avutil/mips: refine msa macros CLIP_*.

Changing details as following: 1. Remove the local variable 'out_m' in 'CLIP_SH' and store the result in source vector. 2. Refine the implementation of macro 'CLIP_SH_0_255' and 'CLIP_SW_0_255'. Performance of VP8 decoding has speed up about 1.1%(from 7.03x to 7.11x). Performance of H264 decoding has speed up about 0.5%(from 4.35x to 4.37x). Performance of Theora decoding has speed up about 0.7%(from 5.79x to 5.83x). 3. Remove redundant macro 'CLIP_SH/Wn_0_255_MAX_SATU' and use 'CLIP_SH/Wn_0_255' instead, because there are no difference in the effect of this two macros. Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
5 years ago · a3e572d96f
--- a/libavcodec/mips/h264dsp_msa.c
+++ b/libavcodec/mips/h264dsp_msa.c
@@ -413,8 +413,7 @@ static void avc_biwgt_8x8_msa(uint8_t *src, uint8_t *dst, int32_t stride,
    tmp7 = __msa_dpadd_s_h(offset, wgt, vec7);
    SRA_4V(tmp0, tmp1, tmp2, tmp3, denom);
    SRA_4V(tmp4, tmp5, tmp6, tmp7, denom);
    CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
    CLIP_SH4_0_255(tmp4, tmp5, tmp6, tmp7);
    CLIP_SH8_0_255(tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7);
    PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, dst0, dst1);
    PCKEV_B2_UB(tmp5, tmp4, tmp7, tmp6, dst2, dst3);
    ST_D8(dst0, dst1, dst2, dst3, 0, 1, 0, 1, 0, 1, 0, 1, dst, stride);
@@ -475,8 +474,7 @@ static void avc_biwgt_8x16_msa(uint8_t *src, uint8_t *dst, int32_t stride,
        SRA_4V(temp0, temp1, temp2, temp3, denom);
        SRA_4V(temp4, temp5, temp6, temp7, denom);
        CLIP_SH4_0_255(temp0, temp1, temp2, temp3);
        CLIP_SH4_0_255(temp4, temp5, temp6, temp7);
        CLIP_SH8_0_255(temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7);
        PCKEV_B4_UB(temp1, temp0, temp3, temp2, temp5, temp4, temp7, temp6,
                    dst0, dst1, dst2, dst3);
        ST_D8(dst0, dst1, dst2, dst3, 0, 1, 0, 1, 0, 1, 0, 1, dst, stride);
@@ -531,7 +529,7 @@ static void avc_biwgt_8x16_msa(uint8_t *src, uint8_t *dst, int32_t stride,
    temp = p1_or_q1_org_in << 1;                              \
    clip3 = clip3 - temp;                                     \
    clip3 = __msa_ave_s_h(p2_or_q2_org_in, clip3);            \
    clip3 = CLIP_SH(clip3, negate_tc_in, tc_in);              \
    CLIP_SH(clip3, negate_tc_in, tc_in);                      \
    p1_or_q1_out = p1_or_q1_org_in + clip3;                   \
 }
@@ -549,7 +547,7 @@ static void avc_biwgt_8x16_msa(uint8_t *src, uint8_t *dst, int32_t stride,
    delta = q0_sub_p0 + p1_sub_q1;                              \
    delta >>= 3;                                                \
                                                                \
    delta = CLIP_SH(delta, negate_threshold_in, threshold_in);  \
    CLIP_SH(delta, negate_threshold_in, threshold_in);          \
                                                                \
    p0_or_q0_out = p0_or_q0_org_in + delta;                     \
    q0_or_p0_out = q0_or_p0_org_in - delta;                     \
@@ -598,7 +596,7 @@ static void avc_biwgt_8x16_msa(uint8_t *src, uint8_t *dst, int32_t stride,
    delta = q0_sub_p0 + p1_sub_q1;                                       \
    delta = __msa_srari_h(delta, 3);                                     \
                                                                         \
    delta = CLIP_SH(delta, -tc, tc);                                     \
    CLIP_SH(delta, -tc, tc);                                             \
                                                                         \
    ILVR_B2_SH(zeros, src1, zeros, src2, res0_r, res1_r);                \
                                                                         \
@@ -662,7 +660,7 @@ static void avc_biwgt_8x16_msa(uint8_t *src, uint8_t *dst, int32_t stride,
    q0_sub_p0 <<= 2;                                                       \
    delta = q0_sub_p0 + p1_sub_q1;                                         \
    delta = __msa_srari_h(delta, 3);                                       \
    delta = CLIP_SH(delta, -tc, tc);                                       \
    CLIP_SH(delta, -tc, tc);                                               \
                                                                           \
    ILVR_B2_SH(zeros, src1, zeros, src2, res0_r, res1_r);                  \
                                                                           \
@@ -1741,7 +1739,7 @@ static void avc_h_loop_filter_luma_mbaff_msa(uint8_t *in, int32_t stride,
    v8i16 tc, tc_orig_r, tc_plus1;
    v16u8 is_tc_orig1, is_tc_orig2, tc_orig = { 0 };
    v8i16 p0_ilvr_q0, p0_add_q0, q0_sub_p0, p1_sub_q1;
    v8u16 src2_r, src3_r;
    v8i16 src2_r, src3_r;
    v8i16 p2_r, p1_r, q2_r, q1_r;
    v16u8 p2, q2, p0, q0;
    v4i32 dst0, dst1;
@@ -1839,8 +1837,8 @@ static void avc_h_loop_filter_luma_mbaff_msa(uint8_t *in, int32_t stride,
    tc_orig_r = (v8i16) __msa_ilvr_b(zeros, (v16i8) tc_orig);
    tc = tc_orig_r;
    p2_r = CLIP_SH(p2_r, -tc_orig_r, tc_orig_r);
    q2_r = CLIP_SH(q2_r, -tc_orig_r, tc_orig_r);
    CLIP_SH(p2_r, -tc_orig_r, tc_orig_r);
    CLIP_SH(q2_r, -tc_orig_r, tc_orig_r);
    p2_r += p1_r;
    q2_r += q1_r;
@@ -1872,14 +1870,13 @@ static void avc_h_loop_filter_luma_mbaff_msa(uint8_t *in, int32_t stride,
                                              (v16i8) is_less_than_beta2);
    tc = (v8i16) __msa_bmnz_v((v16u8) tc, (v16u8) tc_plus1, is_less_than_beta2);
    q0_sub_p0 = CLIP_SH(q0_sub_p0, -tc, tc);
    CLIP_SH(q0_sub_p0, -tc, tc);
    ILVR_B2_UH(zeros, src2, zeros, src3, src2_r, src3_r);
    ILVR_B2_SH(zeros, src2, zeros, src3, src2_r, src3_r);
    src2_r += q0_sub_p0;
    src3_r -= q0_sub_p0;
    src2_r = (v8u16) CLIP_SH_0_255(src2_r);
    src3_r = (v8u16) CLIP_SH_0_255(src3_r);
    CLIP_SH2_0_255(src2_r, src3_r);
    PCKEV_B2_UB(src2_r, src2_r, src3_r, src3_r, p0, q0);
@@ -2509,10 +2506,8 @@ void ff_biweight_h264_pixels16_8_msa(uint8_t *dst, uint8_t *src,
    SRA_4V(tmp4, tmp5, tmp6, tmp7, denom);
    SRA_4V(tmp8, tmp9, tmp10, tmp11, denom);
    SRA_4V(tmp12, tmp13, tmp14, tmp15, denom);
    CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
    CLIP_SH4_0_255(tmp4, tmp5, tmp6, tmp7);
    CLIP_SH4_0_255(tmp8, tmp9, tmp10, tmp11);
    CLIP_SH4_0_255(tmp12, tmp13, tmp14, tmp15);
    CLIP_SH8_0_255(tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7);
    CLIP_SH8_0_255(tmp8, tmp9, tmp10, tmp11, tmp12, tmp13, tmp14, tmp15);
    PCKEV_B4_UB(tmp1, tmp0, tmp3, tmp2, tmp5, tmp4, tmp7, tmp6, dst0, dst1,
                dst2, dst3);
    PCKEV_B4_UB(tmp9, tmp8, tmp11, tmp10, tmp13, tmp12, tmp15, tmp14, dst4,
@@ -2553,10 +2548,8 @@ void ff_biweight_h264_pixels16_8_msa(uint8_t *dst, uint8_t *src,
        SRA_4V(tmp4, tmp5, tmp6, tmp7, denom);
        SRA_4V(tmp8, tmp9, tmp10, tmp11, denom);
        SRA_4V(tmp12, tmp13, tmp14, tmp15, denom);
        CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
        CLIP_SH4_0_255(tmp4, tmp5, tmp6, tmp7);
        CLIP_SH4_0_255(tmp8, tmp9, tmp10, tmp11);
        CLIP_SH4_0_255(tmp12, tmp13, tmp14, tmp15);
        CLIP_SH8_0_255(tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7);
        CLIP_SH8_0_255(tmp8, tmp9, tmp10, tmp11, tmp12, tmp13, tmp14, tmp15);
        PCKEV_B4_UB(tmp1, tmp0, tmp3, tmp2, tmp5, tmp4, tmp7, tmp6, dst0, dst1,
                    dst2, dst3);
        PCKEV_B4_UB(tmp9, tmp8, tmp11, tmp10, tmp13, tmp12, tmp15, tmp14, dst4,
--- a/libavcodec/mips/h264idct_msa.c
+++ b/libavcodec/mips/h264idct_msa.c
@@ -233,8 +233,7 @@ static void avc_idct8_addblk_msa(uint8_t *dst, int16_t *src, int32_t dst_stride)
         res0, res1, res2, res3);
    ADD4(res4, tmp4, res5, tmp5, res6, tmp6, res7, tmp7,
         res4, res5, res6, res7);
    CLIP_SH4_0_255(res0, res1, res2, res3);
    CLIP_SH4_0_255(res4, res5, res6, res7);
    CLIP_SH8_0_255(res0, res1, res2, res3, res4, res5, res6, res7);
    PCKEV_B4_SB(res1, res0, res3, res2, res5, res4, res7, res6,
                dst0, dst1, dst2, dst3);
    ST_D8(dst0, dst1, dst2, dst3, 0, 1, 0, 1, 0, 1, 0, 1, dst, dst_stride)
@@ -263,8 +262,8 @@ static void avc_idct8_dc_addblk_msa(uint8_t *dst, int16_t *src,
         dst0_r, dst1_r, dst2_r, dst3_r);
    ADD4(dst4_r, dc, dst5_r, dc, dst6_r, dc, dst7_r, dc,
         dst4_r, dst5_r, dst6_r, dst7_r);
    CLIP_SH4_0_255(dst0_r, dst1_r, dst2_r, dst3_r);
    CLIP_SH4_0_255(dst4_r, dst5_r, dst6_r, dst7_r);
    CLIP_SH8_0_255(dst0_r, dst1_r, dst2_r, dst3_r,
                   dst4_r, dst5_r, dst6_r, dst7_r);
    PCKEV_B4_SB(dst1_r, dst0_r, dst3_r, dst2_r, dst5_r, dst4_r, dst7_r, dst6_r,
                dst0, dst1, dst2, dst3);
    ST_D8(dst0, dst1, dst2, dst3, 0, 1, 0, 1, 0, 1, 0, 1, dst, dst_stride)
--- a/libavcodec/mips/hevc_idct_msa.c
+++ b/libavcodec/mips/hevc_idct_msa.c
@@ -803,8 +803,9 @@ static void hevc_addblk_16x16_msa(int16_t *coeffs, uint8_t *dst, int32_t stride)
        LD_SH4((coeffs + 8), 16, in1, in3, in5, in7);
        coeffs += 64;
        CLIP_SH4_0_255(dst_r0, dst_l0, dst_r1, dst_l1);
        CLIP_SH4_0_255(dst_r2, dst_l2, dst_r3, dst_l3);
        CLIP_SH8_0_255(dst_r0, dst_l0, dst_r1, dst_l1,
                       dst_r2, dst_l2, dst_r3, dst_l3);
        PCKEV_B4_UB(dst_l0, dst_r0, dst_l1, dst_r1, dst_l2, dst_r2, dst_l3,
                    dst_r3, dst0, dst1, dst2, dst3);
        ST_UB4(dst0, dst1, dst2, dst3, dst, stride);
@@ -825,8 +826,8 @@ static void hevc_addblk_16x16_msa(int16_t *coeffs, uint8_t *dst, int32_t stride)
    dst_r3 += in6;
    dst_l3 += in7;
    CLIP_SH4_0_255(dst_r0, dst_l0, dst_r1, dst_l1);
    CLIP_SH4_0_255(dst_r2, dst_l2, dst_r3, dst_l3);
    CLIP_SH8_0_255(dst_r0, dst_l0, dst_r1, dst_l1,
                   dst_r2, dst_l2, dst_r3, dst_l3);
    PCKEV_B4_UB(dst_l0, dst_r0, dst_l1, dst_r1, dst_l2, dst_r2, dst_l3,
                dst_r3, dst0, dst1, dst2, dst3);
    ST_UB4(dst0, dst1, dst2, dst3, dst, stride);
@@ -873,8 +874,8 @@ static void hevc_addblk_32x32_msa(int16_t *coeffs, uint8_t *dst, int32_t stride)
        LD_SH4((coeffs + 8), 16, in1, in3, in5, in7);
        coeffs += 64;
        CLIP_SH4_0_255(dst_r0, dst_l0, dst_r1, dst_l1);
        CLIP_SH4_0_255(dst_r2, dst_l2, dst_r3, dst_l3);
        CLIP_SH8_0_255(dst_r0, dst_l0, dst_r1, dst_l1,
