Flat shade an ARGB image

BUG=none
TEST=none
Review URL: https://webrtc-codereview.appspot.com/683004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@298 16f28f9a-4ce2-e073-06de-1de4eb20be90

README.chromium

 Name: libyuv
 URL: http://code.google.com/p/libyuv/
-Version: 297
+Version: 298
 License: BSD
 License File: LICENSE
 

include/libyuv/planar_functions.h

@@ -15,7 +15,7 @@
 
 // TODO(fbarchard): Remove the following headers includes
 #include "libyuv/convert.h"
-#include "libyuv/planar_functions.h"
+#include "libyuv/convert_argb.h"
 
 #ifdef __cplusplus
 namespace libyuv {
@@ -188,11 +188,6 @@ int ARGBUnattenuate(const uint8* src_argb, int src_stride_argb,
                     uint8* dst_argb, int dst_stride_argb,
                     int width, int height);
 
-// Multiply ARGB image by ARGB value.
-int ARGBShade(const uint8* src_argb, int src_stride_argb,
-              uint8* dst_argb, int dst_stride_argb,
-              int width, int height, uint32 value);
-
 // Convert MJPG to ARGB.
 int MJPGToARGB(const uint8* sample, size_t sample_size,
                uint8* argb, int argb_stride,
@@ -212,6 +207,11 @@ int ARGBBlur(const uint8* src_argb, int src_stride_argb,
              int32* dst_cumsum, int dst_stride32_cumsum,
              int width, int height, int radius);
 
+// Multiply ARGB image by ARGB value.
+int ARGBShade(const uint8* src_argb, int src_stride_argb,
+              uint8* dst_argb, int dst_stride_argb,
+              int width, int height, uint32 value);
+
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv

include/libyuv/version.h

@@ -11,6 +11,6 @@
 #ifndef INCLUDE_LIBYUV_VERSION_H_  // NOLINT
 #define INCLUDE_LIBYUV_VERSION_H_
 
-#define LIBYUV_VERSION 297
+#define LIBYUV_VERSION 298
 
 #endif  // INCLUDE_LIBYUV_VERSION_H_  NOLINT

source/convert.cc

@@ -784,11 +784,9 @@ int UYVYToI420(const uint8* src_uyvy, int src_stride_uyvy,
   return 0;
 }
 
-// Visual C for x86 defines these.
-#if defined(_M_X64) || defined(_M_IX86)
-#define LIBYUV_LITTLE_ENDIAN
-// GCC provided macros.
-#elif __BYTE_ORDER == __ORDER_LITTLE_ENDIAN__ || __BYTE_ORDER == __LITTLE_ENDIAN
+// Visual C x86 or GCC little endian.
+#if defined(_M_X64) || defined(_M_IX86) || (defined(__BYTE_ORDER) && \
+  (__BYTE_ORDER == __ORDER_LITTLE_ENDIAN__ || __BYTE_ORDER == __LITTLE_ENDIAN))
 #define LIBYUV_LITTLE_ENDIAN
 #endif
 

source/convert_from.cc

@@ -401,18 +401,15 @@ static void I42xToUYVYRow_C(const uint8* src_y,
     }
 }
 
-// Visual C for x86 defines these.
-#if defined(_M_X64) || defined(_M_IX86)
-#define LIBYUV_LITTLE_ENDIAN
-// GCC provided macros.
-#elif __BYTE_ORDER == __ORDER_LITTLE_ENDIAN__ || __BYTE_ORDER == __LITTLE_ENDIAN
+// Visual C x86 or GCC little endian.
+#if defined(_M_X64) || defined(_M_IX86) || (defined(__BYTE_ORDER) && \
+  (__BYTE_ORDER == __ORDER_LITTLE_ENDIAN__ || __BYTE_ORDER == __LITTLE_ENDIAN))
 #define LIBYUV_LITTLE_ENDIAN
 #endif
 
