+add Base implementation of function BgrToYuv420pV2.

docs/2023.html

@@ -35,10 +35,20 @@ <h1>Simd Library Release Notes (2023).</h1>
 <hr/>
 <h3 id="R130">October X, 2023 (version X.X.130)</h3>
 <h4>Algorithms</h4>
+<h5>New features</h5>
+<ul>
+ <li>Base implementation of function BgrToYuv420pV2.</li>
+</ul>
 <ul>
  <li>Error in AVX-512BW optimizations of function SynetSoftmaxLayerForward.</li>
 </ul>
 
+<h4>Test framework</h4>
+<h5>New features</h5>
+<ul>
+ <li>Tests for verifying functionality of function BgrToYuv420pV2.</li>
+</ul>
+
 <a href="#HOME">Home</a>
 <hr/>
 <h3 id="R129">September 4, 2023 (version 5.3.129)</h3>

src/Simd/SimdBase.h

@@ -168,6 +168,9 @@ namespace Simd
 
         void BgrToYuv420p(const uint8_t * bgr, size_t width, size_t height, size_t bgrStride, uint8_t * y, size_t yStride, uint8_t * u, size_t uStride, uint8_t * v, size_t vStride);
 
+        void BgrToYuv420pV2(const uint8_t* bgr, size_t bgrStride, size_t width, size_t height, 
+            uint8_t* y, size_t yStride, uint8_t* u, size_t uStride, uint8_t* v, size_t vStride, SimdYuvType yuvType);
+
         void BgrToYuv422p(const uint8_t * bgr, size_t width, size_t height, size_t bgrStride, uint8_t * y, size_t yStride, uint8_t * u, size_t uStride, uint8_t * v, size_t vStride);
 
         void BgrToYuv444p(const uint8_t * bgr, size_t width, size_t height, size_t bgrStride, uint8_t * y, size_t yStride, uint8_t * u, size_t uStride, uint8_t * v, size_t vStride);

src/Simd/SimdBaseBgrToYuv.cpp

@@ -23,6 +23,7 @@
 * SOFTWARE.
 */
 #include "Simd/SimdConversion.h"
+#include "Simd/SimdYuvToBgr.h"
 
 namespace Simd
 {
@@ -114,5 +115,55 @@ namespace Simd
                 bgr += bgrStride;
             }
         }
+
+        //-------------------------------------------------------------------------------------------------
+
+        template <class YuvType> SIMD_INLINE void BgrToYuv420pV2(const uint8_t* bgr0, size_t bgrStride, uint8_t* y0, size_t yStride, uint8_t* u, uint8_t* v)
+        {
+            const uint8_t* bgr1 = bgr0 + bgrStride;
+            uint8_t* y1 = y0 + yStride;
+
+            y0[0] = BgrToY<YuvType>(bgr0[0], bgr0[1], bgr0[2]);
+            y0[1] = BgrToY<YuvType>(bgr0[2], bgr0[4], bgr0[5]);
+            y1[0] = BgrToY<YuvType>(bgr1[0], bgr1[1], bgr1[2]);
+            y1[1] = BgrToY<YuvType>(bgr1[3], bgr1[4], bgr1[5]);
+
+            int blue = Average(bgr0[0], bgr0[3], bgr1[0], bgr1[3]);
+            int green = Average(bgr0[1], bgr0[4], bgr1[1], bgr1[4]);
+            int red = Average(bgr0[2], bgr0[5], bgr1[2], bgr1[5]);
+
+            u[0] = BgrToU<YuvType>(blue, green, red);
+            v[0] = BgrToV<YuvType>(blue, green, red);
+        }
+
+        template <class YuvType> void BgrToYuv420pV2(const uint8_t* bgr, size_t bgrStride, size_t width, size_t height,
+            uint8_t* y, size_t yStride, uint8_t* u, size_t uStride, uint8_t* v, size_t vStride)
+        {
+            assert((width % 2 == 0) && (height % 2 == 0) && (width >= 2) && (height >= 2));
+
+            for (size_t row = 0; row < height; row += 2)
+            {
+                for (size_t colUV = 0, colY = 0, colBgr = 0; colY < width; colY += 2, colUV++, colBgr += 6)
+                    BgrToYuv420pV2<YuvType>(bgr + colBgr, bgrStride, y + colY, yStride, u + colUV, v + colUV);
+                y += 2 * yStride;
+                u += uStride;
+                v += vStride;
+                bgr += 2 * bgrStride;
+            }
+        }
+
+        void BgrToYuv420pV2(const uint8_t* bgr, size_t bgrStride, size_t width, size_t height,
+            uint8_t* y, size_t yStride, uint8_t* u, size_t uStride, uint8_t* v, size_t vStride, SimdYuvType yuvType)
+        {
+            switch (yuvType)
+            {
+            case SimdYuvBt601: BgrToYuv420pV2<Bt601>(bgr, bgrStride, width, height, y, yStride, u, uStride, v, vStride); break;
+            case SimdYuvBt709: BgrToYuv420pV2<Bt709>(bgr, bgrStride, width, height, y, yStride, u, uStride, v, vStride); break;
+            case SimdYuvBt2020: BgrToYuv420pV2<Bt2020>(bgr, bgrStride, width, height, y, yStride, u, uStride, v, vStride); break;
+            case SimdYuvTrect871: BgrToYuv420pV2<Trect871>(bgr, bgrStride, width, height, y, yStride, u, uStride, v, vStride); break;
+            default:
+                assert(0);
+            }
+        }
     }
 }