                       dst_r2, dst_l2, dst_r3, dst_l3);
        PCKEV_B4_UB(dst_l0, dst_r0, dst_l1, dst_r1, dst_l2, dst_r2, dst_l3,
                    dst_r3, dst0, dst1, dst2, dst3);
        ST_UB2(dst0, dst1, dst, 16);
@@ -905,8 +906,8 @@ static void hevc_addblk_32x32_msa(int16_t *coeffs, uint8_t *dst, int32_t stride)
    LD_SH4(coeffs, 16, in0, in2, in4, in6);
    LD_SH4((coeffs + 8), 16, in1, in3, in5, in7);
    CLIP_SH4_0_255(dst_r0, dst_l0, dst_r1, dst_l1);
    CLIP_SH4_0_255(dst_r2, dst_l2, dst_r3, dst_l3);
    CLIP_SH8_0_255(dst_r0, dst_l0, dst_r1, dst_l1,
                   dst_r2, dst_l2, dst_r3, dst_l3);
    PCKEV_B4_UB(dst_l0, dst_r0, dst_l1, dst_r1, dst_l2, dst_r2, dst_l3,
                dst_r3, dst0, dst1, dst2, dst3);
    ST_UB2(dst0, dst1, dst, 16);
@@ -928,8 +929,8 @@ static void hevc_addblk_32x32_msa(int16_t *coeffs, uint8_t *dst, int32_t stride)
    dst_r3 += in6;
    dst_l3 += in7;
    CLIP_SH4_0_255(dst_r0, dst_l0, dst_r1, dst_l1);
    CLIP_SH4_0_255(dst_r2, dst_l2, dst_r3, dst_l3);
    CLIP_SH8_0_255(dst_r0, dst_l0, dst_r1, dst_l1,
                   dst_r2, dst_l2, dst_r3, dst_l3);
    PCKEV_B4_UB(dst_l0, dst_r0, dst_l1, dst_r1, dst_l2, dst_r2, dst_l3,
                dst_r3, dst0, dst1, dst2, dst3);
    ST_UB2(dst0, dst1, dst, 16);
--- a/libavcodec/mips/hevc_lpf_sao_msa.c
+++ b/libavcodec/mips/hevc_lpf_sao_msa.c
@@ -140,19 +140,19 @@ static void hevc_loopfilter_luma_hor_msa(uint8_t *src, int32_t stride,
            temp1 = ((p3_src + p2_src) << 1) + p2_src + temp0;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 3);
            temp2 = (v8i16) (temp1 - p2_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst0 = (v16u8) (temp2 + (v8i16) p2_src);
            temp1 = temp0 + p2_src;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 2);
            temp2 = (v8i16) (temp1 - p1_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst1 = (v16u8) (temp2 + (v8i16) p1_src);
            temp1 = (temp0 << 1) + p2_src + q1_src;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 3);
            temp2 = (v8i16) (temp1 - p0_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst2 = (v16u8) (temp2 + (v8i16) p0_src);
            dst0 = __msa_bmz_v(dst0, (v16u8) p2_src, (v16u8) p_is_pcm_vec);
@@ -165,19 +165,19 @@ static void hevc_loopfilter_luma_hor_msa(uint8_t *src, int32_t stride,
            temp1 = ((q3_src + q2_src) << 1) + q2_src + temp0;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 3);
            temp2 = (v8i16) (temp1 - q2_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst5 = (v16u8) (temp2 + (v8i16) q2_src);
            temp1 = temp0 + q2_src;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 2);
            temp2 = (v8i16) (temp1 - q1_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst4 = (v16u8) (temp2 + (v8i16) q1_src);
            temp1 = (temp0 << 1) + p1_src + q2_src;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 3);
            temp2 = (v8i16) (temp1 - q0_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst3 = (v16u8) (temp2 + (v8i16) q0_src);
            dst3 = __msa_bmz_v(dst3, (v16u8) q0_src, (v16u8) q_is_pcm_vec);
@@ -218,15 +218,15 @@ static void hevc_loopfilter_luma_hor_msa(uint8_t *src, int32_t stride,
            abs_delta0 = __msa_add_a_h(delta0, (v8i16) zero);
            abs_delta0 = (v8u16) abs_delta0 < temp1;
            delta0 = CLIP_SH(delta0, tc_neg, tc_pos);
            CLIP_SH(delta0, tc_neg, tc_pos);
            temp0 = (v8u16) (delta0 + p0_src);
            temp0 = (v8u16) CLIP_SH_0_255(temp0);
            temp0 = (v8u16) __msa_bmz_v((v16u8) temp0, (v16u8) p0_src,
            temp2 = (v8i16) (delta0 + p0_src);
            CLIP_SH_0_255(temp2);
            temp0 = (v8u16) __msa_bmz_v((v16u8) temp2, (v16u8) p0_src,
                                        (v16u8) p_is_pcm_vec);
            temp2 = (v8i16) (q0_src - delta0);
            temp2 = CLIP_SH_0_255(temp2);
            CLIP_SH_0_255(temp2);
            temp2 = (v8i16) __msa_bmz_v((v16u8) temp2, (v16u8) q0_src,
                                        (v16u8) q_is_pcm_vec);
@@ -252,9 +252,9 @@ static void hevc_loopfilter_luma_hor_msa(uint8_t *src, int32_t stride,
            delta1 -= (v8i16) p1_src;
            delta1 += delta0;
            delta1 >>= 1;
            delta1 = CLIP_SH(delta1, tc_neg, tc_pos);
            CLIP_SH(delta1, tc_neg, tc_pos);
            delta1 = (v8i16) p1_src + (v8i16) delta1;
            delta1 = CLIP_SH_0_255(delta1);
            CLIP_SH_0_255(delta1);
            delta1 = (v8i16) __msa_bmnz_v((v16u8) delta1, (v16u8) p1_src,
                                          (v16u8) p_is_pcm_vec);
@@ -262,9 +262,9 @@ static void hevc_loopfilter_luma_hor_msa(uint8_t *src, int32_t stride,
            delta2 = delta2 - (v8i16) q1_src;
            delta2 = delta2 - delta0;
            delta2 = delta2 >> 1;
            delta2 = CLIP_SH(delta2, tc_neg, tc_pos);
            CLIP_SH(delta2, tc_neg, tc_pos);
            delta2 = (v8i16) q1_src + (v8i16) delta2;
            delta2 = CLIP_SH_0_255(delta2);
            CLIP_SH_0_255(delta2);
            delta2 = (v8i16) __msa_bmnz_v((v16u8) delta2, (v16u8) q1_src,
                                          (v16u8) q_is_pcm_vec);
@@ -298,19 +298,19 @@ static void hevc_loopfilter_luma_hor_msa(uint8_t *src, int32_t stride,
            temp1 = ((p3_src + p2_src) << 1) + p2_src + temp0;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 3);
            temp2 = (v8i16) (temp1 - p2_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst0 = (v16u8) (temp2 + (v8i16) p2_src);
            temp1 = temp0 + p2_src;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 2);
            temp2 = (v8i16) (temp1 - p1_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst1 = (v16u8) (temp2 + (v8i16) p1_src);
            temp1 = (temp0 << 1) + p2_src + q1_src;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 3);
            temp2 = (v8i16) (temp1 - p0_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst2 = (v16u8) (temp2 + (v8i16) p0_src);
            dst0 = __msa_bmz_v(dst0, (v16u8) p2_src, (v16u8) p_is_pcm_vec);
@@ -323,19 +323,19 @@ static void hevc_loopfilter_luma_hor_msa(uint8_t *src, int32_t stride,
            temp1 = ((q3_src + q2_src) << 1) + q2_src + temp0;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 3);
            temp2 = (v8i16) (temp1 - q2_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst5 = (v16u8) (temp2 + (v8i16) q2_src);
            temp1 = temp0 + q2_src;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 2);
            temp2 = (v8i16) (temp1 - q1_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst4 = (v16u8) (temp2 + (v8i16) q1_src);
            temp1 = (temp0 << 1) + p1_src + q2_src;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 3);
            temp2 = (v8i16) (temp1 - q0_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst3 = (v16u8) (temp2 + (v8i16) q0_src);
            dst3 = __msa_bmz_v(dst3, (v16u8) q0_src, (v16u8) q_is_pcm_vec);
@@ -362,15 +362,15 @@ static void hevc_loopfilter_luma_hor_msa(uint8_t *src, int32_t stride,
            abs_delta0 = __msa_add_a_h(delta0, (v8i16) zero);
            abs_delta0 = (v8u16) abs_delta0 < temp1;
            delta0 = CLIP_SH(delta0, tc_neg, tc_pos);
            CLIP_SH(delta0, tc_neg, tc_pos);
            temp0 = (v8u16) (delta0 + p0_src);
            temp0 = (v8u16) CLIP_SH_0_255(temp0);
            temp0 = (v8u16) __msa_bmz_v((v16u8) temp0, (v16u8) p0_src,
            temp2 = (v8i16) (delta0 + p0_src);
            CLIP_SH_0_255(temp2);
            temp0 = (v8u16) __msa_bmz_v((v16u8) temp2, (v16u8) p0_src,
                                        (v16u8) p_is_pcm_vec);
            temp2 = (v8i16) (q0_src - delta0);
            temp2 = CLIP_SH_0_255(temp2);
            CLIP_SH_0_255(temp2);
            temp2 = (v8i16) __msa_bmz_v((v16u8) temp2, (v16u8) q0_src,
                                        (v16u8) q_is_pcm_vec);
@@ -394,9 +394,9 @@ static void hevc_loopfilter_luma_hor_msa(uint8_t *src, int32_t stride,
            delta1 -= (v8i16) p1_src;
            delta1 += delta0;
            delta1 >>= 1;
            delta1 = CLIP_SH(delta1, tc_neg, tc_pos);
            CLIP_SH(delta1, tc_neg, tc_pos);
            delta1 = (v8i16) p1_src + (v8i16) delta1;
            delta1 = CLIP_SH_0_255(delta1);
            CLIP_SH_0_255(delta1);
            delta1 = (v8i16) __msa_bmnz_v((v16u8) delta1, (v16u8) p1_src,
                                          (v16u8) p_is_pcm_vec);
@@ -404,9 +404,9 @@ static void hevc_loopfilter_luma_hor_msa(uint8_t *src, int32_t stride,
            delta2 = delta2 - (v8i16) q1_src;
            delta2 = delta2 - delta0;
            delta2 = delta2 >> 1;
            delta2 = CLIP_SH(delta2, tc_neg, tc_pos);
            CLIP_SH(delta2, tc_neg, tc_pos);
            delta2 = (v8i16) q1_src + (v8i16) delta2;
            delta2 = CLIP_SH_0_255(delta2);
            CLIP_SH_0_255(delta2);
            delta2 = (v8i16) __msa_bmnz_v((v16u8) delta2, (v16u8) q1_src,
                                          (v16u8) q_is_pcm_vec);
@@ -561,19 +561,19 @@ static void hevc_loopfilter_luma_ver_msa(uint8_t *src, int32_t stride,
            temp1 = ((p3_src + p2_src) << 1) + p2_src + temp0;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 3);
            temp2 = (v8i16) (temp1 - p2_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst0 = (v16u8) (temp2 + (v8i16) p2_src);
            temp1 = temp0 + p2_src;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 2);
            temp2 = (v8i16) (temp1 - p1_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst1 = (v16u8) (temp2 + (v8i16) p1_src);
            temp1 = (temp0 << 1) + p2_src + q1_src;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 3);
            temp2 = (v8i16) (temp1 - p0_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst2 = (v16u8) (temp2 + (v8i16) p0_src);
            dst0 = __msa_bmz_v(dst0, (v16u8) p2_src, (v16u8) p_is_pcm_vec);
@@ -585,19 +585,19 @@ static void hevc_loopfilter_luma_ver_msa(uint8_t *src, int32_t stride,
            temp1 = ((q3_src + q2_src) << 1) + q2_src + temp0;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 3);
            temp2 = (v8i16) (temp1 - q2_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst5 = (v16u8) (temp2 + (v8i16) q2_src);
