avcodec/mips: Improve avc chroma hv mc msa functions

Replace generic with block size specific function.

Signed-off-by: Kaustubh Raste <kaustubh.raste@imgtec.com>
Reviewed-by: Manojkumar Bhosale <Manojkumar.Bhosale@imgtec.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

1 files changed, 166 insertions, 143 deletions

diff --git a/libavcodec/mips/h264chroma_msa.c b/libavcodec/mips/h264chroma_msa.c
index 16e2fe4ec7..b8fcf6d012 100644
--- a/libavcodec/mips/h264chroma_msa.c
+++ b/libavcodec/mips/h264chroma_msa.c
@@ -526,8 +526,7 @@ static void avc_chroma_vt_8w_msa(uint8_t *src, uint8_t *dst, int32_t stride,
     }
 }
 
-static void avc_chroma_hv_2x2_msa(uint8_t *src, int32_t src_stride,
-                                  uint8_t *dst, int32_t dst_stride,
+static void avc_chroma_hv_2x2_msa(uint8_t *src, uint8_t *dst, int32_t stride,
                                   uint32_t coef_hor0, uint32_t coef_hor1,
                                   uint32_t coef_ver0, uint32_t coef_ver1)
 {
@@ -544,7 +543,7 @@ static void avc_chroma_hv_2x2_msa(uint8_t *src, int32_t src_stride,
 
     mask = LD_SB(&chroma_mask_arr[48]);
 
-    LD_UB3(src, src_stride, src0, src1, src2);
+    LD_UB3(src, stride, src0, src1, src2);
     VSHF_B2_UB(src0, src1, src1, src2, mask, mask, src0, src1);
     DOTP_UB2_UH(src0, src1, coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1);
     MUL2(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_vt0, res_vt1);
@@ -558,12 +557,11 @@ static void avc_chroma_hv_2x2_msa(uint8_t *src, int32_t src_stride,
     out1 = __msa_copy_u_h(res_vert, 1);
 
     SH(out0, dst);
-    dst += dst_stride;
+    dst += stride;
     SH(out1, dst);
 }
 
-static void avc_chroma_hv_2x4_msa(uint8_t *src, int32_t src_stride,
-                                  uint8_t *dst, int32_t dst_stride,
+static void avc_chroma_hv_2x4_msa(uint8_t *src, uint8_t *dst, int32_t stride,
                                   uint32_t coef_hor0, uint32_t coef_hor1,
                                   uint32_t coef_ver0, uint32_t coef_ver1)
 {
@@ -580,7 +578,7 @@ static void avc_chroma_hv_2x4_msa(uint8_t *src, int32_t src_stride,
 
     mask = LD_SB(&chroma_mask_arr[48]);
 
-    LD_UB5(src, src_stride, src0, src1, src2, src3, src4);
+    LD_UB5(src, stride, src0, src1, src2, src3, src4);
 
