SSE2 optimizations for _6/_16 lowbd lpf functions

Includes vertical and horizontal implementations
and to fix 5/13 TAPs/Parallel deblocking support.

Re-working internals of the filters for better
re-usage across different sizes.

Tests are enabled.

Performance changes, SSE2 over C:
Horizontal methods: up to    3-4x
Vertical   methods: up to 1.5x-2x

Change-Id: I2e36035355d8c23c1d4b0d59d0e23f598e9d0e3f

aom_dsp/aom_dsp_rtcd_defs.pl

@@ -444,6 +444,11 @@ if (aom_config("CONFIG_PARALLEL_DEBLOCKING") ne "yes") {
   $aom_lpf_vertical_16_dual_neon_asm=aom_lpf_vertical_16_dual_neon;
 }
 
+add_proto qw/void aom_lpf_vertical_6/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
+if (aom_config("CONFIG_PARALLEL_DEBLOCKING") eq "yes") {
+  specialize qw/aom_lpf_vertical_6 sse2/;
+}
+
 add_proto qw/void aom_lpf_vertical_8/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
 if (aom_config("CONFIG_PARALLEL_DEBLOCKING") eq "yes") {
   specialize qw/aom_lpf_vertical_8 sse2/;
@@ -485,6 +490,11 @@ if (aom_config("CONFIG_PARALLEL_DEBLOCKING") eq "yes") {
   $aom_lpf_horizontal_16_dual_neon_asm=aom_lpf_horizontal_16_dual_neon;
 }
 
+add_proto qw/void aom_lpf_horizontal_6/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
+if (aom_config("CONFIG_PARALLEL_DEBLOCKING") eq "yes") {
+  specialize qw/aom_lpf_horizontal_6 sse2/;
+}
+
 add_proto qw/void aom_lpf_horizontal_8/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
 if (aom_config("CONFIG_PARALLEL_DEBLOCKING") eq "yes") {
   specialize qw/aom_lpf_horizontal_8 sse2/;

av1/common/av1_loopfilter.c

@@ -42,12 +42,6 @@ static const int delta_lf_id_lut[MAX_MB_PLANE][2] = {
 #endif
 
 #if PARALLEL_DEBLOCKING_5_TAP_CHROMA
-extern void aom_lpf_vertical_6_c(uint8_t *s, int pitch, const uint8_t *blimit,
-                                 const uint8_t *limit, const uint8_t *thresh);
-
-extern void aom_lpf_horizontal_6_c(uint8_t *s, int p, const uint8_t *blimit,
-                                   const uint8_t *limit, const uint8_t *thresh);
-
 extern void aom_highbd_lpf_horizontal_6_c(uint16_t *s, int p,
                                           const uint8_t *blimit,
                                           const uint8_t *limit,
@@ -2233,8 +2227,8 @@ static void av1_filter_block_plane_vert(
                                         params.mblim, params.lim,
                                         params.hev_thr, cm->bit_depth);
           else
-            aom_lpf_vertical_6_c(p, dst_stride, params.mblim, params.lim,
-                                 params.hev_thr);
+            aom_lpf_vertical_6(p, dst_stride, params.mblim, params.lim,
+                               params.hev_thr);
           break;
 #endif
         // apply 8-tap filtering
@@ -2261,13 +2255,8 @@ static void av1_filter_block_plane_vert(
                                        cm->bit_depth);
 #endif
           else
-#if CONFIG_DEBLOCK_13TAP
-            aom_lpf_vertical_16_c(p, dst_stride, params.mblim, params.lim,
-                                  params.hev_thr);
-#else
             aom_lpf_vertical_16(p, dst_stride, params.mblim, params.lim,
                                 params.hev_thr);
-#endif
           break;
         // no filtering
         default: break;
@@ -2335,8 +2324,8 @@ static void av1_filter_block_plane_horz(
                                           params.mblim, params.lim,
                                           params.hev_thr, cm->bit_depth);
           else
-            aom_lpf_horizontal_6_c(p, dst_stride, params.mblim, params.lim,
-                                   params.hev_thr);
+            aom_lpf_horizontal_6(p, dst_stride, params.mblim, params.lim,
+                                 params.hev_thr);
           break;
 #endif
         // apply 8-tap filtering