Commit f14bdd8e authored by Jason Garrett-Glaser's avatar Jason Garrett-Glaser

H.264: Partially inline CABAC residual decoding

Improves CABAC performance about ~1.2%.

Trick originates from x264 and has also been used in ffvp8.  It's useful because
coded block flags are usually zero, so it helps to have the early termination
inlined into the main function.

Originally committed as revision 26375 to svn://svn.ffmpeg.org/ffmpeg/trunk
parent a242ac36
......@@ -1049,29 +1049,6 @@ static av_always_inline void decode_cabac_residual_internal( H264Context *h, DCT
#define CC &h->cabac
#endif
/* cat: 0-> DC 16x16 n = 0
* 1-> AC 16x16 n = luma4x4idx
* 2-> Luma4x4 n = luma4x4idx
* 3-> DC Chroma n = iCbCr
* 4-> AC Chroma n = 16 + 4 * iCbCr + chroma4x4idx
* 5-> Luma8x8 n = 4 * luma8x8idx
*/
/* read coded block flag */
if( is_dc || cat != 5 ) {
if( get_cabac( CC, &h->cabac_state[85 + get_cabac_cbf_ctx( h, cat, n, is_dc ) ] ) == 0 ) {
h->non_zero_count_cache[scan8[n]] = 0;
#ifdef CABAC_ON_STACK
h->cabac.range = cc.range ;
h->cabac.low = cc.low ;
h->cabac.bytestream= cc.bytestream;
#endif
return;
}
}
significant_coeff_ctx_base = h->cabac_state
+ significant_coeff_flag_offset[MB_FIELD][cat];
last_coeff_ctx_base = h->cabac_state
......@@ -1172,19 +1149,50 @@ static av_always_inline void decode_cabac_residual_internal( H264Context *h, DCT
}
static void decode_cabac_residual_dc( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, int max_coeff ) {
static void decode_cabac_residual_dc_internal( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, int max_coeff ) {
decode_cabac_residual_internal(h, block, cat, n, scantable, NULL, max_coeff, 1);
}
static void decode_cabac_residual_nondc( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, const uint32_t *qmul, int max_coeff ) {
static void decode_cabac_residual_nondc_internal( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, const uint32_t *qmul, int max_coeff ) {
decode_cabac_residual_internal(h, block, cat, n, scantable, qmul, max_coeff, 0);
}
/* cat: 0-> DC 16x16 n = 0
* 1-> AC 16x16 n = luma4x4idx
* 2-> Luma4x4 n = luma4x4idx
* 3-> DC Chroma n = iCbCr
* 4-> AC Chroma n = 16 + 4 * iCbCr + chroma4x4idx
* 5-> Luma8x8 n = 4 * luma8x8idx */
/* Partially inline the CABAC residual decode: inline the coded block flag.
* This has very little impact on binary size and improves performance
* because it allows improved constant propagation into get_cabac_cbf_ctx,
* as well as because most blocks have zero CBFs. */
static av_always_inline void decode_cabac_residual_dc( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, int max_coeff ) {
/* read coded block flag */
if( get_cabac( CC, &h->cabac_state[85 + get_cabac_cbf_ctx( h, cat, n, 1 ) ] ) == 0 ) {
h->non_zero_count_cache[scan8[n]] = 0;
return;
}
decode_cabac_residual_dc_internal( h, block, cat, n, scantable, max_coeff );
}
static av_always_inline void decode_cabac_residual_nondc( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, const uint32_t *qmul, int max_coeff ) {
/* read coded block flag */
if( cat != 5 && get_cabac( CC, &h->cabac_state[85 + get_cabac_cbf_ctx( h, cat, n, 0 ) ] ) == 0 ) {
h->non_zero_count_cache[scan8[n]] = 0;
return;
}
decode_cabac_residual_nondc_internal( h, block, cat, n, scantable, qmul, max_coeff );
}
/**
* decodes a macroblock
* @return 0 if OK, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
*/
int ff_h264_decode_mb_cabac(H264Context *h) {
START_TIMER;
MpegEncContext * const s = &h->s;
int mb_xy;
int mb_type, partition_count, cbp = 0;
......@@ -1717,6 +1725,6 @@ decode_intra_mb:
h->ref_count[0] >>= 1;
h->ref_count[1] >>= 1;
}
STOP_TIMER("test");
return 0;
}
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment