ARGBAffineRow_C for non-SSE2 machine.

BUG=62
TEST=TestAffine unittest added to planar_test
Review URL: https://webrtc-codereview.appspot.com/731004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@315 16f28f9a-4ce2-e073-06de-1de4eb20be90

include/libyuv/planar_functions.h

@@ -228,6 +228,13 @@ int ARGBInterpolate(const uint8* src_argb0, int src_stride_argb0,
                     uint8* dst_argb, int dst_stride_argb,
                     int width, int height, int interpolation);
 
+// Row functions for copying a pixels from a source with a slope to a row
+// of destination.  Useful for scaling, rotation, mirror, texture mapping.
+void ARGBAffineRow_C(const uint8* src_argb, int src_argb_stride,
+                     uint8* dst_argb, const float* uv_dudv, int width);
+void ARGBAffineRow_SSE2(const uint8* src_argb, int src_argb_stride,
+                        uint8* dst_argb, const float* uv_dudv, int width);
+
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv

source/row.h

@@ -523,6 +523,9 @@ void ARGBShadeRow_C(const uint8* src_argb, uint8* dst_argb, int width,
                     uint32 value);
 void ARGBShadeRow_SSE2(const uint8* src_argb, uint8* dst_argb, int width,
                        uint32 value);
+
+void ARGBAffineRow_C(const uint8* src_argb, int src_argb_stride,
+                     uint8* dst_argb, const float* uv_dudv, int width);
 void ARGBAffineRow_SSE2(const uint8* src_argb, int src_argb_stride,
                         uint8* dst_argb, const float* uv_dudv, int width);
 

source/row_common.cc

@@ -984,6 +984,25 @@ void ARGBShadeRow_C(const uint8* src_argb, uint8* dst_argb, int width,
 #undef REPEAT8
 #undef SHADE
 
+// Copy pixels from rotated source to destination row with a slope.
+void ARGBAffineRow_C(const uint8* src_argb, int src_argb_stride,
+                     uint8* dst_argb, const float* uv_dudv, int width) {
+  // Render a row of pixels from source into a buffer.
+  float uv[2];
+  uv[0] = uv_dudv[0];
+  uv[1] = uv_dudv[1];
+  for (int i = 0; i < width; ++i) {
+    int x = static_cast<int>(uv[0]);
+    int y = static_cast<int>(uv[1]);
+    *reinterpret_cast<uint32*>(dst_argb) =
+        *reinterpret_cast<const uint32*>(src_argb + y * src_argb_stride +
+                                         x * 4);
+    dst_argb += 4;
+    uv[0] += uv_dudv[2];
+    uv[1] += uv_dudv[3];
+  }
+}
+
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv

unit_test/planar_test.cc

@@ -899,4 +899,44 @@ TEST_F(libyuvTest, TestInterpolate) {
   }
 }
 
+TEST_F(libyuvTest, TestAffine) {
+  SIMD_ALIGNED(uint8 orig_pixels_0[256][4]);
+  SIMD_ALIGNED(uint8 interpolate_pixels_Opt[256][4]);
+  SIMD_ALIGNED(uint8 interpolate_pixels_C[256][4]);
+
+  for (int i = 0; i < 256; ++i) {
+    for (int j = 0; j < 4; ++j) {
+      orig_pixels_0[i][j] = i;
+    }
+  }
+
+  float uv_step[4] = { 0.f, 0.f, 0.75f, 0.f };
+
+  ARGBAffineRow_C(&orig_pixels_0[0][0], 0, &interpolate_pixels_C[0][0],
+                  uv_step, 256);
+  EXPECT_EQ(0u, interpolate_pixels_C[0][0]);
+  EXPECT_EQ(96u, interpolate_pixels_C[128][0]);
+  EXPECT_EQ(191u, interpolate_pixels_C[255][3]);
+
+#if defined(_MSC_VER)
+  ARGBAffineRow_SSE2(&orig_pixels_0[0][0], 0, &interpolate_pixels_Opt[0][0],
+                     uv_step, 256);
+  EXPECT_EQ(0, memcmp(interpolate_pixels_Opt, interpolate_pixels_C, 256 * 4));
+#endif
+
+#if defined(_MSC_VER)
+  int has_sse2 = TestCpuFlag(kCpuHasSSE2);
+  if (has_sse2) {
+    for (int i = 0; i < 1000 * 1280 * 720 / 256; ++i) {
+      ARGBAffineRow_SSE2(&orig_pixels_0[0][0], 0, &interpolate_pixels_Opt[0][0],
+                         uv_step, 256);
+    }
+  } else
+#endif
+  for (int i = 0; i < 1000 * 1280 * 720 / 256; ++i) {
+    ARGBAffineRow_C(&orig_pixels_0[0][0], 0, &interpolate_pixels_C[0][0],
+                    uv_step, 256);
+  }
+}
+
 }  // namespace libyuv