[FFmpeg-cvslog] swscale/rgb : move shuffle func shuffle_bytes_1230, shuffle_bytes_3012, shuffle_bytes_3210 in order to add SIMD

[Martin Vignali]

ffmpeg | branch: master | Martin Vignali <martin.vignali at gmail.com> | Sat Mar 24 20:17:32 2018 +0100| [d4f6640855faf57da1ccc3cade356975fcb00207] | committer: Martin Vignali

swscale/rgb : move shuffle func shuffle_bytes_1230, shuffle_bytes_3012, shuffle_bytes_3210 in order to add SIMD

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=d4f6640855faf57da1ccc3cade356975fcb00207
---

 libswscale/rgb2rgb.c          | 23 ++++-------------------
 libswscale/rgb2rgb.h          |  6 +++---
 libswscale/rgb2rgb_template.c | 21 +++++++++++++++++++++
 3 files changed, 28 insertions(+), 22 deletions(-)

diff --git a/libswscale/rgb2rgb.c b/libswscale/rgb2rgb.c
index 04b7908b5e..eab8e6aebb 100644
--- a/libswscale/rgb2rgb.c
+++ b/libswscale/rgb2rgb.c
@@ -53,6 +53,10 @@ void (*rgb15to32)(const uint8_t *src, uint8_t *dst, int src_size);
 
 void (*shuffle_bytes_0321)(const uint8_t *src, uint8_t *dst, int src_size);
 void (*shuffle_bytes_2103)(const uint8_t *src, uint8_t *dst, int src_size);
+void (*shuffle_bytes_1230)(const uint8_t *src, uint8_t *dst, int src_size);
+void (*shuffle_bytes_3012)(const uint8_t *src, uint8_t *dst, int src_size);
+void (*shuffle_bytes_3210)(const uint8_t *src, uint8_t *dst, int src_size);
+
 
 void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc,
                    const uint8_t *vsrc, uint8_t *dst,
@@ -319,25 +323,6 @@ void rgb12tobgr12(const uint8_t *src, uint8_t *dst, int src_size)
     }
 }
 
-
-#define DEFINE_SHUFFLE_BYTES(a, b, c, d)                                \
-void shuffle_bytes_ ## a ## b ## c ## d(const uint8_t *src,             \
-                                        uint8_t *dst, int src_size)     \
-{                                                                       \
-    int i;                                                              \
-                                                                        \
-    for (i = 0; i < src_size; i += 4) {                                 \
-        dst[i + 0] = src[i + a];                                        \
-        dst[i + 1] = src[i + b];                                        \
-        dst[i + 2] = src[i + c];                                        \
-        dst[i + 3] = src[i + d];                                        \
-    }                                                                   \
-}
-
-DEFINE_SHUFFLE_BYTES(1, 2, 3, 0)
-DEFINE_SHUFFLE_BYTES(3, 0, 1, 2)
-DEFINE_SHUFFLE_BYTES(3, 2, 1, 0)
-
 #define DEFINE_RGB48TOBGR48(need_bswap, swap)                           \
 void rgb48tobgr48_ ## need_bswap(const uint8_t *src,                    \
                                  uint8_t *dst, int src_size)            \
diff --git a/libswscale/rgb2rgb.h b/libswscale/rgb2rgb.h
index 6994839299..08de546598 100644
--- a/libswscale/rgb2rgb.h
+++ b/libswscale/rgb2rgb.h
@@ -76,9 +76,9 @@ void rgb15tobgr15(const uint8_t *src, uint8_t *dst, int src_size);
 void rgb12tobgr12(const uint8_t *src, uint8_t *dst, int src_size);
 void    rgb12to15(const uint8_t *src, uint8_t *dst, int src_size);
 
-void shuffle_bytes_1230(const uint8_t *src, uint8_t *dst, int src_size);
-void shuffle_bytes_3012(const uint8_t *src, uint8_t *dst, int src_size);
-void shuffle_bytes_3210(const uint8_t *src, uint8_t *dst, int src_size);
+extern void (*shuffle_bytes_1230)(const uint8_t *src, uint8_t *dst, int src_size);
+extern void (*shuffle_bytes_3012)(const uint8_t *src, uint8_t *dst, int src_size);
+extern void (*shuffle_bytes_3210)(const uint8_t *src, uint8_t *dst, int src_size);
 
 void ff_rgb24toyv12_c(const uint8_t *src, uint8_t *ydst, uint8_t *udst,
                       uint8_t *vdst, int width, int height, int lumStride,
diff --git a/libswscale/rgb2rgb_template.c b/libswscale/rgb2rgb_template.c
index 499d25b26d..fb7d663ccc 100644
--- a/libswscale/rgb2rgb_template.c
+++ b/libswscale/rgb2rgb_template.c
@@ -342,6 +342,24 @@ static inline void shuffle_bytes_0321_c(const uint8_t *src, uint8_t *dst,
     }
 }
 
+#define DEFINE_SHUFFLE_BYTES(name, a, b, c, d)                          \
+static void shuffle_bytes_##name (const uint8_t *src,                   \
+                                        uint8_t *dst, int src_size)     \
+{                                                                       \
+    int i;                                                              \
+                                                                        \
+    for (i = 0; i < src_size; i += 4) {                                 \
+        dst[i + 0] = src[i + a];                                        \
+        dst[i + 1] = src[i + b];                                        \
+        dst[i + 2] = src[i + c];                                        \
+        dst[i + 3] = src[i + d];                                        \
+    }                                                                   \
+}
+
+DEFINE_SHUFFLE_BYTES(1230_c, 1, 2, 3, 0)
+DEFINE_SHUFFLE_BYTES(3012_c, 3, 0, 1, 2)
+DEFINE_SHUFFLE_BYTES(3210_c, 3, 2, 1, 0)
+
 static inline void rgb24tobgr24_c(const uint8_t *src, uint8_t *dst, int src_size)
 {
     unsigned i;
@@ -949,6 +967,9 @@ static av_cold void rgb2rgb_init_c(void)
 #else
     shuffle_bytes_0321 = shuffle_bytes_0321_c;
     shuffle_bytes_2103 = shuffle_bytes_2103_c;
+    shuffle_bytes_1230 = shuffle_bytes_1230_c;
+    shuffle_bytes_3012 = shuffle_bytes_3012_c;
+    shuffle_bytes_3210 = shuffle_bytes_3210_c;
 #endif
     rgb32tobgr16       = rgb32tobgr16_c;
     rgb32tobgr15       = rgb32tobgr15_c;