1, in2, sat_val)          \
+{                                                       \
+    SAT_SH2(RTYPE, in0, in1, sat_val)                   \
+    in2 = (RTYPE) __msa_sat_s_h((v8i16) in2, sat_val);  \
+}
+#define SAT_SH3_SH(...) SAT_SH3(v8i16, __VA_ARGS__)
+
+#define SAT_SH4(RTYPE, in0, in1, in2, in3, sat_val)  \
+{                                                    \
+    SAT_SH2(RTYPE, in0, in1, sat_val);               \
+    SAT_SH2(RTYPE, in2, in3, sat_val);               \
+}
+#define SAT_SH4_SH(...) SAT_SH4(v8i16, __VA_ARGS__)
+
 /* Description : Indexed halfword element values are replicated to all
                  elements in output vector
    Arguments   : Inputs  - in, idx0, idx1
@@ -1043,6 +1181,7 @@
     SPLATI_H2(RTYPE, in, idx0, idx1, out0, out1);     \
     SPLATI_H2(RTYPE, in, idx2, idx3, out2, out3);     \
 }
+#define SPLATI_H4_SB(...) SPLATI_H4(v16i8, __VA_ARGS__)
 #define SPLATI_H4_SH(...) SPLATI_H4(v8i16, __VA_ARGS__)
 
 /* Description : Indexed word element values are replicated to all
@@ -1097,6 +1236,7 @@
     out2 = (RTYPE) __msa_pckev_b((v16i8) in4, (v16i8) in5);              \
 }
 #define PCKEV_B3_UB(...) PCKEV_B3(v16u8, __VA_ARGS__)
+#define PCKEV_B3_SB(...) PCKEV_B3(v16i8, __VA_ARGS__)
 
 #define PCKEV_B4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7,  \
                  out0, out1, out2, out3)                         \
@@ -1123,6 +1263,7 @@
     out0 = (RTYPE) __msa_pckev_h((v8i16) in0, (v8i16) in1);  \
     out1 = (RTYPE) __msa_pckev_h((v8i16) in2, (v8i16) in3);  \
 }
+#define PCKEV_H2_SH(...) PCKEV_H2(v8i16, __VA_ARGS__)
 #define PCKEV_H2_SW(...) PCKEV_H2(v4i32, __VA_ARGS__)
 
 #define PCKEV_H4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7,  \
@@ -1131,6 +1272,7 @@
     PCKEV_H2(RTYPE, in0, in1, in2, in3, out0, out1);             \
     PCKEV_H2(RTYPE, in4, in5, in6, in7, out2, out3);             \
 }
+#define PCKEV_H4_SH(...) PCKEV_H4(v8i16, __VA_ARGS__)
 #define PCKEV_H4_SW(...) PCKEV_H4(v4i32, __VA_ARGS__)
 
 /* Description : Each byte element is logically xor'ed with immediate 128
@@ -1212,6 +1354,7 @@
     ADDS_SH2(RTYPE, in4, in5, in6, in7, out2, out3);             \
 }
 #define ADDS_SH4_UH(...) ADDS_SH4(v8u16, __VA_ARGS__)
+#define ADDS_SH4_SH(...) ADDS_SH4(v8i16, __VA_ARGS__)
 
 /* Description : Shift left all elements of vector (generic for all data types)
    Arguments   : Inputs  - in0, in1, in2, in3, shift
@@ -1266,6 +1409,64 @@
 }
 #define SRL_H4_UH(...) SRL_H4(v8u16, __VA_ARGS__)
 
