Chiebot-Mirror
/
FFmpeg
mirror of https://github.com/FFmpeg/FFmpeg.git
8
0
Fork 0
Code Issues Projects Releases Wiki Activity

VP8: always_inline some things to force gcc to do the right thing

Browse Source 
Mostly seems to help in the MC code, which gets a hundred cycles faster.

Originally committed as revision 24463 to svn://svn.ffmpeg.org/ffmpeg/trunk
oldabi
Jason Garrett-Glaser 15 years ago
parent 06d50ca804
commit 414ac27d8f
1 changed files with 43 additions and 32 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Unified View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 47
      libavcodec/vp8.c

47
libavcodec/vp8.c
Unescape View File

@ -521,8 +521,8 @@ static int decode_frame_header(VP8Context *s, const uint8_t *buf, int buf_size)
return 0; return 0;
} }
static inline void clamp_mv(VP8Context *s, VP56mv *dst, const VP56mv *src, static av_always_inline
int mb_x, int mb_y) void clamp_mv(VP8Context *s, VP56mv *dst, const VP56mv *src, int mb_x, int mb_y)
{ {
#define MARGIN (16 << 2) #define MARGIN (16 << 2)
dst->x = av_clip(src->x, -((mb_x << 6) + MARGIN), dst->x = av_clip(src->x, -((mb_x << 6) + MARGIN),
@ -531,7 +531,8 @@ static inline void clamp_mv(VP8Context *s, VP56mv *dst, const VP56mv *src,
((s->mb_height - 1 - mb_y) << 6) + MARGIN); ((s->mb_height - 1 - mb_y) << 6) + MARGIN);
} }
static void find_near_mvs(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, static av_always_inline
void find_near_mvs(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y,
VP56mv near[2], VP56mv *best, uint8_t cnt[4]) VP56mv near[2], VP56mv *best, uint8_t cnt[4])
{ {
VP8Macroblock *mb_edge[3] = { mb + 2 /* top */, VP8Macroblock *mb_edge[3] = { mb + 2 /* top */,
@ -614,7 +615,8 @@ static int read_mv_component(VP56RangeCoder *c, const uint8_t *p)
return (x && vp56_rac_get_prob(c, p[1])) ? -x : x; return (x && vp56_rac_get_prob(c, p[1])) ? -x : x;
} }
static const uint8_t *get_submv_prob(uint32_t left, uint32_t top) static av_always_inline
const uint8_t *get_submv_prob(uint32_t left, uint32_t top)
{ {
if (left == top) if (left == top)
return vp8_submv_prob[4-!!left]; return vp8_submv_prob[4-!!left];
@ -627,7 +629,8 @@ static const uint8_t *get_submv_prob(uint32_t left, uint32_t top)
* Split motion vector prediction, 16.4. * Split motion vector prediction, 16.4.
* @returns the number of motion vectors parsed (2, 4 or 16) * @returns the number of motion vectors parsed (2, 4 or 16)
*/ */
static int decode_splitmvs(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb) static av_always_inline
int decode_splitmvs(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb)
{ {
int part_idx = mb->partitioning = int part_idx = mb->partitioning =
vp8_rac_get_tree(c, vp8_mbsplit_tree, vp8_mbsplit_prob); vp8_rac_get_tree(c, vp8_mbsplit_tree, vp8_mbsplit_prob);
@ -678,7 +681,8 @@ static int decode_splitmvs(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb)
return num; return num;
} }
static inline void decode_intra4x4_modes(VP56RangeCoder *c, uint8_t *intra4x4, static av_always_inline
void decode_intra4x4_modes(VP56RangeCoder *c, uint8_t *intra4x4,
int stride, int keyframe) int stride, int keyframe)
{ {
int x, y, t, l, i; int x, y, t, l, i;
@ -700,7 +704,8 @@ static inline void decode_intra4x4_modes(VP56RangeCoder *c, uint8_t *intra4x4,
} }
} }
static void decode_mb_mode(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, static av_always_inline
void decode_mb_mode(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y,
uint8_t *intra4x4, uint8_t *segment) uint8_t *intra4x4, uint8_t *segment)
{ {
VP56RangeCoder *c = &s->c; VP56RangeCoder *c = &s->c;
@ -827,7 +832,8 @@ static int decode_block_coeffs(VP56RangeCoder *c, DCTELEM block[16],
return nonzero; return nonzero;
} }
static void decode_mb_coeffs(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb, static av_always_inline
void decode_mb_coeffs(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb,
uint8_t t_nnz[9], uint8_t l_nnz[9]) uint8_t t_nnz[9], uint8_t l_nnz[9])
{ {
LOCAL_ALIGNED_16(DCTELEM, dc,[16]); LOCAL_ALIGNED_16(DCTELEM, dc,[16]);
@ -925,7 +931,8 @@ void xchg_mb_border(uint8_t *top_border, uint8_t *src_y, uint8_t *src_cb, uint8_
} }
} }
static int check_intra_pred_mode(int mode, int mb_x, int mb_y) static av_always_inline
int check_intra_pred_mode(int mode, int mb_x, int mb_y)
{ {
if (mode == DC_PRED8x8) { if (mode == DC_PRED8x8) {
if (!mb_x) { if (!mb_x) {
@ -937,7 +944,8 @@ static int check_intra_pred_mode(int mode, int mb_x, int mb_y)
return mode; return mode;
} }
static void intra_predict(VP8Context *s, uint8_t *dst[3], VP8Macroblock *mb, static av_always_inline