 #ifdef LIBYUV_LITTLE_ENDIAN
 #define WRITEWORD(p, v) *reinterpret_cast<uint32*>(p) = v
 #else
-
 static inline void WRITEWORD(uint8* p, uint32 v) {
   p[0] = (uint8)(v & 255);
   p[1] = (uint8)((v >> 8) & 255);

source/planar_functions.cc

@@ -60,8 +60,7 @@ int I420ToI400(const uint8* src_y, int src_stride_y,
                uint8*, int,
                uint8*, int,
                int width, int height) {
-  if (!src_y || !dst_y ||
-      width <= 0 || height == 0) {
+  if (!src_y || !dst_y || width <= 0 || height == 0) {
     return -1;
   }
   // Negative height means invert the image.
@@ -112,8 +111,7 @@ int I420Mirror(const uint8* src_y, int src_stride_y,
                uint8* dst_u, int dst_stride_u,
                uint8* dst_v, int dst_stride_v,
                int width, int height) {
-  if (!src_y || !src_u || !src_v ||
-      !dst_y || !dst_u || !dst_v ||
+  if (!src_y || !src_u || !src_v || !dst_y || !dst_u || !dst_v ||
       width <= 0 || height == 0) {
     return -1;
   }
@@ -143,9 +141,7 @@ int I420Mirror(const uint8* src_y, int src_stride_y,
 int ARGBMirror(const uint8* src_argb, int src_stride_argb,
                uint8* dst_argb, int dst_stride_argb,
                int width, int height) {
-  if (!src_argb ||
-      !dst_argb ||
-      width <= 0 || height == 0) {
+  if (!src_argb || !dst_argb || width <= 0 || height == 0) {
     return -1;
   }
   // Negative height means invert the image.
@@ -224,6 +220,9 @@ int ARGBBlend(const uint8* src_argb0, int src_stride_argb0,
 int ARGBToI400(const uint8* src_argb, int src_stride_argb,
                uint8* dst_y, int dst_stride_y,
                int width, int height) {
+  if (!src_argb || !dst_y || width <= 0 || height == 0) {
+    return -1;
+  }
   if (height < 0) {
     height = -height;
     src_argb = src_argb + (height - 1) * src_stride_argb;
@@ -255,6 +254,9 @@ int ARGBToI422(const uint8* src_argb, int src_stride_argb,
                uint8* dst_u, int dst_stride_u,
                uint8* dst_v, int dst_stride_v,
                int width, int height) {
+  if (!src_argb || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
+    return -1;
+  }
   if (height < 0) {
     height = -height;
     src_argb = src_argb + (height - 1) * src_stride_argb;
@@ -298,6 +300,9 @@ int ARGBToI422(const uint8* src_argb, int src_stride_argb,
 int ARGBToRGB24(const uint8* src_argb, int src_stride_argb,
                 uint8* dst_rgb24, int dst_stride_rgb24,
                 int width, int height) {
+  if (!src_argb || !dst_rgb24 || width <= 0 || height == 0) {
+    return -1;
+  }
   if (height < 0) {
     height = -height;
     src_argb = src_argb + (height - 1) * src_stride_argb;
@@ -330,6 +335,9 @@ int ARGBToRGB24(const uint8* src_argb, int src_stride_argb,
 int ARGBToRAW(const uint8* src_argb, int src_stride_argb,
               uint8* dst_raw, int dst_stride_raw,
               int width, int height) {
+  if (!src_argb || !dst_raw || width <= 0 || height == 0) {
+    return -1;
+  }
   if (height < 0) {
     height = -height;
     src_argb = src_argb + (height - 1) * src_stride_argb;
@@ -362,6 +370,9 @@ int ARGBToRAW(const uint8* src_argb, int src_stride_argb,
 int ARGBToRGB565(const uint8* src_argb, int src_stride_argb,
                  uint8* dst_rgb565, int dst_stride_rgb565,
                  int width, int height) {
+  if (!src_argb || !dst_rgb565 || width <= 0 || height == 0) {
+    return -1;
+  }
   if (height < 0) {
     height = -height;
     src_argb = src_argb + (height - 1) * src_stride_argb;
@@ -393,6 +404,9 @@ int ARGBToRGB565(const uint8* src_argb, int src_stride_argb,
 int ARGBToARGB1555(const uint8* src_argb, int src_stride_argb,
                    uint8* dst_argb1555, int dst_stride_argb1555,
                    int width, int height) {
+  if (!src_argb || !dst_argb1555 || width <= 0 || height == 0) {
+    return -1;
+  }
   if (height < 0) {
     height = -height;
     src_argb = src_argb + (height - 1) * src_stride_argb;
@@ -424,6 +438,9 @@ int ARGBToARGB1555(const uint8* src_argb, int src_stride_argb,
 int ARGBToARGB4444(const uint8* src_argb, int src_stride_argb,
                    uint8* dst_argb4444, int dst_stride_argb4444,
                    int width, int height) {
+  if (!src_argb || !dst_argb4444 || width <= 0 || height == 0) {
+    return -1;
+  }
   if (height < 0) {
     height = -height;
