pixblockdsp: Use AV_COPY128U for get_pixels_16_c

Before:
  15543 decicycles in get_pixels, 4193214 runs,   1090 skips
After:
   5713 decicycles in get_pixels, 8387564 runs,   1044 skips

1 files changed, 32 insertions, 5 deletions

diff --git a/libavcodec/pixblockdsp.c b/libavcodec/pixblockdsp.c
index 322e1dd111..f0883d3d08 100644
--- a/libavcodec/pixblockdsp.c
+++ b/libavcodec/pixblockdsp.c
@@ -20,15 +20,42 @@
 
 #include "config.h"
 #include "libavutil/attributes.h"
+#include "libavutil/intreadwrite.h"
 #include "avcodec.h"
 #include "pixblockdsp.h"
 
-#define BIT_DEPTH 16
-#include "pixblockdsp_template.c"
-#undef BIT_DEPTH
+static void get_pixels_16_c(int16_t *av_restrict block, const uint8_t *pixels,
+                            ptrdiff_t line_size)
+{
+    AV_COPY128U(block + 0 * 8, pixels + 0 * line_size);
+    AV_COPY128U(block + 1 * 8, pixels + 1 * line_size);
+    AV_COPY128U(block + 2 * 8, pixels + 2 * line_size);
+    AV_COPY128U(block + 3 * 8, pixels + 3 * line_size);
+    AV_COPY128U(block + 4 * 8, pixels + 4 * line_size);
+    AV_COPY128U(block + 5 * 8, pixels + 5 * line_size);
+    AV_COPY128U(block + 6 * 8, pixels + 6 * line_size);
+    AV_COPY128U(block + 7 * 8, pixels + 7 * line_size);
+}
+
+static void get_pixels_8_c(int16_t *av_restrict block, const uint8_t *pixels,
+                           ptrdiff_t line_size)
+{
+    int i;
 
-#define BIT_DEPTH 8
-#include "pixblockdsp_template.c"
+    /* read the pixels */
+    for (i = 0; i < 8; i++) {
+        block[0] = pixels[0];
+        block[1] = pixels[1];
+        block[2] = pixels[2];
+        block[3] = pixels[3];
+        block[4] = pixels[4];
+        block[5] = pixels[5];
+        block[6] = pixels[6];
+        block[7] = pixels[7];
+        pixels  += line_size;
+        block   += 8;
+    }
+}
 
 static void diff_pixels_c(int16_t *av_restrict block, const uint8_t *s1,
                           const uint8_t *s2, int stride)