libvpx: Pull from upstream

source/libvpx/vpx_scale/generic/yv12extend.c

Index: source/libvpx/vpx_scale/generic/yv12extend.c
===================================================================
--- source/libvpx/vpx_scale/generic/yv12extend.c	(revision 291857)
+++ source/libvpx/vpx_scale/generic/yv12extend.c	(working copy)
@@ -13,6 +13,9 @@
 #include "vpx/vpx_integer.h"
 #include "vpx_mem/vpx_mem.h"
 #include "vpx_scale/yv12config.h"
+#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
+#include "vp9/common/vp9_common.h"
+#endif
 
 static void extend_plane(uint8_t *const src, int src_stride,
                          int width, int height,
@@ -55,6 +58,50 @@
   }
 }
 
+#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
+static void extend_plane_high(uint8_t *const src8, int src_stride,
+                              int width, int height,
+                              int extend_top, int extend_left,
+                              int extend_bottom, int extend_right) {
+  int i;
+  const int linesize = extend_left + extend_right + width;
+  uint16_t *src = CONVERT_TO_SHORTPTR(src8);
+
+  /* copy the left and right most columns out */
+  uint16_t *src_ptr1 = src;
+  uint16_t *src_ptr2 = src + width - 1;
+  uint16_t *dst_ptr1 = src - extend_left;
+  uint16_t *dst_ptr2 = src + width;
+
+  for (i = 0; i < height; ++i) {
+    vpx_memset16(dst_ptr1, src_ptr1[0], extend_left);
+    vpx_memset16(dst_ptr2, src_ptr2[0], extend_right);
+    src_ptr1 += src_stride;
+    src_ptr2 += src_stride;
+    dst_ptr1 += src_stride;
+    dst_ptr2 += src_stride;
+  }
+
+  /* Now copy the top and bottom lines into each line of the respective
+   * borders
+   */
+  src_ptr1 = src - extend_left;
+  src_ptr2 = src + src_stride * (height - 1) - extend_left;
+  dst_ptr1 = src + src_stride * -extend_top - extend_left;
+  dst_ptr2 = src + src_stride * height - extend_left;
+
+  for (i = 0; i < extend_top; ++i) {
+    vpx_memcpy(dst_ptr1, src_ptr1, linesize * sizeof(uint16_t));
+    dst_ptr1 += src_stride;
+  }
+
+  for (i = 0; i < extend_bottom; ++i) {
+    vpx_memcpy(dst_ptr2, src_ptr2, linesize * sizeof(uint16_t));
+    dst_ptr2 += src_stride;
+  }
+}
+#endif
+
 void vp8_yv12_extend_frame_borders_c(YV12_BUFFER_CONFIG *ybf) {
   const int uv_border = ybf->border / 2;
 
@@ -64,6 +111,31 @@
   assert(ybf->y_height - ybf->y_crop_height >= 0);
   assert(ybf->y_width - ybf->y_crop_width >= 0);
 
+#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
+  if (ybf->flags & YV12_FLAG_HIGHBITDEPTH) {
+    extend_plane_high(
+        ybf->y_buffer, ybf->y_stride,
+        ybf->y_crop_width, ybf->y_crop_height,
+        ybf->border, ybf->border,
+        ybf->border + ybf->y_height - ybf->y_crop_height,
+        ybf->border + ybf->y_width - ybf->y_crop_width);
+
+    extend_plane_high(
+        ybf->u_buffer, ybf->uv_stride,
+        (ybf->y_crop_width + 1) / 2, (ybf->y_crop_height + 1) / 2,
+        ybf->border / 2, ybf->border / 2,
+        (ybf->border + ybf->y_height - ybf->y_crop_height + 1) / 2,
+        (ybf->border + ybf->y_width - ybf->y_crop_width + 1) / 2);
+
+    extend_plane_high(
+        ybf->v_buffer, ybf->uv_stride,
+        (ybf->y_crop_width + 1) / 2, (ybf->y_crop_height + 1) / 2,
+        ybf->border / 2, ybf->border / 2,
+        (ybf->border + ybf->y_height - ybf->y_crop_height + 1) / 2,
+        (ybf->border + ybf->y_width - ybf->y_crop_width + 1) / 2);
+    return;
+  }
+#endif
   extend_plane(ybf->y_buffer, ybf->y_stride,
                ybf->y_crop_width, ybf->y_crop_height,
                ybf->border, ybf->border,
@@ -99,6 +171,20 @@
   assert(ybf->y_height - ybf->y_crop_height >= 0);
   assert(ybf->y_width - ybf->y_crop_width >= 0);
 