     VSHF_B2_UB(src0, src1, src2, src3, mask, mask, tmp0, tmp1);
     VSHF_B2_UB(src1, src2, src3, src4, mask, mask, tmp2, tmp3);
@@ -591,83 +589,27 @@ static void avc_chroma_hv_2x4_msa(uint8_t *src, int32_t src_stride,
     res_vt0 += res_vt1;
     res_vt0 = (v8u16) __msa_srari_h((v8i16) res_vt0, 6);
     res_vt0 = __msa_sat_u_h(res_vt0, 7);
-    res = (v8i16) __msa_pckev_b((v16i8) res_vt0, (v16i8) res_vt0);
-
-    ST2x4_UB(res, 0, dst, dst_stride);
-}
-
-static void avc_chroma_hv_2x8_msa(uint8_t *src, int32_t src_stride,
-                                  uint8_t *dst, int32_t dst_stride,
-                                  uint32_t coef_hor0, uint32_t coef_hor1,
-                                  uint32_t coef_ver0, uint32_t coef_ver1)
-{
-    v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
-    v16u8 tmp0, tmp1, tmp2, tmp3;
-    v8u16 res_hz0, res_hz1, res_vt0, res_vt1;
-    v8i16 res;
-    v16i8 mask;
-    v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
-    v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
-    v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
-    v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
-    v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
-
-    mask = LD_SB(&chroma_mask_arr[48]);
-
-    LD_UB5(src, src_stride, src0, src1, src2, src3, src4);
-    src += (5 * src_stride);
-    LD_UB4(src, src_stride, src5, src6, src7, src8);
-
-    VSHF_B2_UB(src0, src1, src2, src3, mask, mask, tmp0, tmp1);
-    VSHF_B2_UB(src1, src2, src3, src4, mask, mask, tmp2, tmp3);
-    ILVR_D2_UB(tmp1, tmp0, tmp3, tmp2, src0, src1);
-    VSHF_B2_UB(src4, src5, src6, src7, mask, mask, tmp0, tmp1);
-    VSHF_B2_UB(src5, src6, src7, src8, mask, mask, tmp2, tmp3);
-    ILVR_D2_UB(tmp1, tmp0, tmp3, tmp2, src4, src5);
-    DOTP_UB2_UH(src0, src1, coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1);
-    MUL2(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_vt0, res_vt1);
-
-    res_vt0 += res_vt1;
-    res_vt0 = (v8u16) __msa_srari_h((v8i16) res_vt0, 6);
-    res_vt0 = __msa_sat_u_h(res_vt0, 7);
 
     res = (v8i16) __msa_pckev_b((v16i8) res_vt0, (v16i8) res_vt0);
 
-    ST2x4_UB(res, 0, dst, dst_stride);
-    dst += (4 * dst_stride);
-
-    DOTP_UB2_UH(src4, src5, coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1);
-    MUL2(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_vt0, res_vt1);
-
-    res_vt0 += res_vt1;
-    res_vt0 = (v8u16) __msa_srari_h((v8i16) res_vt0, 6);
-    res_vt0 = __msa_sat_u_h(res_vt0, 7);
-
-    res = (v8i16) __msa_pckev_b((v16i8) res_vt0, (v16i8) res_vt0);
-
-    ST2x4_UB(res, 0, dst, dst_stride);
+    ST2x4_UB(res, 0, dst, stride);
 }
 
-static void avc_chroma_hv_2w_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
+static void avc_chroma_hv_2w_msa(uint8_t *src, uint8_t *dst, int32_t stride,
                                  uint32_t coef_hor0, uint32_t coef_hor1,
                                  uint32_t coef_ver0, uint32_t coef_ver1,
                                  int32_t height)
 {
     if (2 == height) {
-        avc_chroma_hv_2x2_msa(src, src_stride, dst, dst_stride, coef_hor0,
-                              coef_hor1, coef_ver0, coef_ver1);
+        avc_chroma_hv_2x2_msa(src, dst, stride, coef_hor0, coef_hor1, coef_ver0,
+                              coef_ver1);
     } else if (4 == height) {
-        avc_chroma_hv_2x4_msa(src, src_stride, dst, dst_stride, coef_hor0,
-                              coef_hor1, coef_ver0, coef_ver1);
-    } else if (8 == height) {
-        avc_chroma_hv_2x8_msa(src, src_stride, dst, dst_stride, coef_hor0,
-                              coef_hor1, coef_ver0, coef_ver1);
+        avc_chroma_hv_2x4_msa(src, dst, stride, coef_hor0, coef_hor1, coef_ver0,
+                              coef_ver1);
     }
 }
 
