swscale/output: Add rgba64/rgb48/bgra64/bgr48 output functions with full chroma inter...
authorMichael Niedermayer <michaelni@gmx.at>
Tue, 16 Jun 2015 22:01:47 +0000 (00:01 +0200)
committerMichael Niedermayer <michaelni@gmx.at>
Tue, 16 Jun 2015 22:45:46 +0000 (00:45 +0200)
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
libswscale/output.c
libswscale/utils.c
tests/ref/fate/filter-pixfmts-scale

index 56ed5f6..f63af3b 100644 (file)
@@ -925,6 +925,196 @@ yuv2rgba64_1_c_template(SwsContext *c, const int32_t *buf0,
     }
 }
 
+static av_always_inline void
+yuv2rgba64_full_X_c_template(SwsContext *c, const int16_t *lumFilter,
+                       const int32_t **lumSrc, int lumFilterSize,
+                       const int16_t *chrFilter, const int32_t **chrUSrc,
+                       const int32_t **chrVSrc, int chrFilterSize,
+                       const int32_t **alpSrc, uint16_t *dest, int dstW,
+                       int y, enum AVPixelFormat target, int hasAlpha, int eightbytes)
+{
+    int i;
+    int A = 0xffff<<14;
+
+    for (i = 0; i < dstW; i++) {
+        int j;
+        int Y  = -0x40000000;
+        int U  = -128 << 23; // 19
+        int V  = -128 << 23;
+        int R, G, B;
+
+        for (j = 0; j < lumFilterSize; j++) {
+            Y += lumSrc[j][i]  * (unsigned)lumFilter[j];
+        }
+        for (j = 0; j < chrFilterSize; j++) {;
+            U += chrUSrc[j][i] * (unsigned)chrFilter[j];
+            V += chrVSrc[j][i] * (unsigned)chrFilter[j];
+        }
+
+        if (hasAlpha) {
+            A = -0x40000000;
+            for (j = 0; j < lumFilterSize; j++) {
+                A += alpSrc[j][i] * (unsigned)lumFilter[j];
+            }
+            A >>= 1;
+            A += 0x20002000;
+        }
+
+        // 8bit: 12+15=27; 16-bit: 12+19=31
+        Y  >>= 14; // 10
+        Y += 0x10000;
+        U  >>= 14;
+        V  >>= 14;
+
+        // 8bit: 27 -> 17bit, 16bit: 31 - 14 = 17bit
+        Y -= c->yuv2rgb_y_offset;
+        Y *= c->yuv2rgb_y_coeff;
+        Y += 1 << 13; // 21
+        // 8bit: 17 + 13bit = 30bit, 16bit: 17 + 13bit = 30bit
+
+        R = V * c->yuv2rgb_v2r_coeff;
+        G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
+        B =                            U * c->yuv2rgb_u2b_coeff;
+
+        // 8bit: 30 - 22 = 8bit, 16bit: 30bit - 14 = 16bit
+        output_pixel(&dest[0], av_clip_uintp2(R_B + Y, 30) >> 14);
+        output_pixel(&dest[1], av_clip_uintp2(  G + Y, 30) >> 14);
+        output_pixel(&dest[2], av_clip_uintp2(B_R + Y, 30) >> 14);
+        if (eightbytes) {
+            output_pixel(&dest[3], av_clip_uintp2(A, 30) >> 14);
+            dest += 4;
+        } else {
+            dest += 3;
+        }
+    }
+}
+
+static av_always_inline void
+yuv2rgba64_full_2_c_template(SwsContext *c, const int32_t *buf[2],
+                       const int32_t *ubuf[2], const int32_t *vbuf[2],
+                       const int32_t *abuf[2], uint16_t *dest, int dstW,
+                       int yalpha, int uvalpha, int y,
+                       enum AVPixelFormat target, int hasAlpha, int eightbytes)
+{
+    const int32_t *buf0  = buf[0],  *buf1  = buf[1],