            temp1 = temp0 + q2_src;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 2);
            temp2 = (v8i16) (temp1 - q1_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst4 = (v16u8) (temp2 + (v8i16) q1_src);
            temp1 = (temp0 << 1) + p1_src + q2_src;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 3);
            temp2 = (v8i16) (temp1 - q0_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst3 = (v16u8) (temp2 + (v8i16) q0_src);
            dst3 = __msa_bmz_v(dst3, (v16u8) q0_src, (v16u8) q_is_pcm_vec);
@@ -620,14 +620,14 @@ static void hevc_loopfilter_luma_ver_msa(uint8_t *src, int32_t stride,
            abs_delta0 = __msa_add_a_h(delta0, (v8i16) zero);
            abs_delta0 = (v8u16) abs_delta0 < temp1;
            delta0 = CLIP_SH(delta0, tc_neg, tc_pos);
            temp0 = (v8u16) (delta0 + p0_src);
            temp0 = (v8u16) CLIP_SH_0_255(temp0);
            temp0 = (v8u16) __msa_bmz_v((v16u8) temp0, (v16u8) p0_src,
            CLIP_SH(delta0, tc_neg, tc_pos);
            temp2 = (v8i16) (delta0 + p0_src);
            CLIP_SH_0_255(temp2);
            temp0 = (v8u16) __msa_bmz_v((v16u8) temp2, (v16u8) p0_src,
                                        (v16u8) p_is_pcm_vec);
            temp2 = (v8i16) (q0_src - delta0);
            temp2 = CLIP_SH_0_255(temp2);
            CLIP_SH_0_255(temp2);
            temp2 = (v8i16) __msa_bmz_v((v16u8) temp2, (v16u8) q0_src,
                                        (v16u8) q_is_pcm_vec);
@@ -649,9 +649,9 @@ static void hevc_loopfilter_luma_ver_msa(uint8_t *src, int32_t stride,
            delta1 -= (v8i16) p1_src;
            delta1 += delta0;
            delta1 >>= 1;
            delta1 = CLIP_SH(delta1, tc_neg, tc_pos);
            CLIP_SH(delta1, tc_neg, tc_pos);
            delta1 = (v8i16) p1_src + (v8i16) delta1;
            delta1 = CLIP_SH_0_255(delta1);
            CLIP_SH_0_255(delta1);
            delta1 = (v8i16) __msa_bmnz_v((v16u8) delta1, (v16u8) p1_src,
                                          (v16u8) p_is_pcm_vec);
@@ -659,9 +659,9 @@ static void hevc_loopfilter_luma_ver_msa(uint8_t *src, int32_t stride,
            delta2 = delta2 - (v8i16) q1_src;
            delta2 = delta2 - delta0;
            delta2 = delta2 >> 1;
            delta2 = CLIP_SH(delta2, tc_neg, tc_pos);
            CLIP_SH(delta2, tc_neg, tc_pos);
            delta2 = (v8i16) q1_src + (v8i16) delta2;
            delta2 = CLIP_SH_0_255(delta2);
            CLIP_SH_0_255(delta2);
            delta2 = (v8i16) __msa_bmnz_v((v16u8) delta2, (v16u8) q1_src,
                                          (v16u8) q_is_pcm_vec);
@@ -726,19 +726,19 @@ static void hevc_loopfilter_luma_ver_msa(uint8_t *src, int32_t stride,
            temp1 = ((p3_src + p2_src) << 1) + p2_src + temp0;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 3);
            temp2 = (v8i16) (temp1 - p2_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst0 = (v16u8) (temp2 + (v8i16) p2_src);
            temp1 = temp0 + p2_src;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 2);
            temp2 = (v8i16) (temp1 - p1_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst1 = (v16u8) (temp2 + (v8i16) p1_src);
            temp1 = (temp0 << 1) + p2_src + q1_src;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 3);
            temp2 = (v8i16) (temp1 - p0_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst2 = (v16u8) (temp2 + (v8i16) p0_src);
            dst0 = __msa_bmz_v(dst0, (v16u8) p2_src, (v16u8) p_is_pcm_vec);
@@ -750,19 +750,19 @@ static void hevc_loopfilter_luma_ver_msa(uint8_t *src, int32_t stride,
            temp1 = ((q3_src + q2_src) << 1) + q2_src + temp0;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 3);
            temp2 = (v8i16) (temp1 - q2_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst5 = (v16u8) (temp2 + (v8i16) q2_src);
            temp1 = temp0 + q2_src;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 2);
            temp2 = (v8i16) (temp1 - q1_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst4 = (v16u8) (temp2 + (v8i16) q1_src);
            temp1 = (temp0 << 1) + p1_src + q2_src;
            temp1 = (v8u16) __msa_srari_h((v8i16) temp1, 3);
            temp2 = (v8i16) (temp1 - q0_src);
            temp2 = CLIP_SH(temp2, tc_neg, tc_pos);
            CLIP_SH(temp2, tc_neg, tc_pos);
            dst3 = (v16u8) (temp2 + (v8i16) q0_src);
            dst3 = __msa_bmz_v(dst3, (v16u8) q0_src, (v16u8) q_is_pcm_vec);
@@ -785,15 +785,15 @@ static void hevc_loopfilter_luma_ver_msa(uint8_t *src, int32_t stride,
            abs_delta0 = __msa_add_a_h(delta0, (v8i16) zero);
            abs_delta0 = (v8u16) abs_delta0 < temp1;
            delta0 = CLIP_SH(delta0, tc_neg, tc_pos);
            CLIP_SH(delta0, tc_neg, tc_pos);
            temp0 = (v8u16) (delta0 + p0_src);
            temp0 = (v8u16) CLIP_SH_0_255(temp0);
            temp0 = (v8u16) __msa_bmz_v((v16u8) temp0, (v16u8) p0_src,
            temp2 = (v8i16) (delta0 + p0_src);
            CLIP_SH_0_255(temp2);
            temp0 = (v8u16) __msa_bmz_v((v16u8) temp2, (v16u8) p0_src,
                                        (v16u8) p_is_pcm_vec);
            temp2 = (v8i16) (q0_src - delta0);
            temp2 = CLIP_SH_0_255(temp2);
            CLIP_SH_0_255(temp2);
            temp2 = (v8i16) __msa_bmz_v((v16u8) temp2, (v16u8) q0_src,
                                        (v16u8) q_is_pcm_vec);
@@ -815,9 +815,9 @@ static void hevc_loopfilter_luma_ver_msa(uint8_t *src, int32_t stride,
            delta1 -= (v8i16) p1_src;
            delta1 += delta0;
            delta1 >>= 1;
            delta1 = CLIP_SH(delta1, tc_neg, tc_pos);
            CLIP_SH(delta1, tc_neg, tc_pos);
            delta1 = (v8i16) p1_src + (v8i16) delta1;
            delta1 = CLIP_SH_0_255(delta1);
            CLIP_SH_0_255(delta1);
            delta1 = (v8i16) __msa_bmnz_v((v16u8) delta1, (v16u8) p1_src,
                                          (v16u8) p_is_pcm_vec);
@@ -825,9 +825,9 @@ static void hevc_loopfilter_luma_ver_msa(uint8_t *src, int32_t stride,
            delta2 = delta2 - (v8i16) q1_src;
            delta2 = delta2 - delta0;
            delta2 = delta2 >> 1;
            delta2 = CLIP_SH(delta2, tc_neg, tc_pos);
            CLIP_SH(delta2, tc_neg, tc_pos);
            delta2 = (v8i16) q1_src + (v8i16) delta2;
            delta2 = CLIP_SH_0_255(delta2);
            CLIP_SH_0_255(delta2);
            delta2 = (v8i16) __msa_bmnz_v((v16u8) delta2, (v16u8) q1_src,
                                          (v16u8) q_is_pcm_vec);
            delta1 = (v8i16) __msa_bmz_v((v16u8) delta1, (v16u8) p1_src,
@@ -955,15 +955,15 @@ static void hevc_loopfilter_chroma_hor_msa(uint8_t *src, int32_t stride,
        temp0 <<= 2;
        temp0 += temp1;
        delta = __msa_srari_h((v8i16) temp0, 3);
        delta = CLIP_SH(delta, tc_neg, tc_pos);
        CLIP_SH(delta, tc_neg, tc_pos);
        temp0 = (v8i16) ((v8i16) p0 + delta);
        temp0 = CLIP_SH_0_255(temp0);
        CLIP_SH_0_255(temp0);
        temp0 = (v8i16) __msa_bmz_v((v16u8) temp0, (v16u8) p0,
                                    (v16u8) p_is_pcm_vec);
        temp1 = (v8i16) ((v8i16) q0 - delta);
        temp1 = CLIP_SH_0_255(temp1);
        CLIP_SH_0_255(temp1);
        temp1 = (v8i16) __msa_bmz_v((v16u8) temp1, (v16u8) q0,
                                    (v16u8) q_is_pcm_vec);
@@ -1014,15 +1014,15 @@ static void hevc_loopfilter_chroma_ver_msa(uint8_t *src, int32_t stride,
        temp0 <<= 2;
        temp0 += temp1;
        delta = __msa_srari_h((v8i16) temp0, 3);
        delta = CLIP_SH(delta, tc_neg, tc_pos);
        CLIP_SH(delta, tc_neg, tc_pos);
        temp0 = (v8i16) ((v8i16) p0 + delta);
        temp0 = CLIP_SH_0_255(temp0);
        CLIP_SH_0_255(temp0);
        temp0 = (v8i16) __msa_bmz_v((v16u8) temp0, (v16u8) p0,
                                    (v16u8) p_is_pcm_vec);
        temp1 = (v8i16) ((v8i16) q0 - delta);
        temp1 = CLIP_SH_0_255(temp1);
        CLIP_SH_0_255(temp1);
        temp1 = (v8i16) __msa_bmz_v((v16u8) temp1, (v16u8) q0,
                                    (v16u8) q_is_pcm_vec);
--- a/libavcodec/mips/hevc_mc_bi_msa.c
+++ b/libavcodec/mips/hevc_mc_bi_msa.c
@@ -48,7 +48,7 @@ static const uint8_t ff_hevc_mask_arr[16 * 2] __attribute__((aligned(0x40))) = {
 {                                                                  \
    ADDS_SH2_SH(vec0, in0, vec1, in1, out0, out1);                 \
    SRARI_H2_SH(out0, out1, rnd_val);                              \
    CLIP_SH2_0_255_MAX_SATU(out0, out1);                           \
    CLIP_SH2_0_255(out0, out1);                                    \
 }
 #define HEVC_BI_RND_CLIP4_MAX_SATU(in0, in1, in2, in3, vec0, vec1, vec2,    \
@@ -83,7 +83,7 @@ static void hevc_bi_copy_4w_msa(uint8_t *src0_ptr,
        dst0 <<= 6;
        dst0 += in0;
        dst0 = __msa_srari_h(dst0, 7);
        dst0 = CLIP_SH_0_255_MAX_SATU(dst0);
        CLIP_SH_0_255(dst0);
        dst0 = (v8i16) __msa_pckev_b((v16i8) dst0, (v16i8) dst0);
        ST_W2(dst0, 0, 1, dst, dst_stride);
@@ -739,7 +739,7 @@ static void hevc_hz_bi_8t_12w_msa(uint8_t *src0_ptr,
        HEVC_BI_RND_CLIP2(in0, in1, dst0, dst1, 7, dst0, dst1);
        dst2 = __msa_adds_s_h(in2, dst2);
        dst2 = __msa_srari_h(dst2, 7);
        dst2 = CLIP_SH_0_255(dst2);
        CLIP_SH_0_255(dst2);
        PCKEV_B2_SH(dst1, dst0, dst2, dst2, dst0, dst1);
        tmp2 = __msa_copy_s_d((v2i64) dst0, 0);
@@ -888,7 +888,7 @@ static void hevc_hz_bi_8t_24w_msa(uint8_t *src0_ptr,
        HEVC_BI_RND_CLIP2(in0, in1, dst0, dst1, 7, dst0, dst1);
        dst2 = __msa_adds_s_h(dst2, in2);
        dst2 = __msa_srari_h(dst2, 7);
        dst2 = CLIP_SH_0_255(dst2);
        CLIP_SH_0_255(dst2);
        PCKEV_B2_SB(dst1, dst0, dst2, dst2, tmp0, tmp1);
        dst_val0 = __msa_copy_u_d((v2i64) tmp1, 0);
@@ -1726,7 +1726,7 @@ static void hevc_hv_bi_8t_4w_msa(uint8_t *src0_ptr,
        ADDS_SH2_SH(out0, in0, out1, in1, out0, out1);
        ADDS_SH2_SH(out0, const_vec, out1, const_vec, out0, out1);
        SRARI_H2_SH(out0, out1, 7);
        CLIP_SH2_0_255_MAX_SATU(out0, out1);
        CLIP_SH2_0_255(out0, out1);
        out = (v16u8) __msa_pckev_b((v16i8) out1, (v16i8) out0);