-static void avc_chroma_hv_4x2_msa(uint8_t *src, int32_t src_stride,
-                                  uint8_t *dst, int32_t dst_stride,
+static void avc_chroma_hv_4x2_msa(uint8_t *src, uint8_t *dst, int32_t stride,
                                   uint32_t coef_hor0, uint32_t coef_hor1,
                                   uint32_t coef_ver0, uint32_t coef_ver1)
 {
@@ -682,7 +624,7 @@ static void avc_chroma_hv_4x2_msa(uint8_t *src, int32_t src_stride,
     v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
 
     mask = LD_SB(&chroma_mask_arr[0]);
-    LD_UB3(src, src_stride, src0, src1, src2);
+    LD_UB3(src, stride, src0, src1, src2);
     VSHF_B2_UB(src0, src1, src1, src2, mask, mask, src0, src1);
     DOTP_UB2_UH(src0, src1, coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1);
     MUL2(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_vt0, res_vt1);
@@ -692,18 +634,13 @@ static void avc_chroma_hv_4x2_msa(uint8_t *src, int32_t src_stride,
     res_vt0 = __msa_sat_u_h(res_vt0, 7);
     res = (v4i32) __msa_pckev_b((v16i8) res_vt0, (v16i8) res_vt0);
 
-    ST4x2_UB(res, dst, dst_stride);
+    ST4x2_UB(res, dst, stride);
 }
 
-static void avc_chroma_hv_4x4multiple_msa(uint8_t *src, int32_t src_stride,
-                                          uint8_t *dst, int32_t dst_stride,
-                                          uint32_t coef_hor0,
-                                          uint32_t coef_hor1,
-                                          uint32_t coef_ver0,
-                                          uint32_t coef_ver1,
-                                          int32_t height)
+static void avc_chroma_hv_4x4_msa(uint8_t *src, uint8_t *dst, int32_t stride,
+                                  uint32_t coef_hor0, uint32_t coef_hor1,
+                                  uint32_t coef_ver0, uint32_t coef_ver1)
 {
-    uint32_t row;
     v16u8 src0, src1, src2, src3, src4;
     v8u16 res_hz0, res_hz1, res_hz2, res_hz3;
     v8u16 res_vt0, res_vt1, res_vt2, res_vt3;
@@ -717,55 +654,82 @@ static void avc_chroma_hv_4x4multiple_msa(uint8_t *src, int32_t src_stride,
 
     mask = LD_SB(&chroma_mask_arr[0]);
 
-    src0 = LD_UB(src);
-    src += src_stride;
+    LD_UB5(src, stride, src0, src1, src2, src3, src4);
+    VSHF_B2_UB(src0, src1, src1, src2, mask, mask, src0, src1);
+    VSHF_B2_UB(src2, src3, src3, src4, mask, mask, src2, src3);
+    DOTP_UB4_UH(src0, src1, src2, src3, coeff_hz_vec, coeff_hz_vec,
+                coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1, res_hz2,
+                res_hz3);
+    MUL4(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_hz2, coeff_vt_vec1,
+         res_hz3, coeff_vt_vec0, res_vt0, res_vt1, res_vt2, res_vt3);
+    ADD2(res_vt0, res_vt1, res_vt2, res_vt3, res_vt0, res_vt1);
+    SRARI_H2_UH(res_vt0, res_vt1, 6);
+    SAT_UH2_UH(res_vt0, res_vt1, 7);
+    PCKEV_B2_SW(res_vt0, res_vt0, res_vt1, res_vt1, res0, res1);
+    ST4x4_UB(res0, res1, 0, 1, 0, 1, dst, stride);
+}
 
-    for (row = (height >> 2); row--;) {
-        LD_UB4(src, src_stride, src1, src2, src3, src4);
-        src += (4 * src_stride);
+static void avc_chroma_hv_4x8_msa(uint8_t *src, uint8_t *dst, int32_t stride,
+                                  uint32_t coef_hor0, uint32_t coef_hor1,
+                                  uint32_t coef_ver0, uint32_t coef_ver1)
+{
+    v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8, res0, res1;
+    v8u16 res_hz0, res_hz1, res_hz2, res_hz3, res_hz4, res_hz5, res_hz6, res_hz7;
+    v8u16 res_vt0, res_vt1, res_vt2, res_vt3, res_vt4, res_vt5, res_vt6, res_vt7;
+    v16i8 mask;
+    v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
+    v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
+    v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
+    v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
+    v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
 
