Add support for odd width and height in I410ToI420

Bug: libyuv:950
Change-Id: Ic9a094463af875aefd927023f730b5f35f8551de
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/4154630
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: Frank Barchard <fbarchard@chromium.org>

1 files changed, 82 insertions, 3 deletions

diff --git a/source/scale_common.cc b/source/scale_common.cc
index 55749a2e..f1832403 100644
--- a/source/scale_common.cc
+++ b/source/scale_common.cc
@@ -101,6 +101,30 @@ void ScaleRowDown2_16To8_C(const uint16_t* src_ptr,
   }
 }
 
+void ScaleRowDown2_16To8_Odd_C(const uint16_t* src_ptr,
+                               ptrdiff_t src_stride,
+                               uint8_t* dst,
+                               int dst_width,
+                               int scale) {
+  int x;
+  (void)src_stride;
+  assert(scale >= 256);
+  assert(scale <= 32768);
+  dst_width -= 1;
+  for (x = 0; x < dst_width - 1; x += 2) {
+    dst[0] = STATIC_CAST(uint8_t, C16TO8(src_ptr[1], scale));
+    dst[1] = STATIC_CAST(uint8_t, C16TO8(src_ptr[3], scale));
+    dst += 2;
+    src_ptr += 4;
+  }
+  if (dst_width & 1) {
+    dst[0] = STATIC_CAST(uint8_t, C16TO8(src_ptr[1], scale));
+    dst += 1;
+    src_ptr += 2;
+  }
+  dst[0] = STATIC_CAST(uint8_t, C16TO8(src_ptr[0], scale));
+}
+
 void ScaleRowDown2Linear_C(const uint8_t* src_ptr,
                            ptrdiff_t src_stride,
                            uint8_t* dst,
@@ -158,6 +182,31 @@ void ScaleRowDown2Linear_16To8_C(const uint16_t* src_ptr,
   }
 }
 
+void ScaleRowDown2Linear_16To8_Odd_C(const uint16_t* src_ptr,
+                                     ptrdiff_t src_stride,
+                                     uint8_t* dst,
+                                     int dst_width,
+                                     int scale) {
+  const uint16_t* s = src_ptr;
+  int x;
+  (void)src_stride;
+  assert(scale >= 256);
+  assert(scale <= 32768);
+  dst_width -= 1;
+  for (x = 0; x < dst_width - 1; x += 2) {
+    dst[0] = STATIC_CAST(uint8_t, C16TO8((s[0] + s[1] + 1) >> 1, scale));
+    dst[1] = STATIC_CAST(uint8_t, C16TO8((s[2] + s[3] + 1) >> 1, scale));
+    dst += 2;
+    s += 4;
+  }
+  if (dst_width & 1) {
+    dst[0] = STATIC_CAST(uint8_t, C16TO8((s[0] + s[1] + 1) >> 1, scale));
+    dst += 1;
+    s += 2;
+  }
+  dst[0] = STATIC_CAST(uint8_t, C16TO8(s[0], scale));
+}
+
 void ScaleRowDown2Box_C(const uint8_t* src_ptr,
                         ptrdiff_t src_stride,
                         uint8_t* dst,
@@ -245,6 +294,36 @@ void ScaleRowDown2Box_16To8_C(const uint16_t* src_ptr,
   }
 }
 
+void ScaleRowDown2Box_16To8_Odd_C(const uint16_t* src_ptr,
+                                  ptrdiff_t src_stride,
+                                  uint8_t* dst,
+                                  int dst_width,
+                                  int scale) {
+  const uint16_t* s = src_ptr;
+  const uint16_t* t = src_ptr + src_stride;
+  int x;
+  assert(scale >= 256);
+  assert(scale <= 32768);
+  dst_width -= 1;
+  for (x = 0; x < dst_width - 1; x += 2) {
+    dst[0] = STATIC_CAST(uint8_t,
+                         C16TO8((s[0] + s[1] + t[0] + t[1] + 2) >> 2, scale));
+    dst[1] = STATIC_CAST(uint8_t,
+                         C16TO8((s[2] + s[3] + t[2] + t[3] + 2) >> 2, scale));
+    dst += 2;
+    s += 4;
+    t += 4;
+  }
+  if (dst_width & 1) {
+    dst[0] = STATIC_CAST(uint8_t,
+                         C16TO8((s[0] + s[1] + t[0] + t[1] + 2) >> 2, scale));
+    dst += 1;
+    s += 2;
+    t += 2;
+  }
+  dst[0] = STATIC_CAST(uint8_t, C16TO8((s[0] + t[0] + 1) >> 1, scale));
+}
+
 void ScaleRowDown4_C(const uint8_t* src_ptr,
                      ptrdiff_t src_stride,
                      uint8_t* dst,
@@ -1554,7 +1633,7 @@ void ScalePlaneVertical(int src_height,
                         enum FilterMode filtering) {
   // TODO(fbarchard): Allow higher bpp.
   int dst_width_bytes = dst_width * bpp;
-  void (*InterpolateRow)(uint8_t* dst_argb, const uint8_t* src_argb,
+  void (*InterpolateRow)(uint8_t * dst_argb, const uint8_t* src_argb,
                          ptrdiff_t src_stride, int dst_width,
                          int source_y_fraction) = InterpolateRow_C;
   const int max_y = (src_height > 1) ? ((src_height - 1) << 16) - 1 : 0;
@@ -1633,7 +1712,7 @@ void ScalePlaneVertical_16(int src_height,
                            enum FilterMode filtering) {
   // TODO(fbarchard): Allow higher wpp.
   int dst_width_words = dst_width * wpp;
-  void (*InterpolateRow)(uint16_t* dst_argb, const uint16_t* src_argb,
+  void (*InterpolateRow)(uint16_t * dst_argb, const uint16_t* src_argb,
                          ptrdiff_t src_stride, int dst_width,
                          int source_y_fraction) = InterpolateRow_16_C;
   const int max_y = (src_height > 1) ? ((src_height - 1) << 16) - 1 : 0;
@@ -1712,7 +1791,7 @@ void ScalePlaneVertical_16To8(int src_height,
   // TODO(fbarchard): Allow higher wpp.
   int dst_width_words = dst_width * wpp;
   // TODO(https://crbug.com/libyuv/931): Add NEON 32 bit and AVX2 versions.
-  void (*InterpolateRow_16To8)(uint8_t* dst_argb, const uint16_t* src_argb,
+  void (*InterpolateRow_16To8)(uint8_t * dst_argb, const uint16_t* src_argb,
                                ptrdiff_t src_stride, int scale, int dst_width,
                                int source_y_fraction) = InterpolateRow_16To8_C;
   const int max_y = (src_height > 1) ? ((src_height - 1) << 16) - 1 : 0;