+                  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
+                  *vbuf0 = vbuf[0], *vbuf1 = vbuf[1],
+                  *abuf0 = hasAlpha ? abuf[0] : NULL,
+                  *abuf1 = hasAlpha ? abuf[1] : NULL;
+    int  yalpha1 = 4096 - yalpha;
+    int uvalpha1 = 4096 - uvalpha;
+    int i;
+    int A = 0xffff<<14;
+
+    for (i = 0; i < dstW; i++) {
+        int Y  = (buf0[i]     * yalpha1  + buf1[i]     * yalpha) >> 14;
+        int U  = (ubuf0[i]   * uvalpha1 + ubuf1[i]     * uvalpha + (-128 << 23)) >> 14;
+        int V  = (vbuf0[i]   * uvalpha1 + vbuf1[i]     * uvalpha + (-128 << 23)) >> 14;
+        int R, G, B;
+
+        Y -= c->yuv2rgb_y_offset;
+        Y *= c->yuv2rgb_y_coeff;
+        Y += 1 << 13;
+
+        R = V * c->yuv2rgb_v2r_coeff;
+        G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
+        B =                            U * c->yuv2rgb_u2b_coeff;
+
+        if (hasAlpha) {
+            A = (abuf0[i] * yalpha1 + abuf1[i] * yalpha) >> 1;
+
+            A += 1 << 13;
+        }
+
+        output_pixel(&dest[0], av_clip_uintp2(R_B + Y, 30) >> 14);
+        output_pixel(&dest[1], av_clip_uintp2(  G + Y, 30) >> 14);
+        output_pixel(&dest[2], av_clip_uintp2(B_R + Y, 30) >> 14);
+        if (eightbytes) {
+            output_pixel(&dest[3], av_clip_uintp2(A, 30) >> 14);
+            dest += 4;
+        } else {
+            dest += 3;
+        }
+    }
+}
+
+static av_always_inline void
+yuv2rgba64_full_1_c_template(SwsContext *c, const int32_t *buf0,
+                       const int32_t *ubuf[2], const int32_t *vbuf[2],
+                       const int32_t *abuf0, uint16_t *dest, int dstW,
+                       int uvalpha, int y, enum AVPixelFormat target, int hasAlpha, int eightbytes)
+{
+    const int32_t *ubuf0 = ubuf[0], *vbuf0 = vbuf[0];
+    int i;
+    int A = 0xffff<<14;
+
+    if (uvalpha < 2048) {
+        for (i = 0; i < dstW; i++) {
+            int Y  = (buf0[i]) >> 2;
+            int U  = (ubuf0[i] + (-128 << 11)) >> 2;
+            int V  = (vbuf0[i] + (-128 << 11)) >> 2;
+            int R, G, B;
+
+            Y -= c->yuv2rgb_y_offset;
+            Y *= c->yuv2rgb_y_coeff;
+            Y += 1 << 13;
+
+            if (hasAlpha) {
+                A = abuf0[i] << 11;
+
+                A += 1 << 13;
+            }
+
+            R = V * c->yuv2rgb_v2r_coeff;
+            G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
+            B =                            U * c->yuv2rgb_u2b_coeff;
+
+            output_pixel(&dest[0], av_clip_uintp2(R_B + Y, 30) >> 14);
+            output_pixel(&dest[1], av_clip_uintp2(  G + Y, 30) >> 14);
+            output_pixel(&dest[2], av_clip_uintp2(B_R + Y, 30) >> 14);
+            if (eightbytes) {
+                output_pixel(&dest[3], av_clip_uintp2(A, 30) >> 14);
+                dest += 4;
+            } else {
+                dest += 3;
+            }
+        }
+    } else {
+        const int32_t *ubuf1 = ubuf[1], *vbuf1 = vbuf[1];
+        int A = 0xffff<<14;
+        for (i = 0; i < dstW; i++) {
+            int Y  = (buf0[i]    ) >> 2;