-        VSHF_B2_UB(src0, src1, src1, src2, mask, mask, src0, src1);
-        VSHF_B2_UB(src2, src3, src3, src4, mask, mask, src2, src3);
-        DOTP_UB4_UH(src0, src1, src2, src3, coeff_hz_vec, coeff_hz_vec,
-                    coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1, res_hz2,
-                    res_hz3);
-        MUL4(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_hz2,
-             coeff_vt_vec1, res_hz3, coeff_vt_vec0, res_vt0, res_vt1, res_vt2,
-             res_vt3);
-        ADD2(res_vt0, res_vt1, res_vt2, res_vt3, res_vt0, res_vt1);
-        SRARI_H2_UH(res_vt0, res_vt1, 6);
-        SAT_UH2_UH(res_vt0, res_vt1, 7);
-        PCKEV_B2_SW(res_vt0, res_vt0, res_vt1, res_vt1, res0, res1);
+    mask = LD_SB(&chroma_mask_arr[0]);
 
-        ST4x4_UB(res0, res1, 0, 1, 0, 1, dst, dst_stride);
-        dst += (4 * dst_stride);
-        src0 = src4;
-    }
+    LD_UB5(src, stride, src0, src1, src2, src3, src4);
+    src += (5 * stride);
+    LD_UB4(src, stride, src5, src6, src7, src8);
+
+    VSHF_B2_UB(src0, src1, src1, src2, mask, mask, src0, src1);
+    VSHF_B2_UB(src2, src3, src3, src4, mask, mask, src2, src3);
+    VSHF_B2_UB(src4, src5, src5, src6, mask, mask, src4, src5);
+    VSHF_B2_UB(src6, src7, src7, src8, mask, mask, src6, src7);
+    DOTP_UB4_UH(src0, src1, src2, src3, coeff_hz_vec, coeff_hz_vec,
+                coeff_hz_vec, coeff_hz_vec, res_hz0, res_hz1, res_hz2, res_hz3);
+    DOTP_UB4_UH(src4, src5, src6, src7, coeff_hz_vec, coeff_hz_vec,
+                coeff_hz_vec, coeff_hz_vec, res_hz4, res_hz5, res_hz6, res_hz7);
+    MUL4(res_hz0, coeff_vt_vec1, res_hz1, coeff_vt_vec0, res_hz2, coeff_vt_vec1,
+         res_hz3, coeff_vt_vec0, res_vt0, res_vt1, res_vt2, res_vt3);
+    MUL4(res_hz4, coeff_vt_vec1, res_hz5, coeff_vt_vec0, res_hz6, coeff_vt_vec1,
+         res_hz7, coeff_vt_vec0, res_vt4, res_vt5, res_vt6, res_vt7);
+    ADD2(res_vt0, res_vt1, res_vt2, res_vt3, res_vt0, res_vt1);
+    ADD2(res_vt4, res_vt5, res_vt6, res_vt7, res_vt2, res_vt3);
+    SRARI_H4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 6);
+    SAT_UH4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 7);
+    PCKEV_B2_UB(res_vt1, res_vt0, res_vt3, res_vt2, res0, res1);
+    ST4x8_UB(res0, res1, dst, stride);
 }
 