        ST_W4(out, 0, 1, 2, 3, dst, dst_stride);
        dst += (4 * dst_stride);
@@ -1854,7 +1854,7 @@ static void hevc_hv_bi_8t_8multx1mult_msa(uint8_t *src0_ptr,
            tmp = __msa_pckev_h((v8i16) dst0_l, (v8i16) dst0_r);
            ADDS_SH2_SH(tmp, in0, tmp, const_vec, tmp, tmp);
            tmp = __msa_srari_h(tmp, 7);
            tmp = CLIP_SH_0_255_MAX_SATU(tmp);
            CLIP_SH_0_255(tmp);
            out = (v16u8) __msa_pckev_b((v16i8) tmp, (v16i8) tmp);
            ST_D1(out, 0, dst_tmp);
            dst_tmp += dst_stride;
@@ -2000,7 +2000,7 @@ static void hevc_hv_bi_8t_12w_msa(uint8_t *src0_ptr,
        tmp = __msa_pckev_h((v8i16) dst0_l, (v8i16) dst0_r);
        ADDS_SH2_SH(tmp, in0, tmp, const_vec, tmp, tmp);
        tmp = __msa_srari_h(tmp, 7);
        tmp = CLIP_SH_0_255_MAX_SATU(tmp);
        CLIP_SH_0_255(tmp);
        out = (v16u8) __msa_pckev_b((v16i8) tmp, (v16i8) tmp);
        ST_D1(out, 0, dst_tmp);
        dst_tmp += dst_stride;
@@ -2088,7 +2088,7 @@ static void hevc_hv_bi_8t_12w_msa(uint8_t *src0_ptr,
        ADDS_SH2_SH(out0, in0, out1, in1, out0, out1);
        ADDS_SH2_SH(out0, const_vec, out1, const_vec, out0, out1);
        SRARI_H2_SH(out0, out1, 7);
        CLIP_SH2_0_255_MAX_SATU(out0, out1);
        CLIP_SH2_0_255(out0, out1);
        out = (v16u8) __msa_pckev_b((v16i8) out1, (v16i8) out0);
        ST_W4(out, 0, 1, 2, 3, dst, dst_stride);
        dst += (4 * dst_stride);
@@ -2215,7 +2215,7 @@ static void hevc_hz_bi_4t_4x2_msa(uint8_t *src0_ptr,
    tmp0 = __msa_adds_s_h(tmp0, in0);
    tmp0 = __msa_srari_h(tmp0, 7);
    tmp0 = CLIP_SH_0_255(tmp0);
    CLIP_SH_0_255(tmp0);
    dst0 = __msa_pckev_b((v16i8) tmp0, (v16i8) tmp0);
    ST_W2(dst0, 0, 1, dst, dst_stride);
@@ -2943,7 +2943,7 @@ static void hevc_vt_bi_4t_4x2_msa(uint8_t *src0_ptr,
    DPADD_SB2_SH(src2110, src4332, filt0, filt1, dst10, dst10);
    dst10 = __msa_adds_s_h(dst10, in0);
    dst10 = __msa_srari_h(dst10, 7);
    dst10 = CLIP_SH_0_255(dst10);
    CLIP_SH_0_255(dst10);
    dst10 = (v8i16) __msa_pckev_b((v16i8) dst10, (v16i8) dst10);
    ST_W2(dst10, 0, 1, dst, dst_stride);
@@ -3843,7 +3843,7 @@ static void hevc_hv_bi_4t_4x2_msa(uint8_t *src0_ptr,
    tmp = __msa_pckev_h((v8i16) dst1, (v8i16) dst0);
    tmp = __msa_adds_s_h(tmp, in0);
    tmp = __msa_srari_h(tmp, 7);
    tmp = CLIP_SH_0_255_MAX_SATU(tmp);
    CLIP_SH_0_255(tmp);
    out = (v16u8) __msa_pckev_b((v16i8) tmp, (v16i8) tmp);
    ST_W2(out, 0, 1, dst, dst_stride);
 }
@@ -3919,7 +3919,7 @@ static void hevc_hv_bi_4t_4x4_msa(uint8_t *src0_ptr,
    PCKEV_H2_SH(dst1, dst0, dst3, dst2, tmp0, tmp1);
    ADDS_SH2_SH(tmp0, in0, tmp1, in1, tmp0, tmp1);
    SRARI_H2_SH(tmp0, tmp1, 7);
    CLIP_SH2_0_255_MAX_SATU(tmp0, tmp1);
    CLIP_SH2_0_255(tmp0, tmp1);
    out = (v16u8) __msa_pckev_b((v16i8) tmp1, (v16i8) tmp0);
    ST_W4(out, 0, 1, 2, 3, dst, dst_stride);
 }
@@ -4032,7 +4032,7 @@ static void hevc_hv_bi_4t_4multx8mult_msa(uint8_t *src0_ptr,
        ADDS_SH4_SH(in0, tmp0, in1, tmp1, in2, tmp2, in3, tmp3, tmp0, tmp1,
                    tmp2, tmp3);
        SRARI_H4_SH(tmp0, tmp1, tmp2, tmp3, 7);
        CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
        CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
        PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
        ST_W8(out0, out1, 0, 1, 2, 3, 0, 1, 2, 3, dst, dst_stride);
        dst += (8 * dst_stride);
@@ -4200,7 +4200,7 @@ static void hevc_hv_bi_4t_6w_msa(uint8_t *src0_ptr,
    ADDS_SH4_SH(in0, tmp0, in1, tmp1, in2, tmp2, in3, tmp3, tmp0, tmp1, tmp2,
                tmp3);
    SRARI_H4_SH(tmp0, tmp1, tmp2, tmp3, 7);
    CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
    CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
    PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
    ST_W8(out0, out1, 0, 1, 2, 3, 0, 1, 2, 3, dst, dst_stride);
@@ -4212,7 +4212,7 @@ static void hevc_hv_bi_4t_6w_msa(uint8_t *src0_ptr,
    ADDS_SH2_SH(in4, const_vec, in5, const_vec, in4, in5);
    ADDS_SH2_SH(in4, tmp4, in5, tmp5, tmp4, tmp5);
    SRARI_H2_SH(tmp4, tmp5, 7);
    CLIP_SH2_0_255_MAX_SATU(tmp4, tmp5);
    CLIP_SH2_0_255(tmp4, tmp5);
    out2 = (v16u8) __msa_pckev_b((v16i8) tmp5, (v16i8) tmp4);
    ST_H8(out2, 0, 1, 2, 3, 4, 5, 6, 7, dst + 4, dst_stride);
 }
@@ -4286,7 +4286,7 @@ static void hevc_hv_bi_4t_8x2_msa(uint8_t *src0_ptr,
    PCKEV_H2_SH(dst0_l, dst0_r, dst1_l, dst1_r, tmp0, tmp1);
    ADDS_SH2_SH(in0, tmp0, in1, tmp1, tmp0, tmp1);
    SRARI_H2_SH(tmp0, tmp1, 7);
    CLIP_SH2_0_255_MAX_SATU(tmp0, tmp1);
    CLIP_SH2_0_255(tmp0, tmp1);
    out = (v16u8) __msa_pckev_b((v16i8) tmp1, (v16i8) tmp0);
    ST_D2(out, 0, 1, dst, dst_stride);
 }
@@ -4380,7 +4380,7 @@ static void hevc_hv_bi_4t_8multx4_msa(uint8_t *src0_ptr,
        ADDS_SH4_SH(in0, tmp0, in1, tmp1, in2, tmp2, in3, tmp3,
                    tmp0, tmp1, tmp2, tmp3);
        SRARI_H4_SH(tmp0, tmp1, tmp2, tmp3, 7);
        CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
        CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
        PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
        ST_D4(out0, out1, 0, 1, 0, 1, dst, dst_stride);
        dst += 8;
@@ -4495,8 +4495,8 @@ static void hevc_hv_bi_4t_8x6_msa(uint8_t *src0_ptr,
    ADDS_SH2_SH(in4, tmp4, in5, tmp5, tmp4, tmp5);
    SRARI_H4_SH(tmp0, tmp1, tmp2, tmp3, 7);
    SRARI_H2_SH(tmp4, tmp5, 7);
    CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
    CLIP_SH2_0_255_MAX_SATU(tmp4, tmp5);
    CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
    CLIP_SH2_0_255(tmp4, tmp5);
    PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
    out2 = (v16u8) __msa_pckev_b((v16i8) tmp5, (v16i8) tmp4);
    ST_D4(out0, out1, 0, 1, 0, 1, dst, dst_stride);
@@ -4610,7 +4610,7 @@ static void hevc_hv_bi_4t_8multx4mult_msa(uint8_t *src0_ptr,
            ADDS_SH4_SH(in0, tmp0, in1, tmp1, in2, tmp2, in3, tmp3,
                        tmp0, tmp1, tmp2, tmp3);
            SRARI_H4_SH(tmp0, tmp1, tmp2, tmp3, 7);
            CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
            CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
            PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
            ST_D4(out0, out1, 0, 1, 0, 1, dst_tmp, dst_stride);
            dst_tmp += (4 * dst_stride);
@@ -4760,7 +4760,7 @@ static void hevc_hv_bi_4t_12w_msa(uint8_t *src0_ptr,
        ADDS_SH4_SH(in0, tmp0, in1, tmp1, in2, tmp2, in3, tmp3,
                    tmp0, tmp1, tmp2, tmp3);
        SRARI_H4_SH(tmp0, tmp1, tmp2, tmp3, 7);
        CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
        CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
        PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
        ST_D4(out0, out1, 0, 1, 0, 1, dst_tmp, dst_stride);
        dst_tmp += (4 * dst_stride);
@@ -4846,7 +4846,7 @@ static void hevc_hv_bi_4t_12w_msa(uint8_t *src0_ptr,
        ADDS_SH4_SH(in0, tmp0, in1, tmp1, in2, tmp2, in3, tmp3,
                    tmp0, tmp1, tmp2, tmp3);
        SRARI_H4_SH(tmp0, tmp1, tmp2, tmp3, 7);
        CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
        CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
        PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
        ST_W8(out0, out1, 0, 1, 2, 3, 0, 1, 2, 3, dst, dst_stride);
        dst += (8 * dst_stride);
--- a/libavcodec/mips/hevc_mc_biw_msa.c
+++ b/libavcodec/mips/hevc_mc_biw_msa.c
@@ -66,7 +66,7 @@ static const uint8_t ff_hevc_mask_arr[16 * 2] __attribute__((aligned(0x40))) = {
    out1_l = __msa_dpadd_s_w(offset, (v8i16) out1_l, (v8i16) wgt);   \
    SRAR_W4_SW(out0_r, out1_r, out0_l, out1_l, rnd);                 \
    PCKEV_H2_SH(out0_l, out0_r, out1_l, out1_r, out0, out1);         \
    CLIP_SH2_0_255_MAX_SATU(out0, out1);                             \
    CLIP_SH2_0_255(out0, out1);                                      \
 }
 #define HEVC_BIW_RND_CLIP4_MAX_SATU(in0, in1, in2, in3, vec0, vec1, vec2,  \
@@ -124,7 +124,7 @@ static void hevc_biwgt_copy_4w_msa(uint8_t *src0_ptr,
        dst0_l = __msa_dpadd_s_w(offset_vec, (v8i16) dst0_l, weight_vec);
        SRAR_W2_SW(dst0_r, dst0_l, rnd_vec);
        dst0 = (v8i16) __msa_pckev_h((v8i16) dst0_l, (v8i16) dst0_r);
        dst0 = CLIP_SH_0_255_MAX_SATU(dst0);
        CLIP_SH_0_255(dst0);
        out0 = (v16u8) __msa_pckev_b((v16i8) dst0, (v16i8) dst0);
        ST_W2(out0, 0, 1, dst, dst_stride);
    } else if (4 == height) {
@@ -1069,8 +1069,8 @@ static void hevc_hz_biwgt_8t_24w_msa(uint8_t *src0_ptr,
        dst2_l = __msa_dpadd_s_w(offset_vec, (v8i16) dst2_l,
                                 (v8i16) weight_vec);
        SRAR_W2_SW(dst2_r, dst2_l, rnd_vec);
        dst2_r = (v4i32) __msa_pckev_h((v8i16) dst2_l, (v8i16) dst2_r);
        out2 = CLIP_SH_0_255(dst2_r);
        out2 = __msa_pckev_h((v8i16) dst2_l, (v8i16) dst2_r);
        CLIP_SH_0_255(out2);
        LD_SB2(src0_ptr, 16, src0, src1);
        src0_ptr += src_stride;
@@ -1100,8 +1100,8 @@ static void hevc_hz_biwgt_8t_24w_msa(uint8_t *src0_ptr,
    dst2_r = __msa_dpadd_s_w(offset_vec, (v8i16) dst2_r, (v8i16) weight_vec);
    dst2_l = __msa_dpadd_s_w(offset_vec, (v8i16) dst2_l, (v8i16) weight_vec);
    SRAR_W2_SW(dst2_r, dst2_l, rnd_vec);
    dst2_r = (v4i32) __msa_pckev_h((v8i16) dst2_l, (v8i16) dst2_r);
    out2 = CLIP_SH_0_255(dst2_r);
    out2 = __msa_pckev_h((v8i16) dst2_l, (v8i16) dst2_r);
    CLIP_SH_0_255(out2);
    PCKEV_B2_SH(out1, out0, out2, out2, out0, out2);
    dst_val0 = __msa_copy_u_d((v2i64) out2, 0);
    ST_SH(out0, dst);
@@ -1674,8 +1674,8 @@ static void hevc_vt_biwgt_8t_12w_msa(uint8_t *src0_ptr,
        dst2_l = __msa_dpadd_s_w(offset_vec, (v8i16) dst2_l,
                                 (v8i16) weight_vec);
        SRAR_W2_SW(dst2_r, dst2_l, rnd_vec);
        dst2_r = (v4i32) __msa_pckev_h((v8i16) dst2_l, (v8i16) dst2_r);
        out2 = CLIP_SH_0_255(dst2_r);
        out2 = __msa_pckev_h((v8i16) dst2_l, (v8i16) dst2_r);
        CLIP_SH_0_255(out2);
        PCKEV_B2_SH(out1, out0, out2, out2, out0, out2);
        ST_D2(out0, 0, 1, dst, dst_stride);
        ST_W2(out2, 0, 1, dst + 8, dst_stride);
@@ -2048,7 +2048,7 @@ static void hevc_hv_biwgt_8t_4w_msa(uint8_t *src0_ptr,
        dst2 = __msa_dpadd_s_w(offset_vec, tmp2, weight_vec);