+            int U  = (ubuf0[i] + ubuf1[i] + (-128 << 12)) >> 3;
+            int V  = (vbuf0[i] + vbuf1[i] + (-128 << 12)) >> 3;
+            int R, G, B;
+
+            Y -= c->yuv2rgb_y_offset;
+            Y *= c->yuv2rgb_y_coeff;
+            Y += 1 << 13;
+
+            if (hasAlpha) {
+                A = abuf0[i] << 11;
+
+                A += 1 << 13;
+            }
+
+            R = V * c->yuv2rgb_v2r_coeff;
+            G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
+            B =                            U * c->yuv2rgb_u2b_coeff;
+
+            output_pixel(&dest[0], av_clip_uintp2(R_B + Y, 30) >> 14);
+            output_pixel(&dest[1], av_clip_uintp2(  G + Y, 30) >> 14);
+            output_pixel(&dest[2], av_clip_uintp2(B_R + Y, 30) >> 14);
+            if (eightbytes) {
+                output_pixel(&dest[3], av_clip_uintp2(A, 30) >> 14);
+                dest += 4;
+            } else {
+                dest += 3;
+            }
+        }
+    }
+}
+
 #undef output_pixel
 #undef r_b
 #undef b_r
@@ -988,6 +1178,19 @@ YUV2PACKED16WRAPPER(yuv2, rgba64, bgra64le, AV_PIX_FMT_BGRA64LE, 1, 1)
 YUV2PACKED16WRAPPER(yuv2, rgba64, bgrx64be, AV_PIX_FMT_BGRA64BE, 0, 1)
 YUV2PACKED16WRAPPER(yuv2, rgba64, bgrx64le, AV_PIX_FMT_BGRA64LE, 0, 1)
 
+YUV2PACKED16WRAPPER(yuv2, rgba64_full, rgb48be_full, AV_PIX_FMT_RGB48BE, 0, 0)
+YUV2PACKED16WRAPPER(yuv2, rgba64_full, rgb48le_full, AV_PIX_FMT_RGB48LE, 0, 0)
+YUV2PACKED16WRAPPER(yuv2, rgba64_full, bgr48be_full, AV_PIX_FMT_BGR48BE, 0, 0)
+YUV2PACKED16WRAPPER(yuv2, rgba64_full, bgr48le_full, AV_PIX_FMT_BGR48LE, 0, 0)
+YUV2PACKED16WRAPPER(yuv2, rgba64_full, rgba64be_full, AV_PIX_FMT_RGBA64BE, 1, 1)
+YUV2PACKED16WRAPPER(yuv2, rgba64_full, rgba64le_full, AV_PIX_FMT_RGBA64LE, 1, 1)
+YUV2PACKED16WRAPPER(yuv2, rgba64_full, rgbx64be_full, AV_PIX_FMT_RGBA64BE, 0, 1)
+YUV2PACKED16WRAPPER(yuv2, rgba64_full, rgbx64le_full, AV_PIX_FMT_RGBA64LE, 0, 1)
+YUV2PACKED16WRAPPER(yuv2, rgba64_full, bgra64be_full, AV_PIX_FMT_BGRA64BE, 1, 1)
+YUV2PACKED16WRAPPER(yuv2, rgba64_full, bgra64le_full, AV_PIX_FMT_BGRA64LE, 1, 1)
+YUV2PACKED16WRAPPER(yuv2, rgba64_full, bgrx64be_full, AV_PIX_FMT_BGRA64BE, 0, 1)
+YUV2PACKED16WRAPPER(yuv2, rgba64_full, bgrx64le_full, AV_PIX_FMT_BGRA64LE, 0, 1)
+
 /*
  * Write out 2 RGB pixels in the target pixel format. This function takes a
  * R/G/B LUT as generated by ff_yuv2rgb_c_init_tables(), which takes care of
@@ -1833,7 +2036,64 @@ av_cold void ff_sws_init_output_funcs(SwsContext *c,
                 }
 #endif /* !CONFIG_SMALL */
                 break;
-            case AV_PIX_FMT_RGB24:
+        case AV_PIX_FMT_RGBA64LE:
+#if CONFIG_SWSCALE_ALPHA
+            if (c->alpPixBuf) {
+                *yuv2packedX = yuv2rgba64le_full_X_c;
+                *yuv2packed2 = yuv2rgba64le_full_2_c;
+                *yuv2packed1 = yuv2rgba64le_full_1_c;
+            } else
+#endif /* CONFIG_SWSCALE_ALPHA */
+            {
+                *yuv2packedX = yuv2rgbx64le_full_X_c;
+                *yuv2packed2 = yuv2rgbx64le_full_2_c;
+                *yuv2packed1 = yuv2rgbx64le_full_1_c;
+            }
+            break;
+        case AV_PIX_FMT_RGBA64BE:
+#if CONFIG_SWSCALE_ALPHA