-static void avc_chroma_hv_4w_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
+static void avc_chroma_hv_4w_msa(uint8_t *src, uint8_t *dst, int32_t stride,
                                  uint32_t coef_hor0, uint32_t coef_hor1,
                                  uint32_t coef_ver0, uint32_t coef_ver1,
                                  int32_t height)
 {
     if (2 == height) {
-        avc_chroma_hv_4x2_msa(src, src_stride, dst, dst_stride, coef_hor0,
-                              coef_hor1, coef_ver0, coef_ver1);
-    } else {
-        avc_chroma_hv_4x4multiple_msa(src, src_stride, dst, dst_stride,
-                                      coef_hor0, coef_hor1, coef_ver0,
-                                      coef_ver1, height);
+        avc_chroma_hv_4x2_msa(src, dst, stride, coef_hor0, coef_hor1, coef_ver0,
+                              coef_ver1);
+    } else if (4 == height) {
+        avc_chroma_hv_4x4_msa(src, dst, stride, coef_hor0, coef_hor1, coef_ver0,
+                              coef_ver1);
+    } else if (8 == height) {
+        avc_chroma_hv_4x8_msa(src, dst, stride, coef_hor0, coef_hor1, coef_ver0,
+                              coef_ver1);
     }
 }
 
-static void avc_chroma_hv_8w_msa(uint8_t *src, int32_t src_stride,
-                                 uint8_t *dst, int32_t dst_stride,
-                                 uint32_t coef_hor0, uint32_t coef_hor1,
-                                 uint32_t coef_ver0, uint32_t coef_ver1,
-                                 int32_t height)
+static void avc_chroma_hv_8x4_msa(uint8_t *src, uint8_t *dst, int32_t stride,
+                                  uint32_t coef_hor0, uint32_t coef_hor1,
+                                  uint32_t coef_ver0, uint32_t coef_ver1)
 {
-    uint32_t row;
     v16u8 src0, src1, src2, src3, src4, out0, out1;
     v8u16 res_hz0, res_hz1, res_hz2, res_hz3, res_hz4;
     v8u16 res_vt0, res_vt1, res_vt2, res_vt3;
@@ -779,37 +743,99 @@ static void avc_chroma_hv_8w_msa(uint8_t *src, int32_t src_stride,
     mask = LD_SB(&chroma_mask_arr[32]);
 
     src0 = LD_UB(src);
-    src += src_stride;
+    src += stride;
 
     src0 = (v16u8) __msa_vshf_b(mask, (v16i8) src0, (v16i8) src0);
     res_hz0 = __msa_dotp_u_h(src0, coeff_hz_vec);
 
-    for (row = (height >> 2); row--;) {
-        LD_UB4(src, src_stride, src1, src2, src3, src4);
-        src += (4 * src_stride);
+    LD_UB4(src, stride, src1, src2, src3, src4);
+    src += (4 * stride);
 
-        VSHF_B2_UB(src1, src1, src2, src2, mask, mask, src1, src2);
-        VSHF_B2_UB(src3, src3, src4, src4, mask, mask, src3, src4);
-        DOTP_UB4_UH(src1, src2, src3, src4, coeff_hz_vec, coeff_hz_vec,
-                    coeff_hz_vec, coeff_hz_vec, res_hz1, res_hz2, res_hz3,
-                    res_hz4);
-        MUL4(res_hz1, coeff_vt_vec0, res_hz2, coeff_vt_vec0, res_hz3,
-             coeff_vt_vec0, res_hz4, coeff_vt_vec0, res_vt0, res_vt1, res_vt2,
-             res_vt3);
+    VSHF_B2_UB(src1, src1, src2, src2, mask, mask, src1, src2);
+    VSHF_B2_UB(src3, src3, src4, src4, mask, mask, src3, src4);
+    DOTP_UB4_UH(src1, src2, src3, src4, coeff_hz_vec, coeff_hz_vec,
+                coeff_hz_vec, coeff_hz_vec, res_hz1, res_hz2, res_hz3, res_hz4);
+    MUL4(res_hz1, coeff_vt_vec0, res_hz2, coeff_vt_vec0, res_hz3, coeff_vt_vec0,
+         res_hz4, coeff_vt_vec0, res_vt0, res_vt1, res_vt2, res_vt3);
 
