Add a config flag/code for fast sgr computation

Adds an experiment for fast sgr computation where for the r=2
filter, computation of the A, B stats are computed for every
other row and averaged in between.
The motivation is to improve software performance with hopefully
minimal loss.

Change-Id: Ie36687826524dc18c1fbb7f6becff244187bf8da

av1/encoder/pickrst.c

@@ -367,8 +367,13 @@ static void sgr_filter_block(const sgr_params_type *params, const uint8_t *dat8,
                              int width, int height, int dat_stride,
                              int use_highbd, int bit_depth, int32_t *flt1,
                              int32_t *flt2, int flt_stride) {
+#if CONFIG_FAST_SGR
+  av1_selfguided_restoration_c(dat8, width, height, dat_stride, flt1, flt2,
+                               flt_stride, params, bit_depth, use_highbd);
+#else
   av1_selfguided_restoration(dat8, width, height, dat_stride, flt1, flt2,
                              flt_stride, params, bit_depth, use_highbd);
+#endif  // CONFIG_FAST_SGR
 }
 
 // Apply the self-guided filter across an entire restoration unit.

build/cmake/aom_config_defaults.cmake

@@ -121,6 +121,7 @@ set(CONFIG_EXT_QM 0 CACHE NUMBER "AV1 experiment flag.")
 set(CONFIG_EXT_SKIP 1 CACHE NUMBER "AV1 experiment flag.")
 set(CONFIG_EXT_TILE 0 CACHE NUMBER "AV1 experiment flag.")
 set(CONFIG_EXT_WARPED_MOTION 1 CACHE NUMBER "AV1 experiment flag.")
+set(CONFIG_FAST_SGR 0 CACHE NUMBER "AV1 experiment flag.")
 set(CONFIG_FILTER_INTRA 1 CACHE NUMBER "AV1 experiment flag.")
 set(CONFIG_FP_MB_STATS 0 CACHE NUMBER "AV1 experiment flag.")
 set(CONFIG_FRAME_MARKER 1 CACHE NUMBER "AV1 experiment flag.")