h264dsp: Factorize code into a new function, h264_find_start_code_candidate

This performs the start code search which was previously part of
h264_find_frame_end() - the most CPU intensive part of the function.

By itself, this results in a performance regression:
              Before          After
              Mean   StdDev   Mean   StdDev  Change
Overall time  2925.6 26.2     3068.5 31.7    -4.7%

but this can more than be made up for by platform-optimised
implementations of the function.

Signed-off-by: Martin Storsjö <martin@martin.st>

1 files changed, 9 insertions, 0 deletions

diff --git a/libavcodec/h264dsp.h b/libavcodec/h264dsp.h
index 1f9f8fe..6249ba7 100644
--- a/libavcodec/h264dsp.h
+++ b/libavcodec/h264dsp.h
@@ -105,6 +105,15 @@ typedef struct H264DSPContext {
     /* bypass-transform */
     void (*h264_add_pixels8_clear)(uint8_t *dst, int16_t *block, int stride);
     void (*h264_add_pixels4_clear)(uint8_t *dst, int16_t *block, int stride);
+
+    /**
+     * Search buf from the start for up to size bytes. Return the index
+     * of a zero byte, or >= size if not found. Ideally, use lookahead
+     * to filter out any zero bytes that are known to not be followed by
+     * one or more further zero bytes and a one byte. Better still, filter
+     * out any bytes that form the trailing_zero_8bits syntax element too.
+     */
+    int (*h264_find_start_code_candidate)(const uint8_t *buf, int size);
 } H264DSPContext;
 
 void ff_h264dsp_init(H264DSPContext *c, const int bit_depth,