Refactor Neon averaging subpel variance functions

Merge the computation of vpx_comp_avg_pred into the second pass of the
bilinear filter - avoiding the overhead of loading and storing the
entire block again.

This is a backport of this libaom change[1].

[1] https://aomedia-review.googlesource.com/c/aom/+/166961

Change-Id: I9327ff7382a46d50c42a5213a11379b957146372

1 files changed, 148 insertions, 40 deletions

diff --git a/vpx_dsp/arm/subpel_variance_neon.c b/vpx_dsp/arm/subpel_variance_neon.c
index 60650b703..237f7fad2 100644
--- a/vpx_dsp/arm/subpel_variance_neon.c
+++ b/vpx_dsp/arm/subpel_variance_neon.c
@@ -218,45 +218,153 @@ SPECIALIZED_SUBPEL_VARIANCE_WXH_NEON(32, 64, 1)
 SPECIALIZED_SUBPEL_VARIANCE_WXH_NEON(64, 32, 1)
 SPECIALIZED_SUBPEL_VARIANCE_WXH_NEON(64, 64, 1)
 
-// 4xM filter writes an extra row to fdata because it processes two rows at a
-// time.
-#define SUB_PIXEL_AVG_VARIANCENXM(n, m)                                       \
-  uint32_t vpx_sub_pixel_avg_variance##n##x##m##_neon(                        \
-      const uint8_t *src_ptr, int src_stride, int x_offset, int y_offset,     \
-      const uint8_t *ref_ptr, int ref_stride, uint32_t *sse,                  \
-      const uint8_t *second_pred) {                                           \
-    uint8_t temp0[n * (m + (n == 4 ? 2 : 1))];                                \
-    uint8_t temp1[n * m];                                                     \
-                                                                              \
-    if (n == 4) {                                                             \
-      var_filter_block2d_bil_w4(src_ptr, temp0, src_stride, 1, (m + 2),       \
-                                x_offset);                                    \
-      var_filter_block2d_bil_w4(temp0, temp1, n, n, m, y_offset);             \
-    } else if (n == 8) {                                                      \
-      var_filter_block2d_bil_w8(src_ptr, temp0, src_stride, 1, (m + 1),       \
-                                x_offset);                                    \
-      var_filter_block2d_bil_w8(temp0, temp1, n, n, m, y_offset);             \
-    } else {                                                                  \
-      var_filter_block2d_bil_large(src_ptr, temp0, src_stride, 1, n, (m + 1), \
-                                   x_offset);                                 \
-      var_filter_block2d_bil_large(temp0, temp1, n, n, n, m, y_offset);       \
-    }                                                                         \
-                                                                              \
-    vpx_comp_avg_pred(temp0, second_pred, n, m, temp1, n);                    \
-                                                                              \
-    return vpx_variance##n##x##m(temp0, n, ref_ptr, ref_stride, sse);         \
+// Combine bilinear filter with vpx_comp_avg_pred for blocks having width 4.
+static void avg_pred_var_filter_block2d_bil_w4(const uint8_t *src_ptr,
+                                               uint8_t *dst_ptr, int src_stride,
+                                               int pixel_step, int dst_height,
+                                               int filter_offset,
+                                               const uint8_t *second_pred) {
+  const uint8x8_t f0 = vdup_n_u8(8 - filter_offset);
+  const uint8x8_t f1 = vdup_n_u8(filter_offset);
+
+  int i = dst_height;
+  do {
+    uint8x8_t s0 = load_unaligned_u8(src_ptr, src_stride);
+    uint8x8_t s1 = load_unaligned_u8(src_ptr + pixel_step, src_stride);
+    uint16x8_t blend = vmlal_u8(vmull_u8(s0, f0), s1, f1);
+    uint8x8_t blend_u8 = vrshrn_n_u16(blend, 3);
+
+    uint8x8_t p = vld1_u8(second_pred);
+    uint8x8_t avg = vrhadd_u8(blend_u8, p);
+
+    vst1_u8(dst_ptr, avg);
+
+    src_ptr += 2 * src_stride;
+    dst_ptr += 2 * 4;
+    second_pred += 2 * 4;
+    i -= 2;
+  } while (i != 0);
+}
+
+// Combine bilinear filter with vpx_comp_avg_pred for blocks having width 8.
+static void avg_pred_var_filter_block2d_bil_w8(const uint8_t *src_ptr,
+                                               uint8_t *dst_ptr, int src_stride,
+                                               int pixel_step, int dst_height,
