avcodec/mips: Improve vp9 mc msa functions

Load the specific destination bytes instead of MSA load and pack.
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>

libavcodec/mips/h264qpel_msa.c

@@ -1479,7 +1479,8 @@ static void avc_luma_hz_and_aver_dst_8x8_msa(const uint8_t *src,
                      plus20b, res0, res1, res2, res3);
         SRARI_H4_SH(res0, res1, res2, res3, 5);
         SAT_SH4_SH(res0, res1, res2, res3, 7);
-        CONVERT_UB_AVG_ST8x4_UB(res0, res1, res2, res3, dst0, dst1, dst2, dst3,
+        ILVR_D2_UB(dst1, dst0, dst3, dst2, dst0, dst1);
+        CONVERT_UB_AVG_ST8x4_UB(res0, res1, res2, res3, dst0, dst1,
                                 dst, dst_stride);
 
         dst += (4 * dst_stride);
@@ -1825,8 +1826,8 @@ static void avc_luma_vt_and_aver_dst_8x8_msa(const uint8_t *src,
         SRARI_H4_SH(out0, out1, out2, out3, 5);
         SAT_SH4_SH(out0, out1, out2, out3, 7);
         LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
-
-        CONVERT_UB_AVG_ST8x4_UB(out0, out1, out2, out3, dst0, dst1, dst2, dst3,
+        ILVR_D2_UB(dst1, dst0, dst3, dst2, dst0, dst1);
+        CONVERT_UB_AVG_ST8x4_UB(out0, out1, out2, out3, dst0, dst1,
                                 dst, dst_stride);
         dst += (4 * dst_stride);
 
@@ -2229,7 +2230,8 @@ static void avc_luma_mid_and_aver_dst_8w_msa(const uint8_t *src,
         res3 = AVC_CALC_DPADD_H_6PIX_2COEFF_SH(hz_out3, hz_out4, hz_out5,
                                                hz_out6, hz_out7, hz_out8);
         LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
-        CONVERT_UB_AVG_ST8x4_UB(res0, res1, res2, res3, dst0, dst1, dst2, dst3,
+        ILVR_D2_UB(dst1, dst0, dst3, dst2, dst0, dst1);
+        CONVERT_UB_AVG_ST8x4_UB(res0, res1, res2, res3, dst0, dst1,
                                 dst, dst_stride);
 
         dst += (4 * dst_stride);
@@ -2518,8 +2520,8 @@ static void avc_luma_midv_qrt_and_aver_dst_8w_msa(const uint8_t *src,
         res1 = __msa_aver_s_h(res2, res3);
         res2 = __msa_aver_s_h(res4, res5);
         res3 = __msa_aver_s_h(res6, res7);
-
-        CONVERT_UB_AVG_ST8x4_UB(res0, res1, res2, res3, dst0, dst1, dst2, dst3,
+        ILVR_D2_UB(dst1, dst0, dst3, dst2, dst0, dst1);
+        CONVERT_UB_AVG_ST8x4_UB(res0, res1, res2, res3, dst0, dst1,
                                 dst, dst_stride);
         dst += (4 * dst_stride);
 
@@ -2676,7 +2678,8 @@ static void avc_luma_hv_qrt_and_aver_dst_8x8_msa(const uint8_t *src_x,
         out3 = __msa_srari_h((hz_out3 + vert_out3), 1);
 
         SAT_SH4_SH(out0, out1, out2, out3, 7);
-        CONVERT_UB_AVG_ST8x4_UB(out0, out1, out2, out3, dst0, dst1, dst2, dst3,
+        ILVR_D2_UB(dst1, dst0, dst3, dst2, dst0, dst1);
+        CONVERT_UB_AVG_ST8x4_UB(out0, out1, out2, out3, dst0, dst1,
                                 dst, dst_stride);
         dst += (4 * dst_stride);
 

libavutil/mips/generic_macros_msa.h

@@ -2773,19 +2773,17 @@
 
 /* Description : Converts inputs to unsigned bytes, interleave, average & store
                  as 8x4 unsigned byte block
-   Arguments   : Inputs  - in0, in1, in2, in3, dst0, dst1, dst2, dst3,
-                           pdst, stride
+   Arguments   : Inputs  - in0, in1, in2, in3, dst0, dst1, pdst, stride
 */
 #define CONVERT_UB_AVG_ST8x4_UB(in0, in1, in2, in3,           \
-                                dst0, dst1, dst2, dst3, pdst, stride)  \
+                                dst0, dst1, pdst, stride)     \
 {                                                             \
-    v16u8 tmp0_m, tmp1_m, tmp2_m, tmp3_m;                              \
+    v16u8 tmp0_m, tmp1_m;                                     \
     uint8_t *pdst_m = (uint8_t *) (pdst);                     \
                                                               \
     tmp0_m = PCKEV_XORI128_UB(in0, in1);                      \
     tmp1_m = PCKEV_XORI128_UB(in2, in3);                      \
-    ILVR_D2_UB(dst1, dst0, dst3, dst2, tmp2_m, tmp3_m);                \
-    AVER_UB2_UB(tmp0_m, tmp2_m, tmp1_m, tmp3_m, tmp0_m, tmp1_m);       \
+    AVER_UB2_UB(tmp0_m, dst0, tmp1_m, dst1, tmp0_m, tmp1_m);  \
     ST8x4_UB(tmp0_m, tmp1_m, pdst_m, stride);                 \
 }