+            if (c->alpPixBuf) {
+                *yuv2packedX = yuv2rgba64be_full_X_c;
+                *yuv2packed2 = yuv2rgba64be_full_2_c;
+                *yuv2packed1 = yuv2rgba64be_full_1_c;
+            } else
+#endif /* CONFIG_SWSCALE_ALPHA */
+            {
+                *yuv2packedX = yuv2rgbx64be_full_X_c;
+                *yuv2packed2 = yuv2rgbx64be_full_2_c;
+                *yuv2packed1 = yuv2rgbx64be_full_1_c;
+            }
+            break;
+        case AV_PIX_FMT_BGRA64LE:
+#if CONFIG_SWSCALE_ALPHA
+            if (c->alpPixBuf) {
+                *yuv2packedX = yuv2bgra64le_full_X_c;
+                *yuv2packed2 = yuv2bgra64le_full_2_c;
+                *yuv2packed1 = yuv2bgra64le_full_1_c;
+            } else
+#endif /* CONFIG_SWSCALE_ALPHA */
+            {
+                *yuv2packedX = yuv2bgrx64le_full_X_c;
+                *yuv2packed2 = yuv2bgrx64le_full_2_c;
+                *yuv2packed1 = yuv2bgrx64le_full_1_c;
+            }
+            break;
+        case AV_PIX_FMT_BGRA64BE:
+#if CONFIG_SWSCALE_ALPHA
+            if (c->alpPixBuf) {
+                *yuv2packedX = yuv2bgra64be_full_X_c;
+                *yuv2packed2 = yuv2bgra64be_full_2_c;
+                *yuv2packed1 = yuv2bgra64be_full_1_c;
+            } else
+#endif /* CONFIG_SWSCALE_ALPHA */
+            {
+                *yuv2packedX = yuv2bgrx64be_full_X_c;
+                *yuv2packed2 = yuv2bgrx64be_full_2_c;
+                *yuv2packed1 = yuv2bgrx64be_full_1_c;
+            }
+            break;
+
+        case AV_PIX_FMT_RGB24:
             *yuv2packedX = yuv2rgb24_full_X_c;
             *yuv2packed2 = yuv2rgb24_full_2_c;
             *yuv2packed1 = yuv2rgb24_full_1_c;
@@ -1843,6 +2103,26 @@ av_cold void ff_sws_init_output_funcs(SwsContext *c,
             *yuv2packed2 = yuv2bgr24_full_2_c;
             *yuv2packed1 = yuv2bgr24_full_1_c;
             break;
+        case AV_PIX_FMT_RGB48LE:
+            *yuv2packedX = yuv2rgb48le_full_X_c;
+            *yuv2packed2 = yuv2rgb48le_full_2_c;
+            *yuv2packed1 = yuv2rgb48le_full_1_c;
+            break;
+        case AV_PIX_FMT_BGR48LE:
+            *yuv2packedX = yuv2bgr48le_full_X_c;
+            *yuv2packed2 = yuv2bgr48le_full_2_c;
+            *yuv2packed1 = yuv2bgr48le_full_1_c;
+            break;
+        case AV_PIX_FMT_RGB48BE:
+            *yuv2packedX = yuv2rgb48be_full_X_c;
+            *yuv2packed2 = yuv2rgb48be_full_2_c;
+            *yuv2packed1 = yuv2rgb48be_full_1_c;
+            break;
+        case AV_PIX_FMT_BGR48BE:
+            *yuv2packedX = yuv2bgr48be_full_X_c;
+            *yuv2packed2 = yuv2bgr48be_full_2_c;
+            *yuv2packed1 = yuv2bgr48be_full_1_c;
+            break;
         case AV_PIX_FMT_BGR4_BYTE:
             *yuv2packedX = yuv2bgr4_byte_full_X_c;
             *yuv2packed2 = yuv2bgr4_byte_full_2_c;
index de0c7f9..074f8c0 100644 (file)
@@ -1158,6 +1158,14 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
     if (flags & SWS_FULL_CHR_H_INT &&
         isAnyRGB(dstFormat)        &&
         !isPlanarRGB(dstFormat)    &&
+        dstFormat != AV_PIX_FMT_RGBA64LE &&
+        dstFormat != AV_PIX_FMT_RGBA64BE &&