+#if CONFIG_VP9_HIGHBITDEPTH
+  if (ybf->flags & YV12_FLAG_HIGHBITDEPTH) {
+    extend_plane_high(ybf->y_buffer, ybf->y_stride,
+                      ybf->y_crop_width, ybf->y_crop_height,
+                      ext_size, ext_size,
+                      ext_size + ybf->y_height - ybf->y_crop_height,
+                      ext_size + ybf->y_width - ybf->y_crop_width);
+    extend_plane_high(ybf->u_buffer, ybf->uv_stride,
+                      c_w, c_h, c_et, c_el, c_eb, c_er);
+    extend_plane_high(ybf->v_buffer, ybf->uv_stride,
+                      c_w, c_h, c_et, c_el, c_eb, c_er);
+    return;
+  }
+#endif
   extend_plane(ybf->y_buffer, ybf->y_stride,
                ybf->y_crop_width, ybf->y_crop_height,
                ext_size, ext_size,
@@ -121,6 +207,14 @@
                        VP9INNERBORDERINPIXELS : ybf->border;
   extend_frame(ybf, inner_bw);
 }
+
+#if CONFIG_VP9_HIGHBITDEPTH
+void memcpy_short_addr(uint8_t *dst8, const uint8_t *src8, int num) {
+  uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
+  uint16_t *src = CONVERT_TO_SHORTPTR(src8);
+  vpx_memcpy(dst, src, num * sizeof(uint16_t));
+}
+#endif  // CONFIG_VP9_HIGHBITDEPTH
 #endif  // CONFIG_VP9
 
 // Copies the source image into the destination image and updates the
@@ -140,6 +234,40 @@
   assert(src_ybc->y_height == dst_ybc->y_height);
 #endif
 
+#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
+  if (src_ybc->flags & YV12_FLAG_HIGHBITDEPTH) {
+    assert(dst_ybc->flags & YV12_FLAG_HIGHBITDEPTH);
+    for (row = 0; row < src_ybc->y_height; ++row) {
+      memcpy_short_addr(dst, src, src_ybc->y_width);
+      src += src_ybc->y_stride;
+      dst += dst_ybc->y_stride;
+    }
+
+    src = src_ybc->u_buffer;
+    dst = dst_ybc->u_buffer;
+
+    for (row = 0; row < src_ybc->uv_height; ++row) {
+      memcpy_short_addr(dst, src, src_ybc->uv_width);
+      src += src_ybc->uv_stride;
+      dst += dst_ybc->uv_stride;
+    }
+
+    src = src_ybc->v_buffer;
+    dst = dst_ybc->v_buffer;
+
+    for (row = 0; row < src_ybc->uv_height; ++row) {
+      memcpy_short_addr(dst, src, src_ybc->uv_width);
+      src += src_ybc->uv_stride;
+      dst += dst_ybc->uv_stride;
+    }
+
+    vp8_yv12_extend_frame_borders_c(dst_ybc);
+    return;
+  } else {
+    assert(!(dst_ybc->flags & YV12_FLAG_HIGHBITDEPTH));
+  }
+#endif
+
   for (row = 0; row < src_ybc->y_height; ++row) {
     vpx_memcpy(dst, src, src_ybc->y_width);
     src += src_ybc->y_stride;
@@ -173,6 +301,19 @@
   const uint8_t *src = src_ybc->y_buffer;
   uint8_t *dst = dst_ybc->y_buffer;
 
+#if CONFIG_VP9 && CONFIG_VP9_HIGHBITDEPTH
+  if (src_ybc->flags & YV12_FLAG_HIGHBITDEPTH) {
+    const uint16_t *src16 = CONVERT_TO_SHORTPTR(src);
+    uint16_t *dst16 = CONVERT_TO_SHORTPTR(dst);
+    for (row = 0; row < src_ybc->y_height; ++row) {
+      vpx_memcpy(dst16, src16, src_ybc->y_width * sizeof(uint16_t));
+      src16 += src_ybc->y_stride;
+      dst16 += dst_ybc->y_stride;
+    }
+    return;
+  }
+#endif
+
   for (row = 0; row < src_ybc->y_height; ++row) {
     vpx_memcpy(dst, src, src_ybc->y_width);
     src += src_ybc->y_stride;