src/Simd/SimdBaseBgraToYuv.cpp

@@ -351,6 +351,5 @@ namespace Simd
                 assert(0);
             }
         }
-
     }
 }

src/Simd/SimdLib.cpp

@@ -1592,6 +1592,33 @@ SIMD_API void SimdBgrToYuv420p(const uint8_t * bgr, size_t width, size_t height,
         Base::BgrToYuv420p(bgr, width, height, bgrStride, y, yStride, u, uStride, v, vStride);
 }
 
+SIMD_API void SimdBgrToYuv420pV2(const uint8_t* bgr, size_t bgrStride, size_t width, size_t height,
+    uint8_t* y, size_t yStride, uint8_t* u, size_t uStride, uint8_t* v, size_t vStride, SimdYuvType yuvType)
+{
+    SIMD_EMPTY();
+//#ifdef SIMD_AVX512BW_ENABLE
+//    if (Avx512bw::Enable)
+//        Avx512bw::BgrToYuv420pV2(bgr, bgrStride, width, height, y, yStride, u, uStride, v, vStride, yuvType);
+//    else
+//#endif
+//#ifdef SIMD_AVX2_ENABLE
+//    if (Avx2::Enable && width >= Avx2::DA)
+//        Avx2::BgrToYuv420pV2(bgr, bgrStride, width, height, y, yStride, u, uStride, v, vStride, yuvType);
+//    else
+//#endif
+//#ifdef SIMD_SSE41_ENABLE
+//    if (Sse41::Enable && width >= Sse41::DA)
+//        Sse41::BgrToYuv420pV2(bgr, bgrStride, width, height, y, yStride, u, uStride, v, vStride, yuvType);
+//    else
+//#endif
+//#ifdef SIMD_NEON_ENABLE
+//    if (Neon::Enable && width >= Neon::DA)
+//        Neon::BgrToYuv420pV2(bgr, bgrStride, width, height, y, yStride, u, uStride, v, vStride, yuvType);
+//    else
+//#endif
+        Base::BgrToYuv420pV2(bgr, bgrStride, width, height, y, yStride, u, uStride, v, vStride, yuvType);
+}
+
 SIMD_API void SimdBgrToYuv422p(const uint8_t * bgr, size_t width, size_t height, size_t bgrStride, uint8_t * y, size_t yStride, uint8_t * u, size_t uStride, uint8_t * v, size_t vStride)
 {
     SIMD_EMPTY();

src/Simd/SimdLib.h

@@ -2147,6 +2147,29 @@ extern "C"
     */
     SIMD_API void SimdBgrToYuv420p(const uint8_t * bgr, size_t width, size_t height, size_t bgrStride, uint8_t * y, size_t yStride, uint8_t * u, size_t uStride, uint8_t * v, size_t vStride);
 
