libvpx: Pull from upstream

source/libvpx/vp9/encoder/vp9_aq_variance.c

 /*
  *  Copyright (c) 2013 The WebM project authors. All Rights Reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS.  All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
@@ skipping 64 matching lines @@
       if (rate_ratio[i] == 1.0) {
         continue;
       }
 
       vp9_set_segdata(seg, i, SEG_LVL_ALT_Q, qindex_delta);
       vp9_enable_segfeature(seg, i, SEG_LVL_ALT_Q);
     }
   }
 }
 
+/* TODO(agrange, paulwilkins): The block_variance calls the unoptimized versions
+ * of variance() and highbd_8_variance(). It should not.
+ */
+static void aq_variance(const uint8_t *a, int  a_stride,
+                        const uint8_t *b, int  b_stride,
+                        int  w, int  h, unsigned int *sse, int *sum) {
+  int i, j;
+
+  *sum = 0;
+  *sse = 0;
+
+  for (i = 0; i < h; i++) {
+    for (j = 0; j < w; j++) {
+      const int diff = a[j] - b[j];
+      *sum += diff;
+      *sse += diff * diff;
+    }
+
+    a += a_stride;
+    b += b_stride;
+  }
+}
+
+#if CONFIG_VP9_HIGHBITDEPTH
+static void aq_highbd_variance64(const uint8_t *a8, int  a_stride,
+                                 const uint8_t *b8, int  b_stride,
+                                 int w, int h, uint64_t *sse, uint64_t *sum) {
+  int i, j;
+
+  uint16_t *a = CONVERT_TO_SHORTPTR(a8);
+  uint16_t *b = CONVERT_TO_SHORTPTR(b8);
+  *sum = 0;
+  *sse = 0;
+
+  for (i = 0; i < h; i++) {
+    for (j = 0; j < w; j++) {
+      const int diff = a[j] - b[j];
+      *sum += diff;
+      *sse += diff * diff;
+    }
+    a += a_stride;
+    b += b_stride;
+  }
+}
+
+static void aq_highbd_8_variance(const uint8_t *a8, int  a_stride,
+                                 const uint8_t *b8, int  b_stride,
+                                 int w, int h, unsigned int *sse, int *sum) {
+  uint64_t sse_long = 0;
+  uint64_t sum_long = 0;
+  aq_highbd_variance64(a8, a_stride, b8, b_stride, w, h, &sse_long, &sum_long);
+  *sse = (unsigned int)sse_long;
+  *sum = (int)sum_long;
+}
+#endif  // CONFIG_VP9_HIGHBITDEPTH
 
 static unsigned int block_variance(VP9_COMP *cpi, MACROBLOCK *x,
                                    BLOCK_SIZE bs) {
   MACROBLOCKD *xd = &x->e_mbd;
   unsigned int var, sse;
   int right_overflow = (xd->mb_to_right_edge < 0) ?
       ((-xd->mb_to_right_edge) >> 3) : 0;
   int bottom_overflow = (xd->mb_to_bottom_edge < 0) ?
       ((-xd->mb_to_bottom_edge) >> 3) : 0;
 
   if (right_overflow || bottom_overflow) {
     const int bw = 8 * num_8x8_blocks_wide_lookup[bs] - right_overflow;
     const int bh = 8 * num_8x8_blocks_high_lookup[bs] - bottom_overflow;
     int avg;
 #if CONFIG_VP9_HIGHBITDEPTH
     if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
-      highbd_8_variance(x->plane[0].src.buf, x->plane[0].src.stride,
-                        CONVERT_TO_BYTEPTR(vp9_highbd_64_zeros), 0, bw, bh,
-                        &sse, &avg);
+      aq_highbd_8_variance(x->plane[0].src.buf, x->plane[0].src.stride,
+                           CONVERT_TO_BYTEPTR(vp9_highbd_64_zeros), 0, bw, bh,
+                           &sse, &avg);
       sse >>= 2 * (xd->bd - 8);
       avg >>= (xd->bd - 8);
     } else {
-      variance(x->plane[0].src.buf, x->plane[0].src.stride,
-               vp9_64_zeros, 0, bw, bh, &sse, &avg);
+      aq_variance(x->plane[0].src.buf, x->plane[0].src.stride,
+                  vp9_64_zeros, 0, bw, bh, &sse, &avg);
     }
 #else
-    variance(x->plane[0].src.buf, x->plane[0].src.stride,
-             vp9_64_zeros, 0, bw, bh, &sse, &avg);
+    aq_variance(x->plane[0].src.buf, x->plane[0].src.stride,
+                vp9_64_zeros, 0, bw, bh, &sse, &avg);
 #endif  // CONFIG_VP9_HIGHBITDEPTH
     var = sse - (((int64_t)avg * avg) / (bw * bh));
     return (256 * var) / (bw * bh);
   } else {
 #if CONFIG_VP9_HIGHBITDEPTH
     if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
       var = cpi->fn_ptr[bs].vf(x->plane[0].src.buf,
                                x->plane[0].src.stride,
                                CONVERT_TO_BYTEPTR(vp9_highbd_64_zeros),
                                0, &sse);
@@ skipping 20 matching lines @@
 #define DEFAULT_E_MIDPOINT 10.0
 int vp9_block_energy(VP9_COMP *cpi, MACROBLOCK *x, BLOCK_SIZE bs) {
   double energy;
   double energy_midpoint;
   vp9_clear_system_state();
   energy_midpoint =
     (cpi->oxcf.pass == 2) ? cpi->twopass.mb_av_energy : DEFAULT_E_MIDPOINT;
   energy = vp9_log_block_var(cpi, x, bs) - energy_midpoint;
   return clamp((int)round(energy), ENERGY_MIN, ENERGY_MAX);
 }