+/* Description : Shift right arithmetic rounded halfwords
+   Arguments   : Inputs  - in0, in1, shift
+                 Outputs - in0, in1, (in place)
+                 Return Type - unsigned halfword
+   Details     : Each element of vector 'in0' is shifted right arithmetic by
+                 number of bits respective element holds in vector 'shift'.
+                 The last discarded bit is added to shifted value for rounding
+                 and the result is in place written to 'in0'
+                 Here, 'shift' is a vector passed in
+                 Similar for other pairs
+*/
+#define SRAR_H2(RTYPE, in0, in1, shift)                      \
+{                                                            \
+    in0 = (RTYPE) __msa_srar_h((v8i16) in0, (v8i16) shift);  \
+    in1 = (RTYPE) __msa_srar_h((v8i16) in1, (v8i16) shift);  \
+}
+#define SRAR_H2_UH(...) SRAR_H2(v8u16, __VA_ARGS__)
+#define SRAR_H2_SH(...) SRAR_H2(v8i16, __VA_ARGS__)
+
+#define SRAR_H3(RTYPE, in0, in1, in2, shift)                 \
+{                                                            \
+    SRAR_H2(RTYPE, in0, in1, shift)                          \
+    in2 = (RTYPE) __msa_srar_h((v8i16) in2, (v8i16) shift);  \
+}
+#define SRAR_H3_SH(...) SRAR_H3(v8i16, __VA_ARGS__)
+
+#define SRAR_H4(RTYPE, in0, in1, in2, in3, shift)  \
+{                                                  \
+    SRAR_H2(RTYPE, in0, in1, shift)                \
+    SRAR_H2(RTYPE, in2, in3, shift)                \
+}
+#define SRAR_H4_UH(...) SRAR_H4(v8u16, __VA_ARGS__)
+#define SRAR_H4_SH(...) SRAR_H4(v8i16, __VA_ARGS__)
+/* Description : Shift right arithmetic rounded (immediate)
+   Arguments   : Inputs  - in0, in1, shift
+                 Outputs - in0, in1     (in place)
+                 Return Type - as per RTYPE
+   Details     : Each element of vector 'in0' is shifted right arithmetic by
+                 value in 'shift'.
+                 The last discarded bit is added to shifted value for rounding
+                 and the result is in place written to 'in0'
+                 Similar for other pairs
+*/
+#define SRARI_W2(RTYPE, in0, in1, shift)              \
+{                                                     \
+    in0 = (RTYPE) __msa_srari_w((v4i32) in0, shift);  \
+    in1 = (RTYPE) __msa_srari_w((v4i32) in1, shift);  \
+}
+#define SRARI_W2_SW(...) SRARI_W2(v4i32, __VA_ARGS__)
+
+#define SRARI_W4(RTYPE, in0, in1, in2, in3, shift)  \
+{                                                   \
+    SRARI_W2(RTYPE, in0, in1, shift);               \
+    SRARI_W2(RTYPE, in2, in3, shift);               \
+}
+#define SRARI_W4_SH(...) SRARI_W4(v8i16, __VA_ARGS__)
+#define SRARI_W4_SW(...) SRARI_W4(v4i32, __VA_ARGS__)
+
 /* Description : Multiplication of pairs of vectors
    Arguments   : Inputs  - in0, in1, in2, in3
                  Outputs - out0, out1
@@ -1392,6 +1593,22 @@
     out7 = (v16u8) __msa_ilvod_w((v4i32) tmp3_m, (v4i32) tmp2_m);            \
 }
 
+/* Description : Pack even elements of input vectors & xor with 128
+   Arguments   : Inputs  - in0, in1
+                 Outputs - out_m
+                 Return Type - unsigned byte
+   Details     : Signed byte even elements from 'in0' and 'in1' are packed
+                 together in one vector and the resulted vector is xor'ed with
+                 128 to shift the range from signed to unsigned byte
+*/
+#define PCKEV_XORI128_UB(in0, in1)                            \
+( {                                                           \
+    v16u8 out_m;                                              \
+    out_m = (v16u8) __msa_pckev_b((v16i8) in1, (v16i8) in0);  \
+    out_m = (v16u8) __msa_xori_b((v16u8) out_m, 128);         \
+    out_m;                                                    \
+} )
+
 /* Description : Pack even byte elements, extract 0 & 2 index words from pair
                  of results and store 4 words in destination memory as per
                  stride



More information about the ffmpeg-cvslog mailing list