-        res_vt0 += (res_hz0 * coeff_vt_vec1);
-        res_vt1 += (res_hz1 * coeff_vt_vec1);
-        res_vt2 += (res_hz2 * coeff_vt_vec1);
-        res_vt3 += (res_hz3 * coeff_vt_vec1);
+    res_vt0 += (res_hz0 * coeff_vt_vec1);
+    res_vt1 += (res_hz1 * coeff_vt_vec1);
+    res_vt2 += (res_hz2 * coeff_vt_vec1);
+    res_vt3 += (res_hz3 * coeff_vt_vec1);
 
-        SRARI_H4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 6);
-        SAT_UH4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 7);
-        PCKEV_B2_UB(res_vt1, res_vt0, res_vt3, res_vt2, out0, out1);
-        ST8x4_UB(out0, out1, dst, dst_stride);
+    SRARI_H4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 6);
+    SAT_UH4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 7);
+    PCKEV_B2_UB(res_vt1, res_vt0, res_vt3, res_vt2, out0, out1);
+    ST8x4_UB(out0, out1, dst, stride);
+}
 
-        dst += (4 * dst_stride);
+static void avc_chroma_hv_8x8_msa(uint8_t *src, uint8_t *dst, int32_t stride,
+                                  uint32_t coef_hor0, uint32_t coef_hor1,
+                                  uint32_t coef_ver0, uint32_t coef_ver1)
+{
+    v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
+    v16u8 out0, out1, out2, out3;
+    v8u16 res_hz0, res_hz1, res_hz2, res_hz3, res_hz4;
+    v8u16 res_hz5, res_hz6, res_hz7, res_hz8;
+    v8u16 res_vt0, res_vt1, res_vt2, res_vt3;
+    v8u16 res_vt4, res_vt5, res_vt6, res_vt7;
+    v16i8 mask;
+    v16i8 coeff_hz_vec0 = __msa_fill_b(coef_hor0);
+    v16i8 coeff_hz_vec1 = __msa_fill_b(coef_hor1);
+    v16u8 coeff_hz_vec = (v16u8) __msa_ilvr_b(coeff_hz_vec0, coeff_hz_vec1);
+    v8u16 coeff_vt_vec0 = (v8u16) __msa_fill_h(coef_ver0);
+    v8u16 coeff_vt_vec1 = (v8u16) __msa_fill_h(coef_ver1);
 