+                                               int filter_offset,
+                                               const uint8_t *second_pred) {
+  const uint8x8_t f0 = vdup_n_u8(8 - filter_offset);
+  const uint8x8_t f1 = vdup_n_u8(filter_offset);
+
+  int i = dst_height;
+  do {
+    uint8x8_t s0 = vld1_u8(src_ptr);
+    uint8x8_t s1 = vld1_u8(src_ptr + pixel_step);
+    uint16x8_t blend = vmlal_u8(vmull_u8(s0, f0), s1, f1);
+    uint8x8_t blend_u8 = vrshrn_n_u16(blend, 3);
+
+    uint8x8_t p = vld1_u8(second_pred);
+    uint8x8_t avg = vrhadd_u8(blend_u8, p);
+
+    vst1_u8(dst_ptr, avg);
+
+    src_ptr += src_stride;
+    dst_ptr += 8;
+    second_pred += 8;
+  } while (--i > 0);
+}
+
+// Combine bilinear filter with vpx_comp_avg_pred for large blocks.
+static void avg_pred_var_filter_block2d_bil_large(
+    const uint8_t *src_ptr, uint8_t *dst_ptr, int src_stride, int pixel_step,
+    int dst_width, int dst_height, int filter_offset,
+    const uint8_t *second_pred) {
+  const uint8x8_t f0 = vdup_n_u8(8 - filter_offset);
+  const uint8x8_t f1 = vdup_n_u8(filter_offset);
+
+  int i = dst_height;
+  do {
+    int j = 0;
+    do {
+      uint8x16_t s0 = vld1q_u8(src_ptr + j);
+      uint8x16_t s1 = vld1q_u8(src_ptr + j + pixel_step);
+      uint16x8_t blend_l =
+          vmlal_u8(vmull_u8(vget_low_u8(s0), f0), vget_low_u8(s1), f1);
+      uint16x8_t blend_h =
+          vmlal_u8(vmull_u8(vget_high_u8(s0), f0), vget_high_u8(s1), f1);
+      uint8x16_t blend_u8 =
+          vcombine_u8(vrshrn_n_u16(blend_l, 3), vrshrn_n_u16(blend_h, 3));
+
+      uint8x16_t p = vld1q_u8(second_pred);
+      uint8x16_t avg = vrhaddq_u8(blend_u8, p);
+
+      vst1q_u8(dst_ptr + j, avg);
+
+      j += 16;
+      second_pred += 16;
+    } while (j < dst_width);
+
+    src_ptr += src_stride;
+    dst_ptr += dst_width;
+  } while (--i != 0);
+}
+
+// Combine bilinear filter with vpx_comp_avg_pred for blocks having width 16.
+static void avg_pred_var_filter_block2d_bil_w16(
+    const uint8_t *src_ptr, uint8_t *dst_ptr, int src_stride, int pixel_step,
+    int dst_height, int filter_offset, const uint8_t *second_pred) {
+  avg_pred_var_filter_block2d_bil_large(src_ptr, dst_ptr, src_stride,
+                                        pixel_step, 16, dst_height,
+                                        filter_offset, second_pred);
+}
+
+// Combine bilinear filter with vpx_comp_avg_pred for blocks having width 32.
+static void avg_pred_var_filter_block2d_bil_w32(
+    const uint8_t *src_ptr, uint8_t *dst_ptr, int src_stride, int pixel_step,
+    int dst_height, int filter_offset, const uint8_t *second_pred) {
+  avg_pred_var_filter_block2d_bil_large(src_ptr, dst_ptr, src_stride,
+                                        pixel_step, 32, dst_height,
+                                        filter_offset, second_pred);
+}
+
+// Combine bilinear filter with vpx_comp_avg_pred for blocks having width 64.
+static void avg_pred_var_filter_block2d_bil_w64(
+    const uint8_t *src_ptr, uint8_t *dst_ptr, int src_stride, int pixel_step,
+    int dst_height, int filter_offset, const uint8_t *second_pred) {
+  avg_pred_var_filter_block2d_bil_large(src_ptr, dst_ptr, src_stride,
+                                        pixel_step, 64, dst_height,
+                                        filter_offset, second_pred);
+}
+
+#define SUBPEL_AVG_VARIANCE_WXH_NEON(w, h, padding)                         \
+  unsigned int vpx_sub_pixel_avg_variance##w##x##h##_neon(                  \
+      const uint8_t *src, int source_stride, int xoffset, int yoffset,      \
+      const uint8_t *ref, int ref_stride, uint32_t *sse,                    \
+      const uint8_t *second_pred) {                                         \
+    uint8_t tmp0[w * (h + padding)];                                        \
+    uint8_t tmp1[w * h];                                                    \
+    var_filter_block2d_bil_w##w(src, tmp0, source_stride, 1, (h + padding), \
+                                xoffset);                                   \
+    avg_pred_var_filter_block2d_bil_w##w(tmp0, tmp1, w, w, h, yoffset,      \
+                                         second_pred);                      \
+    return vpx_variance##w##x##h(tmp1, w, ref, ref_stride, sse);            \
   }
 