+    /*! @ingroup bgr_conversion
+
+        \fn void SimdBgrToYuv420pV2(const uint8_t * bgr, size_t bgrStride, size_t width, size_t height, uint8_t * y, size_t yStride, uint8_t * u, size_t uStride, uint8_t * v, size_t vStride, SimdYuvType yuvType);
+
+        \short Converts 24-bit BGR image to YUV420P.
+
+        The input BGR and output Y images must have the same width and height.
+        The input U and V images must have the same width and height (half size relative to Y component).
+
+        \param [in] bgr - a pointer to pixels data of input 24-bit BGR image.
+        \param [in] bgrStride - a row size of the BGR image.
+        \param [in] width - an image width.
+        \param [in] height - an image height.
+        \param [out] y - a pointer to pixels data of output 8-bit image with Y color plane.
+        \param [in] yStride - a row size of the y image.
+        \param [out] u - a pointer to pixels data of output 8-bit image with U color plane.
+        \param [in] uStride - a row size of the u image.
+        \param [out] v - a pointer to pixels data of output 8-bit image with V color plane.
+        \param [in] vStride - a row size of the v image.
+        \param [in] yuvType - a type of output YUV image (see descriprion of ::SimdYuvType).
+    */
+    SIMD_API void SimdBgrToYuv420pV2(const uint8_t* bgr, size_t bgrStride, size_t width, size_t height, uint8_t* y, size_t yStride, uint8_t* u, size_t uStride, uint8_t* v, size_t vStride, SimdYuvType yuvType);
+
     /*! @ingroup bgr_conversion
 
         \fn void SimdBgrToYuv422p(const uint8_t * bgr, size_t width, size_t height, size_t bgrStride, uint8_t * y, size_t yStride, uint8_t * u, size_t uStride, uint8_t * v, size_t vStride);

src/Test/Test.cpp

@@ -115,6 +115,7 @@ namespace Test
     TEST_ADD_GROUP_A0(BgraToYuv444p);
     TEST_ADD_GROUP_A0(BgraToYuv444pV2);
     TEST_ADD_GROUP_A0(BgrToYuv420p);
+    TEST_ADD_GROUP_A0(BgrToYuv420pV2);
     TEST_ADD_GROUP_A0(BgrToYuv422p);
     TEST_ADD_GROUP_A0(BgrToYuv444p);
     TEST_ADD_GROUP_A0(Uyvy422ToYuv420p);

src/Test/TestAnyToYuv.cpp

@@ -499,6 +499,35 @@ namespace Test
         return result;
     }
 
+    bool BgrToYuv420pV2AutoTest()
+    {
+        bool result = true;
+
+        result = result && AnyToYuvV2AutoTest(View::Bgr24, 2, 2, FUNC_YUV2(Simd::Base::BgrToYuv420pV2), FUNC_YUV2(SimdBgrToYuv420pV2));
+
+//#ifdef SIMD_SSE41_ENABLE
+//        if (Simd::Sse41::Enable)
+//            result = result && AnyToYuvV2AutoTest(View::Bgr24, 2, 2, FUNC_YUV2(Simd::Sse41::BgrToYuv420pV2), FUNC_YUV2(SimdBgrToYuv420pV2));
+//#endif 
+//
+//#ifdef SIMD_AVX2_ENABLE
+//        if (Simd::Avx2::Enable)
+//            result = result && AnyToYuvV2AutoTest(View::Bgr24, 2, 2, FUNC_YUV2(Simd::Avx2::BgrToYuv420pV2), FUNC_YUV2(SimdBgrToYuv420pV2));
+//#endif 
+//
+//#ifdef SIMD_AVX512BW_ENABLE
+//        if (Simd::Avx512bw::Enable)
+//            result = result && AnyToYuvV2AutoTest(View::Bgr24, 2, 2, FUNC_YUV2(Simd::Avx512bw::BgrToYuv420pV2), FUNC_YUV2(SimdBgrToYuv420pV2));
+//#endif 
+//
+//#ifdef SIMD_NEON_ENABLE
+//        if (Simd::Neon::Enable)
+//            result = result && AnyToYuvV2AutoTest(View::Bgr24, 2, 2, FUNC_YUV2(Simd::Neon::BgrToYuv420pV2), FUNC_YUV2(SimdBgrToYuv420pV2));
+//#endif
+
+        return result;
+    }
+
     //-------------------------------------------------------------------------------------------------
 
     namespace