mips msa vp9 convolve8 avg horiz optimization

average improvement ~5x-8x

Change-Id: I179a69ec620fbd69979bd128f05d18113618aab4

test/convolve_test.cc

@@ -1818,7 +1818,7 @@ INSTANTIATE_TEST_CASE_P(DSPR2, ConvolveTest, ::testing::Values(
 #if HAVE_MSA
 const ConvolveFunctions convolve8_msa(
     vp9_convolve_copy_msa, vp9_convolve_avg_msa,
-    vp9_convolve8_horiz_msa, vp9_convolve8_avg_horiz_c,
+    vp9_convolve8_horiz_msa, vp9_convolve8_avg_horiz_msa,
     vp9_convolve8_vert_msa, vp9_convolve8_avg_vert_msa,
     vp9_convolve8_msa, vp9_convolve8_avg_c, 0);
 

vp9/common/vp9_rtcd_defs.pl

@@ -304,7 +304,7 @@ add_proto qw/void vp9_convolve8_avg/, "const uint8_t *src, ptrdiff_t src_stride,
 specialize qw/vp9_convolve8_avg sse2 ssse3 neon dspr2/;
 
 add_proto qw/void vp9_convolve8_avg_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h";
-specialize qw/vp9_convolve8_avg_horiz sse2 ssse3 neon dspr2/;
+specialize qw/vp9_convolve8_avg_horiz sse2 ssse3 neon dspr2 msa/;
 
 add_proto qw/void vp9_convolve8_avg_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h";
 specialize qw/vp9_convolve8_avg_vert sse2 ssse3 neon dspr2 msa/;

vp9/vp9_common.mk

@@ -132,6 +132,7 @@ VP9_COMMON_SRCS-$(HAVE_DSPR2)  += common/mips/dspr2/vp9_mblpf_vert_loopfilter_ds
 
 # common (msa)
 VP9_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/vp9_macros_msa.h
+VP9_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/vp9_convolve8_avg_horiz_msa.c
 VP9_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/vp9_convolve8_avg_vert_msa.c
 VP9_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/vp9_convolve8_horiz_msa.c
 VP9_COMMON_SRCS-$(HAVE_MSA) += common/mips/msa/vp9_convolve8_msa.c