        dst3 = __msa_dpadd_s_w(offset_vec, tmp3, weight_vec);
        SRAR_W4_SW(dst0, dst1, dst2, dst3, rnd_vec);
        CLIP_SW4_0_255_MAX_SATU(dst0, dst1, dst2, dst3);
        CLIP_SW4_0_255(dst0, dst1, dst2, dst3);
        PCKEV_H2_SH(dst1, dst0, dst3, dst2, tmp0, tmp1);
        out = (v16u8) __msa_pckev_b((v16i8) tmp1, (v16i8) tmp0);
        ST_W4(out, 0, 1, 2, 3, dst, dst_stride);
@@ -2226,7 +2226,7 @@ static void hevc_hv_biwgt_8t_8multx2mult_msa(uint8_t *src0_ptr,
            dst1_r = __msa_dpadd_s_w(offset_vec, tmp2, weight_vec);
            dst1_l = __msa_dpadd_s_w(offset_vec, tmp3, weight_vec);
            SRAR_W4_SW(dst0_l, dst0_r, dst1_l, dst1_r, rnd_vec);
            CLIP_SW4_0_255_MAX_SATU(dst0_l, dst0_r, dst1_l, dst1_r);
            CLIP_SW4_0_255(dst0_l, dst0_r, dst1_l, dst1_r);
            PCKEV_H2_SH(dst0_l, dst0_r, dst1_l, dst1_r, tmp0, tmp1);
            out = (v16u8) __msa_pckev_b((v16i8) tmp1, (v16i8) tmp0);
            ST_D2(out, 0, 1, dst_tmp, dst_stride);
@@ -2412,7 +2412,7 @@ static void hevc_hv_biwgt_8t_12w_msa(uint8_t *src0_ptr,
        dst2 = __msa_dpadd_s_w(offset_vec, tmp2, weight_vec);
        dst3 = __msa_dpadd_s_w(offset_vec, tmp3, weight_vec);
        SRAR_W4_SW(dst1, dst0, dst3, dst2, rnd_vec);
        CLIP_SW4_0_255_MAX_SATU(dst1, dst0, dst3, dst2);
        CLIP_SW4_0_255(dst1, dst0, dst3, dst2);
        PCKEV_H2_SH(dst1, dst0, dst3, dst2, tmp0, tmp1);
        out = (v16u8) __msa_pckev_b((v16i8) tmp1, (v16i8) tmp0);
        ST_D2(out, 0, 1, dst_tmp, dst_stride);
@@ -2503,7 +2503,7 @@ static void hevc_hv_biwgt_8t_12w_msa(uint8_t *src0_ptr,
        dst2 = __msa_dpadd_s_w(offset_vec, tmp2, weight_vec);
        dst3 = __msa_dpadd_s_w(offset_vec, tmp3, weight_vec);
        SRAR_W4_SW(dst0, dst1, dst2, dst3, rnd_vec);
        CLIP_SW4_0_255_MAX_SATU(dst0, dst1, dst2, dst3);
        CLIP_SW4_0_255(dst0, dst1, dst2, dst3);
        PCKEV_H2_SH(dst1, dst0, dst3, dst2, tmp0, tmp1);
        out = (v16u8) __msa_pckev_b((v16i8) tmp1, (v16i8) tmp0);
        ST_W4(out, 0, 1, 2, 3, dst, dst_stride);
@@ -2683,8 +2683,8 @@ static void hevc_hz_biwgt_4t_4x2_msa(uint8_t *src0_ptr,
    dst0_r = __msa_dpadd_s_w(offset_vec, (v8i16) dst0_r, (v8i16) weight_vec);
    dst0_l = __msa_dpadd_s_w(offset_vec, (v8i16) dst0_l, (v8i16) weight_vec);
    SRAR_W2_SW(dst0_r, dst0_l, rnd_vec);
    dst0_r = (v4i32) __msa_pckev_h((v8i16) dst0_l, (v8i16) dst0_r);
    out0 = CLIP_SH_0_255(dst0_r);
    out0 = __msa_pckev_h((v8i16) dst0_l, (v8i16) dst0_r);
    CLIP_SH_0_255(out0);
    out0 = (v8i16) __msa_pckev_b((v16i8) out0, (v16i8) out0);
    ST_W2(out0, 0, 1, dst, dst_stride);
 }
@@ -3554,8 +3554,8 @@ static void hevc_vt_biwgt_4t_4x2_msa(uint8_t *src0_ptr,
    dst10_r = __msa_dpadd_s_w(offset_vec, (v8i16) dst10_r, (v8i16) weight_vec);
    dst10_l = __msa_dpadd_s_w(offset_vec, (v8i16) dst10_l, (v8i16) weight_vec);
    SRAR_W2_SW(dst10_r, dst10_l, rnd_vec);
    dst10_r = (v4i32) __msa_pckev_h((v8i16) dst10_l, (v8i16) dst10_r);
    out = CLIP_SH_0_255(dst10_r);
    out = __msa_pckev_h((v8i16) dst10_l, (v8i16) dst10_r);
    CLIP_SH_0_255(out);
    out = (v8i16) __msa_pckev_b((v16i8) out, (v16i8) out);
    ST_W2(out, 0, 1, dst, dst_stride);
 }
@@ -4575,7 +4575,7 @@ static void hevc_hv_biwgt_4t_4x2_msa(uint8_t *src0_ptr,
    dst1 = __msa_dpadd_s_w(offset_vec, tmp1, weight_vec);
    SRAR_W2_SW(dst0, dst1, rnd_vec);
    tmp = __msa_pckev_h((v8i16) dst1, (v8i16) dst0);
    tmp = CLIP_SH_0_255_MAX_SATU(tmp);
    CLIP_SH_0_255(tmp);
    out = (v16u8) __msa_pckev_b((v16i8) tmp, (v16i8) tmp);
    ST_W2(out, 0, 1, dst, dst_stride);
 }
@@ -4672,7 +4672,7 @@ static void hevc_hv_biwgt_4t_4x4_msa(uint8_t *src0_ptr,
    dst3 = __msa_dpadd_s_w(offset_vec, tmp3, weight_vec);
    SRAR_W4_SW(dst0, dst1, dst2, dst3, rnd_vec);
    PCKEV_H2_SH(dst1, dst0, dst3, dst2, tmp0, tmp1);
    CLIP_SH2_0_255_MAX_SATU(tmp0, tmp1);
    CLIP_SH2_0_255(tmp0, tmp1);
    out = (v16u8) __msa_pckev_b((v16i8) tmp1, (v16i8) tmp0);
    ST_W4(out, 0, 1, 2, 3, dst, dst_stride);
 }
@@ -4810,7 +4810,7 @@ static void hevc_hv_biwgt_4t_4multx8mult_msa(uint8_t *src0_ptr,
        SRAR_W4_SW(dst4, dst5, dst6, dst7, rnd_vec);
        PCKEV_H4_SH(dst1, dst0, dst3, dst2, dst5, dst4, dst7, dst6, tmp0, tmp1,
                    tmp2, tmp3);
        CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
        CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
        PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
        ST_W8(out0, out1, 0, 1, 2, 3, 0, 1, 2, 3, dst, dst_stride);
        dst += (8 * dst_stride);
@@ -5008,7 +5008,7 @@ static void hevc_hv_biwgt_4t_6w_msa(uint8_t *src0_ptr,
    SRAR_W4_SW(dst4, dst5, dst6, dst7, rnd_vec);
    PCKEV_H4_SH(dst1, dst0, dst3, dst2, dst5, dst4, dst7, dst6, tmp0, tmp1,
                tmp2, tmp3);
    CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
    CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
    PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
    ST_W8(out0, out1, 0, 1, 2, 3, 0, 1, 2, 3, dst, dst_stride);
@@ -5030,7 +5030,7 @@ static void hevc_hv_biwgt_4t_6w_msa(uint8_t *src0_ptr,
    SRAR_W4_SW(dst0, dst1, dst2, dst3, rnd_vec);
    PCKEV_H2_SH(dst1, dst0, dst3, dst2, tmp4, tmp5);
    CLIP_SH2_0_255_MAX_SATU(tmp4, tmp5);
    CLIP_SH2_0_255(tmp4, tmp5);
    out2 = (v16u8) __msa_pckev_b((v16i8) tmp5, (v16i8) tmp4);
    ST_H8(out2, 0, 1, 2, 3, 4, 5, 6, 7, dst + 4, dst_stride);
 }
@@ -5126,7 +5126,7 @@ static void hevc_hv_biwgt_4t_8x2_msa(uint8_t *src0_ptr,
    dst1_l = __msa_dpadd_s_w(offset_vec, tmp3, weight_vec);
    SRAR_W4_SW(dst0_r, dst0_l, dst1_r, dst1_l, rnd_vec);
    PCKEV_H2_SH(dst0_l, dst0_r, dst1_l, dst1_r, tmp0, tmp1);
    CLIP_SH2_0_255_MAX_SATU(tmp0, tmp1);
    CLIP_SH2_0_255(tmp0, tmp1);
    out = (v16u8) __msa_pckev_b((v16i8) tmp1, (v16i8) tmp0);
    ST_D2(out, 0, 1, dst, dst_stride);
 }
@@ -5248,7 +5248,7 @@ static void hevc_hv_biwgt_4t_8multx4_msa(uint8_t *src0_ptr,
        SRAR_W4_SW(dst4, dst5, dst6, dst7, rnd_vec);
        PCKEV_H4_SH(dst1, dst0, dst3, dst2, dst5, dst4, dst7, dst6,
                    tmp0, tmp1, tmp2, tmp3);
        CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
        CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
        PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
        ST_D4(out0, out1, 0, 1, 0, 1, dst, dst_stride);
        dst += 8;
@@ -5387,7 +5387,7 @@ static void hevc_hv_biwgt_4t_8x6_msa(uint8_t *src0_ptr,
    SRAR_W4_SW(dst4, dst5, dst6, dst7, rnd_vec);
    PCKEV_H4_SH(dst1, dst0, dst3, dst2, dst5, dst4, dst7, dst6,
                tmp0, tmp1, tmp2, tmp3);
    CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
    CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
    PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
    PCKEV_H2_SW(dst4_l, dst4_r, dst5_l, dst5_r, dst0, dst1);
@@ -5399,7 +5399,7 @@ static void hevc_hv_biwgt_4t_8x6_msa(uint8_t *src0_ptr,
    dst3 = __msa_dpadd_s_w(offset_vec, tmp3, weight_vec);
    SRAR_W4_SW(dst0, dst1, dst2, dst3, rnd_vec);
    PCKEV_H2_SH(dst1, dst0, dst3, dst2, tmp4, tmp5);
    CLIP_SH2_0_255_MAX_SATU(tmp4, tmp5);
    CLIP_SH2_0_255(tmp4, tmp5);
    out2 = (v16u8) __msa_pckev_b((v16i8) tmp5, (v16i8) tmp4);
    ST_D4(out0, out1, 0, 1, 0, 1, dst, dst_stride);
    ST_D2(out2, 0, 1, dst + 4 * dst_stride, dst_stride);
@@ -5537,7 +5537,7 @@ static void hevc_hv_biwgt_4t_8multx4mult_msa(uint8_t *src0_ptr,
            SRAR_W4_SW(dst4, dst5, dst6, dst7, rnd_vec);
            PCKEV_H4_SH(dst1, dst0, dst3, dst2, dst5, dst4, dst7, dst6,
                        tmp0, tmp1, tmp2, tmp3);
            CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
            CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
            PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
            ST_D4(out0, out1, 0, 1, 0, 1, dst_tmp, dst_stride);
            dst_tmp += (4 * dst_stride);
@@ -5724,7 +5724,7 @@ static void hevc_hv_biwgt_4t_12w_msa(uint8_t *src0_ptr,
        SRAR_W4_SW(dst4, dst5, dst6, dst7, rnd_vec);
        PCKEV_H4_SH(dst1, dst0, dst3, dst2, dst5, dst4, dst7, dst6,
                    tmp0, tmp1, tmp2, tmp3);
        CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
        CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
        PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
        ST_D4(out0, out1, 0, 1, 0, 1, dst_tmp, dst_stride);
        dst_tmp += (4 * dst_stride);
@@ -5820,7 +5820,7 @@ static void hevc_hv_biwgt_4t_12w_msa(uint8_t *src0_ptr,
        SRAR_W4_SW(dst4, dst5, dst6, dst7, rnd_vec);
        PCKEV_H4_SH(dst1, dst0, dst3, dst2, dst5, dst4, dst7, dst6,
                    tmp0, tmp1, tmp2, tmp3);
        CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
        CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
        PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
        ST_W8(out0, out1, 0, 1, 2, 3, 0, 1, 2, 3, dst, dst_stride);
        dst += (8 * dst_stride);
--- a/libavcodec/mips/hevc_mc_uniw_msa.c
+++ b/libavcodec/mips/hevc_mc_uniw_msa.c
@@ -41,7 +41,7 @@ static const uint8_t ff_hevc_mask_arr[16 * 2] __attribute__((aligned(0x40))) = {
    SRAR_W4_SW(in0_r_m, in1_r_m, in0_l_m, in1_l_m, rnd_w);                    \
    PCKEV_H2_SH(in0_l_m, in0_r_m, in1_l_m, in1_r_m, out0_h, out1_h);          \
    ADDS_SH2_SH(out0_h, offset_h, out1_h, offset_h, out0_h, out1_h);          \
    CLIP_SH2_0_255_MAX_SATU(out0_h, out1_h);                                  \
    CLIP_SH2_0_255(out0_h, out1_h);                                           \
 }
 #define HEVC_UNIW_RND_CLIP4_MAX_SATU_H(in0_h, in1_h, in2_h, in3_h, wgt_w,  \
@@ -88,7 +88,7 @@ static void hevc_uniwgt_copy_4w_msa(uint8_t *src,
        SRAR_W2_SW(dst0_r, dst0_l, rnd_vec);
        dst0 = __msa_pckev_h((v8i16) dst0_l, (v8i16) dst0_r);
        dst0 += offset_vec;
        dst0 = CLIP_SH_0_255_MAX_SATU(dst0);
        CLIP_SH_0_255(dst0);
        out0 = (v16u8) __msa_pckev_b((v16i8) dst0, (v16i8) dst0);
        ST_W2(out0, 0, 1, dst, dst_stride);
    } else if (4 == height) {
@@ -1863,7 +1863,7 @@ static void hevc_hv_uniwgt_8t_4w_msa(uint8_t *src,
        SRAR_W4_SW(dst0_r, dst1_r, dst2_r, dst3_r, rnd_vec);
        ADD2(dst0_r, offset_vec, dst1_r, offset_vec, dst0_r, dst1_r);
        ADD2(dst2_r, offset_vec, dst3_r, offset_vec, dst2_r, dst3_r);