+        dstFormat != AV_PIX_FMT_BGRA64LE &&
+        dstFormat != AV_PIX_FMT_BGRA64BE &&
+        dstFormat != AV_PIX_FMT_RGB48LE &&
+        dstFormat != AV_PIX_FMT_RGB48BE &&
+        dstFormat != AV_PIX_FMT_BGR48LE &&
+        dstFormat != AV_PIX_FMT_BGR48BE &&
         dstFormat != AV_PIX_FMT_RGBA  &&
         dstFormat != AV_PIX_FMT_ARGB  &&
         dstFormat != AV_PIX_FMT_BGRA  &&
index 4d1cb7d..553a4c5 100644 (file)
@@ -6,8 +6,8 @@ bgr0                243d58ca64f97b2f415b4c63cb79f0e1
 bgr24               18744aaab4b8bce065a7144dc0ccf921
 bgr444be            920760bee08c4fa161bf060e21ebba92
 bgr444le            01be36a28ebca1a11eb4d192986cd4e9
-bgr48be             a6fee4ac9f70d0da6a4b3a0e6353ca7f
-bgr48le             9c5d30b3b31ceaf3009fc7f1cf1cf7b6
+bgr48be             3ae02769c69d2512eaa26fff65763acb
+bgr48le             a6ce2344f07b77438258b6787fe5c24c
 bgr4_byte           01efea74088e5e3343c19ee053b95f31
 bgr555be            ab353278d103d379e1ec86e5cabb645f
 bgr555le            16ccbf59297e4b9ab25fd8af5a84a95d
@@ -15,8 +15,8 @@ bgr565be            3477e19fc11f95285836f30fdff26c1d
 bgr565le            82a81e7c9d4e0431fa22f4df9694afdc
 bgr8                2c57e76ccf04d51de6acafcf35d6fa70
 bgra                d8316272bc3a360ef9dff3ecc84520a3
-bgra64be            688499004461a2ce9debadb36dbcde5b
-bgra64le            c80dda435633c301e14d5b46a7edcf8d
+bgra64be            4e6a1b9f9c18b881c27d76611d45f737
+bgra64le            efeee0abcc658ebcff049d5e74d74943
 gbrap               e97ea4a104467c482173b7eaa57c14e3
 gbrp                dc3387f925f972c61aae7eb23cdc19f0
 gbrp10be            3a6d59192b6bb89ab42252b2b4818519
@@ -39,8 +39,8 @@ rgb0                fbd27e98154efb7535826afed41e9bb0
 rgb24               e022e741451e81f2ecce1c7240b93e87
 rgb444be            db52b9ecdf98479b693e3f4bd9e77bac
 rgb444le            63288425c05f146cde5c82b85bb126e0
-rgb48be             c2e456838a71237cb1398ab5a7c35a6e
-rgb48le             6ef772549307349c599f419313c75b7a
+rgb48be             45b25016f10d54cf36eef3479afd8249
+rgb48le             40577b147620ecfb115717473d000697
 rgb4_byte           9e540a2e7193ebcbf1c7f85d192a0c4e
 rgb555be            cb5407a0d40f3d0120155daeaaa9a222
 rgb555le            c15540d1fc887882c35860634009c439
@@ -48,11 +48,11 @@ rgb565be            c69fa7d6e458509de65e911d147629a8
 rgb565le            a4a6ef89cdc10282b428cb1392f2a353
 rgb8                bcdc033b4ef0979d060dbc8893d4db58
 rgba                85bb5d03cea1c6e8002ced3373904336
-rgba64be            21611863fbbe149416a11e95877824ac
-rgba64le            35c195a441e5f8ca8e7e4ed098ecf0c1
+rgba64be            ee73e57923af984b31cc7795d13929da
+rgba64le            783d2779adfafe3548bdb671ec0de69e
 uyvy422             aeb4ba4f9f003ae21f6d18089198244f
-xyz12be             f6350b9a2f5add20d3d67f59c100166f
-xyz12le             982935a6ea6a297fd7be8aee0fda9870
+xyz12be             c7ba8345998c0141ddc079cdd29b1a40
+xyz12le             95f5d3a0de834cc495c9032a14987cde
 yuv410p             e8f49b5fb9335b62c074f7f8bb0234fc
 yuv411p             5af32557c93beb482e26e7af693104c6
 yuv420p             5d3ac239c3712143560b1dfbd48a7ddd