-        res_hz0 = res_hz4;
+    mask = LD_SB(&chroma_mask_arr[32]);
+
+    LD_UB5(src, stride, src0, src1, src2, src3, src4);
+    src += (5 * stride);
+    LD_UB4(src, stride, src5, src6, src7, src8);
+    src0 = (v16u8) __msa_vshf_b(mask, (v16i8) src0, (v16i8) src0);
+    VSHF_B2_UB(src1, src1, src2, src2, mask, mask, src1, src2);
+    VSHF_B2_UB(src3, src3, src4, src4, mask, mask, src3, src4);
+    VSHF_B2_UB(src5, src5, src6, src6, mask, mask, src5, src6);
+    VSHF_B2_UB(src7, src7, src8, src8, mask, mask, src7, src8);
+    res_hz0 = __msa_dotp_u_h(src0, coeff_hz_vec);
+    DOTP_UB4_UH(src1, src2, src3, src4, coeff_hz_vec, coeff_hz_vec,
+                coeff_hz_vec, coeff_hz_vec, res_hz1, res_hz2, res_hz3,
+                res_hz4);
+    DOTP_UB4_UH(src5, src6, src7, src8, coeff_hz_vec, coeff_hz_vec,
+                coeff_hz_vec, coeff_hz_vec, res_hz5, res_hz6, res_hz7, res_hz8);
+    MUL4(res_hz1, coeff_vt_vec0, res_hz2, coeff_vt_vec0, res_hz3,
+         coeff_vt_vec0, res_hz4, coeff_vt_vec0, res_vt0, res_vt1, res_vt2,
+         res_vt3);
+    MUL4(res_hz5, coeff_vt_vec0, res_hz6, coeff_vt_vec0, res_hz7,
+         coeff_vt_vec0, res_hz8, coeff_vt_vec0, res_vt4, res_vt5, res_vt6,
+         res_vt7);
+    res_vt0 += (res_hz0 * coeff_vt_vec1);
+    res_vt1 += (res_hz1 * coeff_vt_vec1);
+    res_vt2 += (res_hz2 * coeff_vt_vec1);
+    res_vt3 += (res_hz3 * coeff_vt_vec1);
+    res_vt4 += (res_hz4 * coeff_vt_vec1);
+    res_vt5 += (res_hz5 * coeff_vt_vec1);
+    res_vt6 += (res_hz6 * coeff_vt_vec1);
+    res_vt7 += (res_hz7 * coeff_vt_vec1);
+    SRARI_H4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 6);
+    SRARI_H4_UH(res_vt4, res_vt5, res_vt6, res_vt7, 6);
+    SAT_UH4_UH(res_vt0, res_vt1, res_vt2, res_vt3, 7);
+    SAT_UH4_UH(res_vt4, res_vt5, res_vt6, res_vt7, 7);
+    PCKEV_B2_UB(res_vt1, res_vt0, res_vt3, res_vt2, out0, out1);
+    PCKEV_B2_UB(res_vt5, res_vt4, res_vt7, res_vt6, out2, out3);
+    ST8x8_UB(out0, out1, out2, out3, dst, stride);
+}
+
+static void avc_chroma_hv_8w_msa(uint8_t *src, uint8_t *dst, int32_t stride,
+                                 uint32_t coef_hor0, uint32_t coef_hor1,
+                                 uint32_t coef_ver0, uint32_t coef_ver1,
+                                 int32_t height)
+{
+    if (4 == height) {
+        avc_chroma_hv_8x4_msa(src, dst, stride, coef_hor0, coef_hor1, coef_ver0,
+                              coef_ver1);
+    } else if (8 == height) {
+        avc_chroma_hv_8x8_msa(src, dst, stride, coef_hor0, coef_hor1, coef_ver0,
+                              coef_ver1);
     }
 }
 
@@ -1896,8 +1922,7 @@ void ff_put_h264_chroma_mc8_msa(uint8_t *dst, uint8_t *src,
     av_assert2(x < 8 && y < 8 && x >= 0 && y >= 0);
 
     if (x && y) {
-        avc_chroma_hv_8w_msa(src, stride, dst,
-                             stride, x, (8 - x), y, (8 - y), height);
+        avc_chroma_hv_8w_msa(src, dst, stride, x, (8 - x), y, (8 - y), height);
     } else if (x) {
         avc_chroma_hz_8w_msa(src, dst, stride, x, (8 - x), height);
     } else if (y) {
@@ -1915,8 +1940,7 @@ void ff_put_h264_chroma_mc4_msa(uint8_t *dst, uint8_t *src,
     av_assert2(x < 8 && y < 8 && x >= 0 && y >= 0);
 
     if (x && y) {
-        avc_chroma_hv_4w_msa(src, stride, dst,
-                             stride, x, (8 - x), y, (8 - y), height);
+        avc_chroma_hv_4w_msa(src, dst, stride, x, (8 - x), y, (8 - y), height);
     } else if (x) {
         avc_chroma_hz_4w_msa(src, dst, stride, x, (8 - x), height);
     } else if (y) {
@@ -1939,8 +1963,7 @@ void ff_put_h264_chroma_mc2_msa(uint8_t *dst, uint8_t *src,
     av_assert2(x < 8 && y < 8 && x >= 0 && y >= 0);
 
     if (x && y) {
-        avc_chroma_hv_2w_msa(src, stride, dst,
-                             stride, x, (8 - x), y, (8 - y), height);
+        avc_chroma_hv_2w_msa(src, dst, stride, x, (8 - x), y, (8 - y), height);
     } else if (x) {
         avc_chroma_hz_2w_msa(src, dst, stride, x, (8 - x), height);
     } else if (y) {