        CLIP_SW4_0_255_MAX_SATU(dst0_r, dst1_r, dst2_r, dst3_r);
        CLIP_SW4_0_255(dst0_r, dst1_r, dst2_r, dst3_r);
        PCKEV_H2_SW(dst1_r, dst0_r, dst3_r, dst2_r, dst0_r, dst1_r);
        out = (v16u8) __msa_pckev_b((v16i8) dst1_r, (v16i8) dst0_r);
        ST_W4(out, 0, 1, 2, 3, dst, dst_stride);
@@ -2014,7 +2014,7 @@ static void hevc_hv_uniwgt_8t_8multx2mult_msa(uint8_t *src,
            SRAR_W4_SW(dst0_r, dst1_r, dst0_l, dst1_l, rnd_vec);
            ADD2(dst0_r, offset_vec, dst0_l, offset_vec, dst0_r, dst0_l);
            ADD2(dst1_r, offset_vec, dst1_l, offset_vec, dst1_r, dst1_l);
            CLIP_SW4_0_255_MAX_SATU(dst0_r, dst1_r, dst0_l, dst1_l);
            CLIP_SW4_0_255(dst0_r, dst1_r, dst0_l, dst1_l);
            PCKEV_H2_SW(dst0_l, dst0_r, dst1_l, dst1_r, dst0_r, dst1_r);
            dst0_r = (v4i32) __msa_pckev_b((v16i8) dst1_r, (v16i8) dst0_r);
@@ -2165,7 +2165,7 @@ static void hevc_hv_uniwgt_8t_12w_msa(uint8_t *src,
        MUL2(dst0_r, weight_vec, dst0_l, weight_vec, dst0_r, dst0_l);
        SRAR_W2_SW(dst0_r, dst0_l, rnd_vec);
        ADD2(dst0_r, offset_vec, dst0_l, offset_vec, dst0_r, dst0_l);
        CLIP_SW2_0_255_MAX_SATU(dst0_r, dst0_l);
        CLIP_SW2_0_255(dst0_r, dst0_l);
        dst0_r = (v4i32) __msa_pckev_h((v8i16) dst0_l, (v8i16) dst0_r);
        out = (v16u8) __msa_pckev_b((v16i8) dst0_r, (v16i8) dst0_r);
        ST_D1(out, 0, dst_tmp);
@@ -2246,7 +2246,7 @@ static void hevc_hv_uniwgt_8t_12w_msa(uint8_t *src,
        SRAR_W4_SW(dst0_r, dst1_r, dst2_r, dst3_r, rnd_vec);
        ADD2(dst0_r, offset_vec, dst1_r, offset_vec, dst0_r, dst1_r);
        ADD2(dst2_r, offset_vec, dst3_r, offset_vec, dst2_r, dst3_r);
        CLIP_SW4_0_255_MAX_SATU(dst0_r, dst1_r, dst2_r, dst3_r);
        CLIP_SW4_0_255(dst0_r, dst1_r, dst2_r, dst3_r);
        PCKEV_H2_SW(dst1_r, dst0_r, dst3_r, dst2_r, dst0_r, dst1_r);
        out = (v16u8) __msa_pckev_b((v16i8) dst1_r, (v16i8) dst0_r);
        ST_W4(out, 0, 1, 2, 3, dst, dst_stride);
@@ -2394,7 +2394,7 @@ static void hevc_hz_uniwgt_4t_4x2_msa(uint8_t *src,
    SRAR_W2_SW(dst0_r, dst0_l, rnd_vec);
    dst0 = __msa_pckev_h((v8i16) dst0_l, (v8i16) dst0_r);
    dst0 = __msa_adds_s_h(dst0, offset_vec);
    dst0 = CLIP_SH_0_255_MAX_SATU(dst0);
    CLIP_SH_0_255(dst0);
    out = (v16u8) __msa_pckev_b((v16i8) dst0, (v16i8) dst0);
    ST_W2(out, 0, 1, dst, dst_stride);
    dst += (4 * dst_stride);
@@ -3295,7 +3295,7 @@ static void hevc_vt_uniwgt_4t_4x2_msa(uint8_t *src,
    SRAR_W2_SW(dst0_r, dst0_l, rnd_vec);
    dst0 = __msa_pckev_h((v8i16) dst0_l, (v8i16) dst0_r);
    dst0 = __msa_adds_s_h(dst0, offset_vec);
    dst0 = CLIP_SH_0_255_MAX_SATU(dst0);
    CLIP_SH_0_255(dst0);
    out = (v16u8) __msa_pckev_b((v16i8) dst0, (v16i8) dst0);
    ST_W2(out, 0, 1, dst, dst_stride);
 }
@@ -4247,7 +4247,7 @@ static void hevc_hv_uniwgt_4t_4x2_msa(uint8_t *src,
    SRAR_W2_SW(dst0, dst1, rnd_vec);
    tmp = __msa_pckev_h((v8i16) dst1, (v8i16) dst0);
    tmp += offset_vec;
    tmp = CLIP_SH_0_255_MAX_SATU(tmp);
    CLIP_SH_0_255(tmp);
    out = (v16u8) __msa_pckev_b((v16i8) tmp, (v16i8) tmp);
    ST_W2(out, 0, 1, dst, dst_stride);
 }
@@ -4316,7 +4316,7 @@ static void hevc_hv_uniwgt_4t_4x4_msa(uint8_t *src,
    SRAR_W4_SW(dst0, dst1, dst2, dst3, rnd_vec);
    PCKEV_H2_SH(dst1, dst0, dst3, dst2, tmp0, tmp1);
    ADD2(tmp0, offset_vec, tmp1, offset_vec, tmp0, tmp1);
    CLIP_SH2_0_255_MAX_SATU(tmp0, tmp1);
    CLIP_SH2_0_255(tmp0, tmp1);
    out = (v16u8) __msa_pckev_b((v16i8) tmp1, (v16i8) tmp0);
    ST_W4(out, 0, 1, 2, 3, dst, dst_stride);
 }
@@ -4417,7 +4417,7 @@ static void hevc_hv_uniwgt_4t_4multx8mult_msa(uint8_t *src,
                    tmp2, tmp3);
        ADD2(tmp0, offset_vec, tmp1, offset_vec, tmp0, tmp1);
        ADD2(tmp2, offset_vec, tmp3, offset_vec, tmp2, tmp3);
        CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
        CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
        PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
        ST_W8(out0, out1, 0, 1, 2, 3, 0, 1, 2, 3, dst, dst_stride);
        dst += (8 * dst_stride);
@@ -4574,8 +4574,8 @@ static void hevc_hv_uniwgt_4t_6w_msa(uint8_t *src,
    ADD2(tmp0, offset_vec, tmp1, offset_vec, tmp0, tmp1);
    ADD2(tmp2, offset_vec, tmp3, offset_vec, tmp2, tmp3);
    ADD2(tmp4, offset_vec, tmp5, offset_vec, tmp4, tmp5);
    CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
    CLIP_SH2_0_255_MAX_SATU(tmp4, tmp5);
    CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
    CLIP_SH2_0_255(tmp4, tmp5);
    PCKEV_B3_UB(tmp1, tmp0, tmp3, tmp2, tmp5, tmp4, out0, out1, out2);
    ST_W8(out0, out1, 0, 1, 2, 3, 0, 1, 2, 3, dst, dst_stride);
    ST_H8(out2, 0, 1, 2, 3, 4, 5, 6, 7, dst + 4, dst_stride);
@@ -4652,7 +4652,7 @@ static void hevc_hv_uniwgt_4t_8x2_msa(uint8_t *src,
    SRAR_W4_SW(dst0_r, dst0_l, dst1_r, dst1_l, rnd_vec);
    PCKEV_H2_SH(dst0_l, dst0_r, dst1_l, dst1_r, tmp0, tmp1);
    ADD2(tmp0, offset_vec, tmp1, offset_vec, tmp0, tmp1);
    CLIP_SH2_0_255_MAX_SATU(tmp0, tmp1);
    CLIP_SH2_0_255(tmp0, tmp1);
    out = (v16u8) __msa_pckev_b((v16i8) tmp1, (v16i8) tmp0);
    ST_D2(out, 0, 1, dst, dst_stride);
 }
@@ -4745,7 +4745,7 @@ static void hevc_hv_uniwgt_4t_8multx4_msa(uint8_t *src,
                    dst3_r, tmp0, tmp1, tmp2, tmp3);
        ADD2(tmp0, offset_vec, tmp1, offset_vec, tmp0, tmp1);
        ADD2(tmp2, offset_vec, tmp3, offset_vec, tmp2, tmp3);
        CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
        CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
        PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
        ST_D4(out0, out1, 0, 1, 0, 1, dst, dst_stride);
        dst += 8;
@@ -4861,8 +4861,8 @@ static void hevc_hv_uniwgt_4t_8x6_msa(uint8_t *src,
    ADD2(tmp0, offset_vec, tmp1, offset_vec, tmp0, tmp1);
    ADD2(tmp2, offset_vec, tmp3, offset_vec, tmp2, tmp3);
    ADD2(tmp4, offset_vec, tmp5, offset_vec, tmp4, tmp5);
    CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
    CLIP_SH2_0_255_MAX_SATU(tmp4, tmp5);
    CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
    CLIP_SH2_0_255(tmp4, tmp5);
    PCKEV_B3_UB(tmp1, tmp0, tmp3, tmp2, tmp5, tmp4, out0, out1, out2);
    ST_D4(out0, out1, 0, 1, 0, 1, dst, dst_stride);
    ST_D2(out2, 0, 1, dst + 4 * dst_stride, dst_stride);
@@ -4973,7 +4973,7 @@ static void hevc_hv_uniwgt_4t_8multx4mult_msa(uint8_t *src,
                        dst3_r, tmp0, tmp1, tmp2, tmp3);
            ADD2(tmp0, offset_vec, tmp1, offset_vec, tmp0, tmp1);
            ADD2(tmp2, offset_vec, tmp3, offset_vec, tmp2, tmp3);
            CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
            CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
            PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
            ST_D4(out0, out1, 0, 1, 0, 1, dst_tmp, dst_stride);
            dst_tmp += (4 * dst_stride);
@@ -5120,7 +5120,7 @@ static void hevc_hv_uniwgt_4t_12w_msa(uint8_t *src,
                    dst3_r, tmp0, tmp1, tmp2, tmp3);
        ADD2(tmp0, offset_vec, tmp1, offset_vec, tmp0, tmp1);
        ADD2(tmp2, offset_vec, tmp3, offset_vec, tmp2, tmp3);
        CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
        CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
        PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
        ST_D4(out0, out1, 0, 1, 0, 1, dst_tmp, dst_stride);
        dst_tmp += (4 * dst_stride);
@@ -5187,7 +5187,7 @@ static void hevc_hv_uniwgt_4t_12w_msa(uint8_t *src,
                    tmp2, tmp3);
        ADD2(tmp0, offset_vec, tmp1, offset_vec, tmp0, tmp1);
        ADD2(tmp2, offset_vec, tmp3, offset_vec, tmp2, tmp3);
        CLIP_SH4_0_255_MAX_SATU(tmp0, tmp1, tmp2, tmp3);
        CLIP_SH4_0_255(tmp0, tmp1, tmp2, tmp3);
        PCKEV_B2_UB(tmp1, tmp0, tmp3, tmp2, out0, out1);
        ST_W8(out0, out1, 0, 1, 2, 3, 0, 1, 2, 3, dst, dst_stride);
        dst += (8 * dst_stride);
--- a/libavcodec/mips/hevcpred_msa.c
+++ b/libavcodec/mips/hevcpred_msa.c
@@ -83,7 +83,7 @@ static void hevc_intra_pred_vert_4x4_msa(const uint8_t *src_top,
        vec2 -= vec0;
        vec2 >>= 1;
        vec2 += vec1;
        vec2 = CLIP_SH_0_255(vec2);
        CLIP_SH_0_255(vec2);
        for (col = 0; col < 4; col++) {
            dst[stride * col] = (uint8_t) vec2[col];
@@ -122,7 +122,7 @@ static void hevc_intra_pred_vert_8x8_msa(const uint8_t *src_top,
        vec2 -= vec0;
        vec2 >>= 1;
        vec2 += vec1;
        vec2 = CLIP_SH_0_255(vec2);
        CLIP_SH_0_255(vec2);
        val0 = vec2[0];
        val1 = vec2[1];
@@ -214,7 +214,7 @@ static void hevc_intra_pred_horiz_4x4_msa(const uint8_t *src_top,
        src0_r -= src_top_val;
        src0_r >>= 1;
        src0_r += src_left_val;
        src0_r = CLIP_SH_0_255(src0_r);
        CLIP_SH_0_255(src0_r);
        src0 = __msa_pckev_b((v16i8) src0_r, (v16i8) src0_r);
        val0 = __msa_copy_s_w((v4i32) src0, 0);
        SW(val0, dst);
@@ -254,7 +254,7 @@ static void hevc_intra_pred_horiz_8x8_msa(const uint8_t *src_top,
        src0_r -= src_top_val;
        src0_r >>= 1;
        src0_r += src_left_val;
        src0_r = CLIP_SH_0_255(src0_r);
        CLIP_SH_0_255(src0_r);
        src0 = __msa_pckev_b((v16i8) src0_r, (v16i8) src0_r);
        val0 = __msa_copy_s_d((v2i64) src0, 0);
        SD(val0, dst);
--- a/libavcodec/mips/idctdsp_msa.c
+++ b/libavcodec/mips/idctdsp_msa.c
@@ -28,8 +28,7 @@ static void put_pixels_clamped_msa(const int16_t *block, uint8_t *pixels,
    v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
    LD_SH8(block, 8, in0, in1, in2, in3, in4, in5, in6, in7);
    CLIP_SH4_0_255(in0, in1, in2, in3);
    CLIP_SH4_0_255(in4, in5, in6, in7);
    CLIP_SH8_0_255(in0, in1, in2, in3, in4, in5, in6, in7);
    PCKEV_B4_SH(in0, in0, in1, in1, in2, in2, in3, in3, in0, in1, in2, in3);
    PCKEV_B4_SH(in4, in4, in5, in5, in6, in6, in7, in7, in4, in5, in6, in7);
@@ -63,8 +62,7 @@ static void put_signed_pixels_clamped_msa(const int16_t *block, uint8_t *pixels,
    in6 += 128;
    in7 += 128;
    CLIP_SH4_0_255(in0, in1, in2, in3);
    CLIP_SH4_0_255(in4, in5, in6, in7);
    CLIP_SH8_0_255(in0, in1, in2, in3, in4, in5, in6, in7);
    PCKEV_B4_SH(in0, in0, in1, in1, in2, in2, in3, in3, in0, in1, in2, in3);