void intra_predict(VP8Context *s, uint8_t *dst[3], VP8Macroblock *mb,
uint8_t *intra4x4, int mb_x, int mb_y) uint8_t *intra4x4, int mb_x, int mb_y)
{ {
int x, y, mode, nnz, tr; int x, y, mode, nnz, tr;
@ -1020,7 +1028,8 @@ static void intra_predict(VP8Context *s, uint8_t *dst[3], VP8Macroblock *mb,
* @param linesize size of a single line of plane data, including padding * @param linesize size of a single line of plane data, including padding
* @param mc_func motion compensation function pointers (bilinear or sixtap MC) * @param mc_func motion compensation function pointers (bilinear or sixtap MC)
*/ */
static inline void vp8_mc(VP8Context *s, int luma, static av_always_inline
void vp8_mc(VP8Context *s, int luma,
uint8_t *dst, uint8_t *src, const VP56mv *mv, uint8_t *dst, uint8_t *src, const VP56mv *mv,
int x_off, int y_off, int block_w, int block_h, int x_off, int y_off, int block_w, int block_h,
int width, int height, int linesize, int width, int height, int linesize,
@ -1048,7 +1057,8 @@ static inline void vp8_mc(VP8Context *s, int luma,
mc_func[0][0](dst, linesize, src + y_off * linesize + x_off, linesize, block_h, 0, 0); mc_func[0][0](dst, linesize, src + y_off * linesize + x_off, linesize, block_h, 0, 0);
} }
static inline void vp8_mc_part(VP8Context *s, uint8_t *dst[3], static av_always_inline
void vp8_mc_part(VP8Context *s, uint8_t *dst[3],
AVFrame *ref_frame, int x_off, int y_off, AVFrame *ref_frame, int x_off, int y_off,
int bx_off, int by_off, int bx_off, int by_off,
int block_w, int block_h, int block_w, int block_h,
@ -1083,7 +1093,7 @@ static inline void vp8_mc_part(VP8Context *s, uint8_t *dst[3],
/* Fetch pixels for estimated mv 4 macroblocks ahead. /* Fetch pixels for estimated mv 4 macroblocks ahead.
* Optimized for 64-byte cache lines. Inspired by ffh264 prefetch_motion. */ * Optimized for 64-byte cache lines. Inspired by ffh264 prefetch_motion. */
static inline void prefetch_motion(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, int mb_xy, int ref) static av_always_inline void prefetch_motion(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, int mb_xy, int ref)
{ {
/* Don't prefetch refs that haven't been used very often this frame. */ /* Don't prefetch refs that haven't been used very often this frame. */
if (s->ref_count[ref-1] > (mb_xy >> 5)) { if (s->ref_count[ref-1] > (mb_xy >> 5)) {
@ -1101,7 +1111,8 @@ static inline void prefetch_motion(VP8Context *s, VP8Macroblock *mb, int mb_x, i
/** /**
* Apply motion vectors to prediction buffer, chapter 18. * Apply motion vectors to prediction buffer, chapter 18.
*/ */
static void inter_predict(VP8Context *s, uint8_t *dst[3], VP8Macroblock *mb, static av_always_inline
void inter_predict(VP8Context *s, uint8_t *dst[3], VP8Macroblock *mb,
int mb_x, int mb_y) int mb_x, int mb_y)
{ {
int x_off = mb_x << 4, y_off = mb_y << 4; int x_off = mb_x << 4, y_off = mb_y << 4;
@ -1185,7 +1196,7 @@ static void inter_predict(VP8Context *s, uint8_t *dst[3], VP8Macroblock *mb,
} }
} }
static void idct_mb(VP8Context *s, uint8_t *dst[3], VP8Macroblock *mb) static av_always_inline void idct_mb(VP8Context *s, uint8_t *dst[3], VP8Macroblock *mb)
{ {
int x, y, ch; int x, y, ch;
@ -1236,7 +1247,7 @@ static void idct_mb(VP8Context *s, uint8_t *dst[3], VP8Macroblock *mb)
} }
} }
static void filter_level_for_mb(VP8Context *s, VP8Macroblock *mb, VP8FilterStrength *f ) static av_always_inline void filter_level_for_mb(VP8Context *s, VP8Macroblock *mb, VP8FilterStrength *f )
{ {
int interior_limit, filter_level; int interior_limit, filter_level;
@ -1276,7 +1287,7 @@ static void filter_level_for_mb(VP8Context *s, VP8Macroblock *mb, VP8FilterStren
f->inner_filter = !mb->skip || mb->mode == MODE_I4x4 || mb->mode == VP8_MVMODE_SPLIT; f->inner_filter = !mb->skip || mb->mode == MODE_I4x4 || mb->mode == VP8_MVMODE_SPLIT;
} }
static void filter_mb(VP8Context *s, uint8_t *dst[3], VP8FilterStrength *f, int mb_x, int mb_y) static av_always_inline void filter_mb(VP8Context *s, uint8_t *dst[3], VP8FilterStrength *f, int mb_x, int mb_y)
{ {
int mbedge_lim, bedge_lim, hev_thresh; int mbedge_lim, bedge_lim, hev_thresh;
int filter_level = f->filter_level; int filter_level = f->filter_level;
@ -1345,7 +1356,7 @@ static void filter_mb(VP8Context *s, uint8_t *dst[3], VP8FilterStrength *f, int
} }
} }
static void filter_mb_simple(VP8Context *s, uint8_t *dst, VP8FilterStrength *f, int mb_x, int mb_y) static av_always_inline void filter_mb_simple(VP8Context *s, uint8_t *dst, VP8FilterStrength *f, int mb_x, int mb_y)
{ {
int mbedge_lim, bedge_lim; int mbedge_lim, bedge_lim;
int filter_level = f->filter_level; int filter_level = f->filter_level;

Write Preview
Loading…
Cancel
Save