     src_argb = src_argb + (height - 1) * src_stride_argb;
@@ -457,6 +474,9 @@ int NV12ToRGB565(const uint8* src_y, int src_stride_y,
                  const uint8* src_uv, int src_stride_uv,
                  uint8* dst_rgb565, int dst_stride_rgb565,
                  int width, int height) {
+  if (!src_y || !src_uv || !dst_rgb565 || width <= 0 || height == 0) {
+    return -1;
+  }
   // Negative height means invert the image.
   if (height < 0) {
     height = -height;
@@ -499,6 +519,9 @@ int NV21ToRGB565(const uint8* src_y, int src_stride_y,
                  const uint8* src_vu, int src_stride_vu,
                  uint8* dst_rgb565, int dst_stride_rgb565,
                  int width, int height) {
+  if (!src_y || !src_vu || !dst_rgb565 || width <= 0 || height == 0) {
+    return -1;
+  }
   // Negative height means invert the image.
   if (height < 0) {
     height = -height;
@@ -762,6 +785,9 @@ int ARGBRect(uint8* dst_argb, int dst_stride_argb,
 int ARGBAttenuate(const uint8* src_argb, int src_stride_argb,
                   uint8* dst_argb, int dst_stride_argb,
                   int width, int height) {
+  if (!src_argb || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
   if (height < 0) {
     height = -height;
     src_argb = src_argb + (height - 1) * src_stride_argb;
@@ -796,6 +822,9 @@ int ARGBAttenuate(const uint8* src_argb, int src_stride_argb,
 int ARGBUnattenuate(const uint8* src_argb, int src_stride_argb,
                     uint8* dst_argb, int dst_stride_argb,
                     int width, int height) {
+  if (!src_argb || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
   if (height < 0) {
     height = -height;
     src_argb = src_argb + (height - 1) * src_stride_argb;
@@ -866,7 +895,8 @@ int ARGBSepia(uint8* dst_argb, int dst_stride_argb,
 int ARGBColorMatrix(uint8* dst_argb, int dst_stride_argb,
                     const int8* matrix_argb,
                     int dst_x, int dst_y, int width, int height) {
-  if (!dst_argb || width <= 0 || height <= 0 || dst_x < 0 || dst_y < 0) {
+  if (!dst_argb || !matrix_argb || width <= 0 || height <= 0 ||
+      dst_x < 0 || dst_y < 0) {
     return -1;
   }
   void (*ARGBColorMatrixRow)(uint8* dst_argb, const int8* matrix_argb,
@@ -890,7 +920,8 @@ int ARGBColorMatrix(uint8* dst_argb, int dst_stride_argb,
 int ARGBColorTable(uint8* dst_argb, int dst_stride_argb,
                    const uint8* table_argb,
                    int dst_x, int dst_y, int width, int height) {
-  if (!dst_argb || width <= 0 || height <= 0 || dst_x < 0 || dst_y < 0) {
+  if (!dst_argb || !table_argb || width <= 0 || height <= 0 ||
+      dst_x < 0 || dst_y < 0) {
     return -1;
   }
   void (*ARGBColorTableRow)(uint8* dst_argb, const uint8* table_argb,
@@ -972,6 +1003,9 @@ int ARGBBlur(const uint8* src_argb, int src_stride_argb,
              uint8* dst_argb, int dst_stride_argb,
              int32* dst_cumsum, int dst_stride32_cumsum,
              int width, int height, int radius) {
+  if (!src_argb || !dst_argb || width <= 0 || height == 0) {
+    return -1;
+  }
   void (*ComputeCumulativeSumRow)(const uint8* row, int32* cumsum,
       const int32* previous_cumsum, int width) = ComputeCumulativeSumRow_C;
   void (*CumulativeSumToAverage)(const int32* topleft, const int32* botleft,
@@ -1052,6 +1086,30 @@ int ARGBBlur(const uint8* src_argb, int src_stride_argb,
 int ARGBShade(const uint8* src_argb, int src_stride_argb,
               uint8* dst_argb, int dst_stride_argb,
               int width, int height, uint32 value) {
+  if (!src_argb || !dst_argb || width <= 0 || height == 0 || value == 0u) {
+    return -1;
+  }
+  if (height < 0) {
+    height = -height;
+    src_argb = src_argb + (height - 1) * src_stride_argb;
+    src_stride_argb = -src_stride_argb;
+  }
+  void (*ARGBShadeRow)(const uint8* src_argb, uint8* dst_argb,
+                       int width, uint32 value) = ARGBShadeRow_C;
+#if defined(HAS_ARGBSHADE_SSE2)
+  if (TestCpuFlag(kCpuHasSSE2) && IS_ALIGNED(width, 4) &&
+      IS_ALIGNED(src_argb, 16) && IS_ALIGNED(src_stride_argb, 16) &&
+      IS_ALIGNED(dst_argb, 16) && IS_ALIGNED(dst_stride_argb, 16)) {
+    ARGBShadeRow = ARGBShadeRow_SSE2;
+  }
+#endif
+
+  for (int y = 0; y < height; ++y) {
+    ARGBShadeRow(src_argb, dst_argb, width, value);
+    src_argb += src_stride_argb;
+    dst_argb += dst_stride_argb;
+  }
+  return 0;
 }
 