    PCKEV_B4_SH(in4, in4, in5, in5, in6, in6, in7, in7, in4, in5, in6, in7);
@@ -109,8 +107,7 @@ static void add_pixels_clamped_msa(const int16_t *block, uint8_t *pixels,
    in6 += (v8i16) pix6;
    in7 += (v8i16) pix7;
    CLIP_SH4_0_255(in0, in1, in2, in3);
    CLIP_SH4_0_255(in4, in5, in6, in7);
    CLIP_SH8_0_255(in0, in1, in2, in3, in4, in5, in6, in7);
    PCKEV_B4_SH(in0, in0, in1, in1, in2, in2, in3, in3, in0, in1, in2, in3);
    PCKEV_B4_SH(in4, in4, in5, in5, in6, in6, in7, in7, in4, in5, in6, in7);
--- a/libavcodec/mips/qpeldsp_msa.c
+++ b/libavcodec/mips/qpeldsp_msa.c
@@ -96,7 +96,7 @@
    DPADD_UB2_UH(sum2_r, sum1_r, coef2, coef1, sum0_r, sum3_r);         \
    res0_r = (v8i16) (sum0_r - sum3_r);                                 \
    res0_r = __msa_srari_h(res0_r, 5);                                  \
    res0_r = CLIP_SH_0_255(res0_r);                                     \
    CLIP_SH_0_255(res0_r);                                              \
    out = (v16u8) __msa_pckev_b((v16i8) res0_r, (v16i8) res0_r);        \
                                                                        \
    out;                                                                \
@@ -118,7 +118,7 @@
    res0_r = (v8i16) (sum0_r - sum3_r);                                   \
    res0_r += 15;                                                         \
    res0_r >>= 5;                                                         \
    res0_r = CLIP_SH_0_255(res0_r);                                       \
    CLIP_SH_0_255(res0_r);                                                \
    out = (v16u8) __msa_pckev_b((v16i8) res0_r, (v16i8) res0_r);          \
                                                                          \
    out;                                                                  \
--- a/libavcodec/mips/simple_idct_msa.c
+++ b/libavcodec/mips/simple_idct_msa.c
@@ -336,35 +336,26 @@ static void simple_idct_put_msa(uint8_t *dst, int32_t dst_stride,
    SRA_4V(temp2_r, temp2_l, temp3_r, temp3_l, 20);
    SRA_4V(a3_r, a3_l, a2_r, a2_l, 20);
    SRA_4V(a1_r, a1_l, a0_r, a0_l, 20);
    PCKEV_H4_SW(temp0_l, temp0_r, temp1_l, temp1_r, temp2_l, temp2_r,
                temp3_l, temp3_r, temp0_r, temp1_r, temp2_r, temp3_r);
    PCKEV_H4_SW(a0_l, a0_r, a1_l, a1_r, a2_l, a2_r, a3_l, a3_r,
                a0_r, a1_r, a2_r, a3_r);
    temp0_r = (v4i32) CLIP_SH_0_255(temp0_r);
    temp1_r = (v4i32) CLIP_SH_0_255(temp1_r);
    temp2_r = (v4i32) CLIP_SH_0_255(temp2_r);
    temp3_r = (v4i32) CLIP_SH_0_255(temp3_r);
    PCKEV_B4_SW(temp0_r, temp0_r, temp1_r, temp1_r,
                temp2_r, temp2_r, temp3_r, temp3_r,
                temp0_r, temp1_r, temp2_r, temp3_r);
    tmp0 = __msa_copy_u_d((v2i64) temp0_r, 1);
    tmp1 = __msa_copy_u_d((v2i64) temp1_r, 1);
    tmp2 = __msa_copy_u_d((v2i64) temp2_r, 1);
    tmp3 = __msa_copy_u_d((v2i64) temp3_r, 1);
    SD4(tmp0, tmp1, tmp2, tmp3, dst, dst_stride);
    dst += 4 * dst_stride;
    a0_r = (v4i32) CLIP_SH_0_255(a0_r);
    a1_r = (v4i32) CLIP_SH_0_255(a1_r);
    a2_r = (v4i32) CLIP_SH_0_255(a2_r);
    a3_r = (v4i32) CLIP_SH_0_255(a3_r);
    PCKEV_B4_SW(a0_r, a0_r, a1_r, a1_r,
                a2_r, a2_r, a3_r, a3_r, a0_r, a1_r, a2_r, a3_r);
    tmp3 = __msa_copy_u_d((v2i64) a0_r, 1);
    tmp2 = __msa_copy_u_d((v2i64) a1_r, 1);
    tmp1 = __msa_copy_u_d((v2i64) a2_r, 1);
    tmp0 = __msa_copy_u_d((v2i64) a3_r, 1);
    PCKEV_H4_SH(temp0_l, temp0_r, temp1_l, temp1_r, temp2_l, temp2_r,
                temp3_l, temp3_r, in0, in1, in2, in3);
    PCKEV_H4_SH(a0_l, a0_r, a1_l, a1_r, a2_l, a2_r, a3_l, a3_r,
                in4, in5, in6, in7);
    CLIP_SH4_0_255(in0, in1, in2, in3);
    PCKEV_B4_SH(in0, in0, in1, in1, in2, in2, in3, in3,
                in0, in1, in2, in3);
    tmp0 = __msa_copy_u_d((v2i64) in0, 1);
    tmp1 = __msa_copy_u_d((v2i64) in1, 1);
    tmp2 = __msa_copy_u_d((v2i64) in2, 1);
    tmp3 = __msa_copy_u_d((v2i64) in3, 1);
    SD4(tmp0, tmp1, tmp2, tmp3, dst, dst_stride);
    dst += 4 * dst_stride;
    CLIP_SH4_0_255(in4, in5, in6, in7);
    PCKEV_B4_SH(in4, in4, in5, in5, in6, in6, in7, in7,
                in4, in5, in6, in7);
    tmp3 = __msa_copy_u_d((v2i64) in4, 1);
    tmp2 = __msa_copy_u_d((v2i64) in5, 1);
    tmp1 = __msa_copy_u_d((v2i64) in6, 1);
    tmp0 = __msa_copy_u_d((v2i64) in7, 1);
    SD4(tmp0, tmp1, tmp2, tmp3, dst + 4 * dst_stride, dst_stride);
 }
 static void simple_idct_add_msa(uint8_t *dst, int32_t dst_stride,
@@ -516,21 +507,17 @@ static void simple_idct_add_msa(uint8_t *dst, int32_t dst_stride,
                temp3_l, temp3_r, temp0_r, temp1_r, temp2_r, temp3_r);
    ILVR_B4_SW(zero, in0, zero, in1, zero, in2, zero, in3,
               temp0_l, temp1_l, temp2_l, temp3_l);
    temp0_r = (v4i32) ((v8i16) (temp0_r) + (v8i16) (temp0_l));
    temp1_r = (v4i32) ((v8i16) (temp1_r) + (v8i16) (temp1_l));
    temp2_r = (v4i32) ((v8i16) (temp2_r) + (v8i16) (temp2_l));
    temp3_r = (v4i32) ((v8i16) (temp3_r) + (v8i16) (temp3_l));
    temp0_r = (v4i32) CLIP_SH_0_255(temp0_r);
    temp1_r = (v4i32) CLIP_SH_0_255(temp1_r);
    temp2_r = (v4i32) CLIP_SH_0_255(temp2_r);
    temp3_r = (v4i32) CLIP_SH_0_255(temp3_r);
    PCKEV_B4_SW(temp0_r, temp0_r, temp1_r, temp1_r,
                temp2_r, temp2_r, temp3_r, temp3_r,
                temp0_r, temp1_r, temp2_r, temp3_r);
    tmp0 = __msa_copy_u_d((v2i64) temp0_r, 1);
    tmp1 = __msa_copy_u_d((v2i64) temp1_r, 1);
    tmp2 = __msa_copy_u_d((v2i64) temp2_r, 1);
    tmp3 = __msa_copy_u_d((v2i64) temp3_r, 1);
    in0 = (v8i16) (temp0_r) + (v8i16) (temp0_l);
    in1 = (v8i16) (temp1_r) + (v8i16) (temp1_l);
    in2 = (v8i16) (temp2_r) + (v8i16) (temp2_l);
    in3 = (v8i16) (temp3_r) + (v8i16) (temp3_l);
    CLIP_SH4_0_255(in0, in1, in2, in3);
    PCKEV_B4_SH(in0, in0, in1, in1, in2, in2, in3, in3,
                in0, in1, in2, in3);
    tmp0 = __msa_copy_u_d((v2i64) in0, 1);
    tmp1 = __msa_copy_u_d((v2i64) in1, 1);
    tmp2 = __msa_copy_u_d((v2i64) in2, 1);
    tmp3 = __msa_copy_u_d((v2i64) in3, 1);
    SD4(tmp0, tmp1, tmp2, tmp3, dst, dst_stride);
    SRA_4V(a3_r, a3_l, a2_r, a2_l, 20);
@@ -540,20 +527,17 @@ static void simple_idct_add_msa(uint8_t *dst, int32_t dst_stride,
                a0_r, a1_r, a2_r, a3_r);
    ILVR_B4_SW(zero, in4, zero, in5, zero, in6, zero, in7,
               a3_l, a2_l, a1_l, a0_l);
    a3_r = (v4i32) ((v8i16) (a3_r) + (v8i16) (a3_l));
    a2_r = (v4i32) ((v8i16) (a2_r) + (v8i16) (a2_l));
    a1_r = (v4i32) ((v8i16) (a1_r) + (v8i16) (a1_l));
    a0_r = (v4i32) ((v8i16) (a0_r) + (v8i16) (a0_l));
    a3_r = (v4i32) CLIP_SH_0_255(a3_r);
    a2_r = (v4i32) CLIP_SH_0_255(a2_r);
    a1_r = (v4i32) CLIP_SH_0_255(a1_r);
    a0_r = (v4i32) CLIP_SH_0_255(a0_r);
    PCKEV_B4_SW(a0_r, a0_r, a1_r, a1_r,
                a2_r, a2_r, a3_r, a3_r, a0_r, a1_r, a2_r, a3_r);
    tmp0 = __msa_copy_u_d((v2i64) a3_r, 1);
    tmp1 = __msa_copy_u_d((v2i64) a2_r, 1);
    tmp2 = __msa_copy_u_d((v2i64) a1_r, 1);
    tmp3 = __msa_copy_u_d((v2i64) a0_r, 1);
    in4 = (v8i16) (a3_r) + (v8i16) (a3_l);
    in5 = (v8i16) (a2_r) + (v8i16) (a2_l);
    in6 = (v8i16) (a1_r) + (v8i16) (a1_l);
    in7 = (v8i16) (a0_r) + (v8i16) (a0_l);
    CLIP_SH4_0_255(in4, in5, in6, in7);
    PCKEV_B4_SH(in4, in4, in5, in5, in6, in6, in7, in7,
                in4, in5, in6, in7);
    tmp0 = __msa_copy_u_d((v2i64) in4, 1);
    tmp1 = __msa_copy_u_d((v2i64) in5, 1);
    tmp2 = __msa_copy_u_d((v2i64) in6, 1);
    tmp3 = __msa_copy_u_d((v2i64) in7, 1);
    SD4(tmp0, tmp1, tmp2, tmp3, dst + 4 * dst_stride, dst_stride);
 }
--- a/libavcodec/mips/vp3dsp_idct_msa.c
+++ b/libavcodec/mips/vp3dsp_idct_msa.c
@@ -187,14 +187,7 @@ static void idct_msa(uint8_t *dst, int stride, int16_t *input, int type)
        G += c5;
        H += c6;
    }
    A = CLIP_SW_0_255(A);
    B = CLIP_SW_0_255(B);
    C = CLIP_SW_0_255(C);
    D = CLIP_SW_0_255(D);
    E = CLIP_SW_0_255(E);
    F = CLIP_SW_0_255(F);
    G = CLIP_SW_0_255(G);
    H = CLIP_SW_0_255(H);
    CLIP_SW8_0_255(A, B, C, D, E, F, G, H);
    sign_l = __msa_or_v((v16u8)r1_r, (v16u8)r2_r);
    sign_l = __msa_or_v(sign_l, (v16u8)r3_r);
    sign_l = __msa_or_v(sign_l, (v16u8)r0_l);
@@ -205,7 +198,7 @@ static void idct_msa(uint8_t *dst, int stride, int16_t *input, int type)
    Add = ((r0_r * cnst46341w) + (8 << 16)) >> 20;
    if (type == 1) {
        Bdd = Add + cnst128w;
        Bdd = CLIP_SW_0_255(Bdd);
        CLIP_SW_0_255(Bdd);
        Ad = Bdd;
        Bd = Bdd;
        Cd = Bdd;
@@ -223,14 +216,7 @@ static void idct_msa(uint8_t *dst, int stride, int16_t *input, int type)
        Fd = Add + c5;
        Gd = Add + c6;
        Hd = Add + c7;
        Ad = CLIP_SW_0_255(Ad);
        Bd = CLIP_SW_0_255(Bd);
        Cd = CLIP_SW_0_255(Cd);
        Dd = CLIP_SW_0_255(Dd);
        Ed = CLIP_SW_0_255(Ed);
        Fd = CLIP_SW_0_255(Fd);
        Gd = CLIP_SW_0_255(Gd);
        Hd = CLIP_SW_0_255(Hd);
        CLIP_SW8_0_255(Ad, Bd, Cd, Dd, Ed, Fd, Gd, Hd);
    }
    Ad = (v4i32)__msa_and_v((v16u8)Ad, (v16u8)sign_t);
    Bd = (v4i32)__msa_and_v((v16u8)Bd, (v16u8)sign_t);
@@ -309,14 +295,7 @@ static void idct_msa(uint8_t *dst, int stride, int16_t *input, int type)
        G += c5;
        H += c6;
    }
    A = CLIP_SW_0_255(A);
    B = CLIP_SW_0_255(B);
    C = CLIP_SW_0_255(C);
    D = CLIP_SW_0_255(D);
    E = CLIP_SW_0_255(E);
    F = CLIP_SW_0_255(F);
    G = CLIP_SW_0_255(G);
    H = CLIP_SW_0_255(H);
    CLIP_SW8_0_255(A, B, C, D, E, F, G, H);
    sign_l = __msa_or_v((v16u8)r5_r, (v16u8)r6_r);
    sign_l = __msa_or_v(sign_l, (v16u8)r7_r);
    sign_l = __msa_or_v(sign_l, (v16u8)r4_l);
@@ -327,7 +306,7 @@ static void idct_msa(uint8_t *dst, int stride, int16_t *input, int type)
    Add = ((r4_r * cnst46341w) + (8 << 16)) >> 20;
    if (type == 1) {
        Bdd = Add + cnst128w;
        Bdd = CLIP_SW_0_255(Bdd);
        CLIP_SW_0_255(Bdd);
        Ad = Bdd;
        Bd = Bdd;
        Cd = Bdd;
@@ -345,14 +324,7 @@ static void idct_msa(uint8_t *dst, int stride, int16_t *input, int type)
        Fd = Add + c5;
        Gd = Add + c6;
        Hd = Add + c7;
        Ad = CLIP_SW_0_255(Ad);
        Bd = CLIP_SW_0_255(Bd);
        Cd = CLIP_SW_0_255(Cd);
        Dd = CLIP_SW_0_255(Dd);
        Ed = CLIP_SW_0_255(Ed);
        Fd = CLIP_SW_0_255(Fd);
        Gd = CLIP_SW_0_255(Gd);