-SUB_PIXEL_AVG_VARIANCENXM(4, 4)
-SUB_PIXEL_AVG_VARIANCENXM(4, 8)
-SUB_PIXEL_AVG_VARIANCENXM(8, 4)
-SUB_PIXEL_AVG_VARIANCENXM(8, 8)
-SUB_PIXEL_AVG_VARIANCENXM(8, 16)
-SUB_PIXEL_AVG_VARIANCENXM(16, 8)
-SUB_PIXEL_AVG_VARIANCENXM(16, 16)
-SUB_PIXEL_AVG_VARIANCENXM(16, 32)
-SUB_PIXEL_AVG_VARIANCENXM(32, 16)
-SUB_PIXEL_AVG_VARIANCENXM(32, 32)
-SUB_PIXEL_AVG_VARIANCENXM(32, 64)
-SUB_PIXEL_AVG_VARIANCENXM(64, 32)
-SUB_PIXEL_AVG_VARIANCENXM(64, 64)
+// 4x<h> blocks are processed two rows at a time, so require an extra row of
+// padding.
+SUBPEL_AVG_VARIANCE_WXH_NEON(4, 4, 2)
+SUBPEL_AVG_VARIANCE_WXH_NEON(4, 8, 2)
+
+SUBPEL_AVG_VARIANCE_WXH_NEON(8, 4, 1)
+SUBPEL_AVG_VARIANCE_WXH_NEON(8, 8, 1)
+SUBPEL_AVG_VARIANCE_WXH_NEON(8, 16, 1)
+
+SUBPEL_AVG_VARIANCE_WXH_NEON(16, 8, 1)
+SUBPEL_AVG_VARIANCE_WXH_NEON(16, 16, 1)
+SUBPEL_AVG_VARIANCE_WXH_NEON(16, 32, 1)
+
+SUBPEL_AVG_VARIANCE_WXH_NEON(32, 16, 1)
+SUBPEL_AVG_VARIANCE_WXH_NEON(32, 32, 1)
+SUBPEL_AVG_VARIANCE_WXH_NEON(32, 64, 1)
+
+SUBPEL_AVG_VARIANCE_WXH_NEON(64, 32, 1)
+SUBPEL_AVG_VARIANCE_WXH_NEON(64, 64, 1)