Add compute_stats SSE4_1 and AVX2 code
1. add compute_stats unittest
2. add compute_stats SSE4_1
3. add compute_stats AVX2
Encoder speedup about 0.9% without rd change
test sequence: BasketballDrill_832x480_50.y4m
test command line:./aomenc --cpu-used=1 --psnr -D \
-q --end-usage=vbr --target-bitrate=800 --limit=20 \
BasketballDrill_832x480_50.y4m -otest.webm
Change-Id: Ic0799997c1075a139869b45ba84af00c5475964a
diff --git a/av1/common/av1_rtcd_defs.pl b/av1/common/av1_rtcd_defs.pl
index fa8b349..c1eb1d0 100755
--- a/av1/common/av1_rtcd_defs.pl
+++ b/av1/common/av1_rtcd_defs.pl
@@ -76,12 +76,12 @@
specialize qw/av1_highbd_wiener_convolve_add_src ssse3/;
specialize qw/av1_highbd_wiener_convolve_add_src avx2/;
+
# directional intra predictor functions
add_proto qw/void av1_dr_prediction_z1/, "uint8_t *dst, ptrdiff_t stride, int bw, int bh, const uint8_t *above, const uint8_t *left, int upsample_above, int dx, int dy";
add_proto qw/void av1_dr_prediction_z2/, "uint8_t *dst, ptrdiff_t stride, int bw, int bh, const uint8_t *above, const uint8_t *left, int upsample_above, int upsample_left, int dx, int dy";
add_proto qw/void av1_dr_prediction_z3/, "uint8_t *dst, ptrdiff_t stride, int bw, int bh, const uint8_t *above, const uint8_t *left, int upsample_left, int dx, int dy";
-
# FILTER_INTRA predictor functions
add_proto qw/void av1_filter_intra_predictor/, "uint8_t *dst, ptrdiff_t stride, TX_SIZE tx_size, const uint8_t *above, const uint8_t *left, int mode";
specialize qw/av1_filter_intra_predictor sse4_1/;
@@ -251,6 +251,8 @@
add_proto qw/uint32_t av1_get_crc32c_value/, "void *crc_calculator, uint8_t *p, int length";
specialize qw/av1_get_crc32c_value sse4_2/;
+ add_proto qw/void compute_stats/, "int wiener_win, const uint8_t *dgd8, const uint8_t *src8, int h_start, int h_end, int v_start, int v_end, int dgd_stride, int src_stride, double *M, double *H";
+ specialize qw/compute_stats sse4_1 avx2/;
}
# end encoder functions