        Hd = CLIP_SW_0_255(Hd);
        CLIP_SW8_0_255(Ad, Bd, Cd, Dd, Ed, Fd, Gd, Hd);
    }
    Ad = (v4i32)__msa_and_v((v16u8)Ad, (v16u8)sign_t);
    Bd = (v4i32)__msa_and_v((v16u8)Bd, (v16u8)sign_t);
@@ -436,14 +408,7 @@ void ff_vp3_idct_dc_add_msa(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
    e5 += dc;
    e6 += dc;
    e7 += dc;
    e0 = CLIP_SW_0_255(e0);
    e1 = CLIP_SW_0_255(e1);
    e2 = CLIP_SW_0_255(e2);
    e3 = CLIP_SW_0_255(e3);
    e4 = CLIP_SW_0_255(e4);
    e5 = CLIP_SW_0_255(e5);
    e6 = CLIP_SW_0_255(e6);
    e7 = CLIP_SW_0_255(e7);
    CLIP_SW8_0_255(e0, e1, e2, e3, e4, e5, e6, e7);
    /* Left part */
    ILVL_H4_SW(zero, c0, zero, c1, zero, c2, zero, c3,
@@ -458,14 +423,7 @@ void ff_vp3_idct_dc_add_msa(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
    r5 += dc;
    r6 += dc;
    r7 += dc;
    r0 = CLIP_SW_0_255(r0);
    r1 = CLIP_SW_0_255(r1);
    r2 = CLIP_SW_0_255(r2);
    r3 = CLIP_SW_0_255(r3);
    r4 = CLIP_SW_0_255(r4);
    r5 = CLIP_SW_0_255(r5);
    r6 = CLIP_SW_0_255(r6);
    r7 = CLIP_SW_0_255(r7);
    CLIP_SW8_0_255(r0, r1, r2, r3, r4, r5, r6, r7);
    VSHF_B2_SB(e0, r0, e1, r1, mask, mask, d0, d1);
    VSHF_B2_SB(e2, r2, e3, r3, mask, mask, d2, d3);
    VSHF_B2_SB(e4, r4, e5, r5, mask, mask, d4, d5);
@@ -516,10 +474,7 @@ void ff_vp3_v_loop_filter_msa(uint8_t *first_pixel, ptrdiff_t stride,
    f1 += e1;
    g0 -= e0;
    g1 -= e1;
    f0 = CLIP_SW_0_255(f0);
    f1 = CLIP_SW_0_255(f1);
    g0 = CLIP_SW_0_255(g0);
    g1 = CLIP_SW_0_255(g1);
    CLIP_SW4_0_255(f0, f1, g0, g1);
    VSHF_B2_SB(f0, f1, g0, g1, mask, mask, d1, d2);
    /* Final move to first_pixel */
@@ -563,10 +518,7 @@ void ff_vp3_h_loop_filter_msa(uint8_t *first_pixel, ptrdiff_t stride,
    f1 += e1;
    g0 -= e0;
    g1 -= e1;
    f0 = CLIP_SW_0_255(f0);
    f1 = CLIP_SW_0_255(f1);
    g0 = CLIP_SW_0_255(g0);
    g1 = CLIP_SW_0_255(g1);
    CLIP_SW4_0_255(f0, f1, g0, g1);
    VSHF_B2_SB(f0, g0, f1, g1, mask, mask, d1, d2);
    /* Final move to first_pixel */
    ST_H4(d1, 0, 1, 2, 3, first_pixel - 1, stride);
--- a/libavcodec/mips/vp8_idct_msa.c
+++ b/libavcodec/mips/vp8_idct_msa.c
@@ -71,10 +71,7 @@ void ff_vp8_idct_add_msa(uint8_t *dst, int16_t input[16], ptrdiff_t stride)
    ILVR_H4_SW(zero, res0, zero, res1, zero, res2, zero, res3,
               res0, res1, res2, res3);
    ADD4(res0, vt0, res1, vt1, res2, vt2, res3, vt3, res0, res1, res2, res3);
    res0 = CLIP_SW_0_255(res0);
    res1 = CLIP_SW_0_255(res1);
    res2 = CLIP_SW_0_255(res2);
    res3 = CLIP_SW_0_255(res3);
    CLIP_SW4_0_255(res0, res1, res2, res3);
    VSHF_B2_SB(res0, res1, res2, res3, mask, mask, dest0, dest1);
    ST_W2(dest0, 0, 1, dst, stride);
    ST_W2(dest1, 0, 1, dst + 2 * stride, stride);
--- a/libavcodec/mips/vp9_idct_msa.c
+++ b/libavcodec/mips/vp9_idct_msa.c
@@ -763,13 +763,13 @@ static void vp9_iadst8x8_colcol_addblk_msa(int16_t *input, uint8_t *dst,
    res0 = (v8i16) __msa_ilvr_b((v16i8) zero, (v16i8) dst0);
    res0 += out0;
    res0 = CLIP_SH_0_255(res0);
    CLIP_SH_0_255(res0);
    res0 = (v8i16) __msa_pckev_b((v16i8) res0, (v16i8) res0);
    ST_D1(res0, 0, dst);
    res7 = (v8i16) __msa_ilvr_b((v16i8) zero, (v16i8) dst7);
    res7 += out7;
    res7 = CLIP_SH_0_255(res7);
    CLIP_SH_0_255(res7);
    res7 = (v8i16) __msa_pckev_b((v16i8) res7, (v16i8) res7);
    ST_D1(res7, 0, dst + 7 * dst_stride);
@@ -1192,8 +1192,7 @@ static void vp9_idct16x16_1_add_msa(int16_t *input, uint8_t *dst,
             res3);
        ADD4(res4, vec, res5, vec, res6, vec, res7, vec, res4, res5, res6,
             res7);
        CLIP_SH4_0_255(res0, res1, res2, res3);
        CLIP_SH4_0_255(res4, res5, res6, res7);
        CLIP_SH8_0_255(res0, res1, res2, res3, res4, res5, res6, res7);
        PCKEV_B4_UB(res4, res0, res5, res1, res6, res2, res7, res3,
                    tmp0, tmp1, tmp2, tmp3);
        ST_UB4(tmp0, tmp1, tmp2, tmp3, dst, dst_stride);
@@ -1981,8 +1980,7 @@ static void vp9_idct32x32_1_add_msa(int16_t *input, uint8_t *dst,
             res3);
        ADD4(res4, vec, res5, vec, res6, vec, res7, vec, res4, res5, res6,
             res7);
        CLIP_SH4_0_255(res0, res1, res2, res3);
        CLIP_SH4_0_255(res4, res5, res6, res7);
        CLIP_SH8_0_255(res0, res1, res2, res3, res4, res5, res6, res7);
        PCKEV_B4_UB(res4, res0, res5, res1, res6, res2, res7, res3,
                    tmp0, tmp1, tmp2, tmp3);
--- a/libavutil/mips/generic_macros_msa.h
+++ b/libavutil/mips/generic_macros_msa.h
@@ -933,99 +933,78 @@
 /* Description : Clips all halfword elements of input vector between min & max
                 out = ((in) < (min)) ? (min) : (((in) > (max)) ? (max) : (in))
   Arguments   : Inputs  - in       (input vector)
                         - min      (min threshold)
                         - max      (max threshold)
                 Outputs - out_m    (output vector with clipped elements)
   Arguments   : Inputs  - in    (input vector)
                         - min   (min threshold)
                         - max   (max threshold)
                 Outputs - in    (output vector with clipped elements)
                 Return Type - signed halfword
 */
 #define CLIP_SH(in, min, max)                           \
 ( {                                                     \
    v8i16 out_m;                                        \
                                                        \
    out_m = __msa_max_s_h((v8i16) min, (v8i16) in);     \
    out_m = __msa_min_s_h((v8i16) max, (v8i16) out_m);  \
    out_m;                                              \
 } )
 #define CLIP_SH(in, min, max)                     \
 {                                                 \
    in = __msa_max_s_h((v8i16) min, (v8i16) in);  \
    in = __msa_min_s_h((v8i16) max, (v8i16) in);  \
 }
 /* Description : Clips all signed halfword elements of input vector
                 between 0 & 255
   Arguments   : Inputs  - in       (input vector)
                 Outputs - out_m    (output vector with clipped elements)
                 Return Type - signed halfword
   Arguments   : Inputs  - in    (input vector)
                 Outputs - in    (output vector with clipped elements)
                 Return Type - signed halfwords
 */
 #define CLIP_SH_0_255(in)                                 \
 ( {                                                       \
    v8i16 max_m = __msa_ldi_h(255);                       \
    v8i16 out_m;                                          \
                                                          \
    out_m = __msa_maxi_s_h((v8i16) in, 0);                \
    out_m = __msa_min_s_h((v8i16) max_m, (v8i16) out_m);  \
    out_m;                                                \
 } )
 #define CLIP_SH_0_255(in)                       \
 {                                               \
    in = __msa_maxi_s_h((v8i16) in, 0);         \
    in = (v8i16) __msa_sat_u_h((v8u16) in, 7);  \
 }
 #define CLIP_SH2_0_255(in0, in1)  \
 {                                 \
    in0 = CLIP_SH_0_255(in0);     \
    in1 = CLIP_SH_0_255(in1);     \
    CLIP_SH_0_255(in0);           \
    CLIP_SH_0_255(in1);           \
 }
 #define CLIP_SH4_0_255(in0, in1, in2, in3)  \
 {                                           \
    CLIP_SH2_0_255(in0, in1);               \
    CLIP_SH2_0_255(in2, in3);               \
 }
 #define CLIP_SH_0_255_MAX_SATU(in)                    \
 ( {                                                   \
    v8i16 out_m;                                      \
                                                      \
    out_m = __msa_maxi_s_h((v8i16) in, 0);            \
    out_m = (v8i16) __msa_sat_u_h((v8u16) out_m, 7);  \
    out_m;                                            \
 } )
 #define CLIP_SH2_0_255_MAX_SATU(in0, in1)  \
 {                                          \
    in0 = CLIP_SH_0_255_MAX_SATU(in0);     \
    in1 = CLIP_SH_0_255_MAX_SATU(in1);     \
 }
 #define CLIP_SH4_0_255_MAX_SATU(in0, in1, in2, in3)  \
 {                                                    \
    CLIP_SH2_0_255_MAX_SATU(in0, in1);               \
    CLIP_SH2_0_255_MAX_SATU(in2, in3);               \
 #define CLIP_SH8_0_255(in0, in1, in2, in3,  \
                       in4, in5, in6, in7)  \
 {                                           \
    CLIP_SH4_0_255(in0, in1, in2, in3);     \
    CLIP_SH4_0_255(in4, in5, in6, in7);     \
 }
 /* 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)
   Arguments   : Inputs  - in    (input vector)
                 Outputs - in    (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;                                                \
 } )
 #define CLIP_SW_0_255(in)                       \
 {                                               \
    in = __msa_maxi_s_w((v4i32) in, 0);         \
    in = (v4i32) __msa_sat_u_w((v4u32) in, 7);  \
 }
 #define CLIP_SW_0_255_MAX_SATU(in)                    \
 ( {                                                   \
    v4i32 out_m;                                      \
                                                      \
    out_m = __msa_maxi_s_w((v4i32) in, 0);            \
    out_m = (v4i32) __msa_sat_u_w((v4u32) out_m, 7);  \
    out_m;                                            \
 } )
 #define CLIP_SW2_0_255_MAX_SATU(in0, in1)  \
 {                                          \
    in0 = CLIP_SW_0_255_MAX_SATU(in0);     \
    in1 = CLIP_SW_0_255_MAX_SATU(in1);     \
 #define CLIP_SW2_0_255(in0, in1)  \
 {                                 \
    CLIP_SW_0_255(in0);           \
    CLIP_SW_0_255(in1);           \
 }
 #define CLIP_SW4_0_255_MAX_SATU(in0, in1, in2, in3)  \
 {                                                    \
    CLIP_SW2_0_255_MAX_SATU(in0, in1);               \
    CLIP_SW2_0_255_MAX_SATU(in2, in3);               \
 #define CLIP_SW4_0_255(in0, in1, in2, in3)  \
 {                                           \
    CLIP_SW2_0_255(in0, in1);               \
    CLIP_SW2_0_255(in2, in3);               \
 }
 #define CLIP_SW8_0_255(in0, in1, in2, in3,  \
                       in4, in5, in6, in7)  \
 {                                           \
    CLIP_SW4_0_255(in0, in1, in2, in3);     \
    CLIP_SW4_0_255(in4, in5, in6, in7);     \
 }
 /* Description : Addition of 4 signed word elements