 #ifdef __cplusplus

source/row.h

@@ -85,6 +85,7 @@ extern "C" {
 // The following are Windows only:
 #if !defined(YUV_DISABLE_ASM) && defined(_M_IX86)
 #define HAS_ARGBCOLORTABLEROW_X86
+#define HAS_ARGBSHADE_SSE2
 #endif
 
 // The following are disabled when SSSE3 is available:
@@ -516,6 +517,11 @@ void CumulativeSumToAverage_C(const int32* topleft, const int32* botleft,
 void ComputeCumulativeSumRow_C(const uint8* row, int32* cumsum,
                                const int32* previous_cumsum, int width);
 
+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);
+
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv

source/row_common.cc

@@ -956,6 +956,32 @@ void CumulativeSumToAverage_C(const int32* tl, const int32* bl,
   }
 }
 
+#define REPEAT8(v) (v) | ((v) << 8)
+#define SHADE(f, v) v * f >> 24
+
+void ARGBShadeRow_C(const uint8* src_argb, uint8* dst_argb, int width,
+                    uint32 value) {
+  const uint32 b_scale = REPEAT8(value & 0xff);
+  const uint32 g_scale = REPEAT8((value >> 8) & 0xff);
+  const uint32 r_scale = REPEAT8((value >> 16) & 0xff);
+  const uint32 a_scale = REPEAT8(value >> 24);
+
+  for (int i = 0; i < width; ++i) {
+    const uint32 b = REPEAT8(src_argb[0]);
+    const uint32 g = REPEAT8(src_argb[1]);
+    const uint32 r = REPEAT8(src_argb[2]);
+    const uint32 a = REPEAT8(src_argb[3]);
+    dst_argb[0] = SHADE(b, b_scale);
+    dst_argb[1] = SHADE(g, g_scale);
+    dst_argb[2] = SHADE(r, r_scale);
+    dst_argb[3] = SHADE(a, a_scale);
+    src_argb += 4;
+    dst_argb += 4;
+  }
+}
+#undef REPEAT8
+#undef SHADE
+
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv

unit_test/planar_test.cc

@@ -331,7 +331,6 @@ TESTATOBRANDOM(RGB565, 2, 2, ARGB, 4)
 TESTATOBRANDOM(ARGB1555, 2, 2, ARGB, 4)
 TESTATOBRANDOM(ARGB4444, 2, 2, ARGB, 4)
 
-
 TEST_F(libyuvTest, TestAttenuate) {
   SIMD_ALIGNED(uint8 orig_pixels[256][4]);
   SIMD_ALIGNED(uint8 atten_pixels[256][4]);
@@ -649,4 +648,56 @@ TEST_F(libyuvTest, TestARGBMirror) {
   }
 }
 
+TEST_F(libyuvTest, TestShade) {
+  SIMD_ALIGNED(uint8 orig_pixels[256][4]);
+  SIMD_ALIGNED(uint8 shade_pixels[256][4]);
+
+  // Test unattenuation clamps
+  orig_pixels[0][0] = 10u;
+  orig_pixels[0][1] = 20u;
+  orig_pixels[0][2] = 40u;
+  orig_pixels[0][3] = 80u;
+  // Test unattenuation transparent and opaque are unaffected
+  orig_pixels[1][0] = 0u;
+  orig_pixels[1][1] = 0u;
+  orig_pixels[1][2] = 0u;
+  orig_pixels[1][3] = 255u;
+  orig_pixels[2][0] = 0u;
+  orig_pixels[2][1] = 0u;
+  orig_pixels[2][2] = 0u;
+  orig_pixels[2][3] = 0u;
+  orig_pixels[3][0] = 0u;
+  orig_pixels[3][1] = 0u;
+  orig_pixels[3][2] = 0u;
+  orig_pixels[3][3] = 0u;
+  ARGBShade(&orig_pixels[0][0], 0, &shade_pixels[0][0], 0, 4, 1, 0x80ffffff);
+  EXPECT_EQ(10u, shade_pixels[0][0]);
+  EXPECT_EQ(20u, shade_pixels[0][1]);
+  EXPECT_EQ(40u, shade_pixels[0][2]);
+  EXPECT_EQ(40u, shade_pixels[0][3]);
+  EXPECT_EQ(0u, shade_pixels[1][0]);
+  EXPECT_EQ(0u, shade_pixels[1][1]);
+  EXPECT_EQ(0u, shade_pixels[1][2]);
+  EXPECT_EQ(128u, shade_pixels[1][3]);
+  EXPECT_EQ(0u, shade_pixels[2][0]);
+  EXPECT_EQ(0u, shade_pixels[2][1]);
+  EXPECT_EQ(0u, shade_pixels[2][2]);
+  EXPECT_EQ(0u, shade_pixels[2][3]);
+  EXPECT_EQ(0u, shade_pixels[3][0]);
+  EXPECT_EQ(0u, shade_pixels[3][1]);
+  EXPECT_EQ(0u, shade_pixels[3][2]);
+  EXPECT_EQ(0u, shade_pixels[3][3]);
+
+  ARGBShade(&orig_pixels[0][0], 0, &shade_pixels[0][0], 0, 4, 1, 0x80808080);
+  EXPECT_EQ(5u, shade_pixels[0][0]);
+  EXPECT_EQ(10u, shade_pixels[0][1]);
+  EXPECT_EQ(20u, shade_pixels[0][2]);
+  EXPECT_EQ(40u, shade_pixels[0][3]);
+
+  for (int i = 0; i < 1000 * 1280 * 720 / 256; ++i) {
+    ARGBShade(&orig_pixels[0][0], 0, &shade_pixels[0][0], 0, 256, 1,
+              0x80808080);
+  }
+}
+
 }  // namespace libyuv