[FFmpeg-devel] [PATCH] VC1: merge put/add_pixels with IDCT8x8.
Kostya
kostya.shishkov
Sat Feb 19 18:51:39 CET 2011
On Sat, Feb 19, 2011 at 12:20:06PM -0500, Ronald S. Bultje wrote:
> ---
> libavcodec/dsputil.c | 28 +++++++-------
> libavcodec/dsputil.h | 8 ++++
> libavcodec/vc1.c | 37 ++++++++++++++++++-
> libavcodec/vc1dec.c | 101 ++++++++++++++++++++++---------------------------
> libavcodec/vc1dsp.c | 58 ++++++++++++++++++++++++-----
> libavcodec/vc1dsp.h | 8 +++-
> 6 files changed, 158 insertions(+), 82 deletions(-)
>
> diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c
> index f7151b8..1f7bd4c 100644
> --- a/libavcodec/dsputil.c
> +++ b/libavcodec/dsputil.c
> @@ -437,8 +437,8 @@ static void diff_pixels_c(DCTELEM *restrict block, const uint8_t *s1,
> }
>
>
> -static void put_pixels_clamped_c(const DCTELEM *block, uint8_t *restrict pixels,
> - int line_size)
> +void ff_put_pixels_clamped_c(const DCTELEM *block, uint8_t *restrict pixels,
> + int line_size)
> {
> int i;
> uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
> @@ -493,9 +493,9 @@ static void put_pixels_clamped2_c(const DCTELEM *block, uint8_t *restrict pixels
> }
> }
>
> -static void put_signed_pixels_clamped_c(const DCTELEM *block,
> - uint8_t *restrict pixels,
> - int line_size)
> +void ff_put_signed_pixels_clamped_c(const DCTELEM *block,
> + uint8_t *restrict pixels,
> + int line_size)
> {
> int i, j;
>
> @@ -535,8 +535,8 @@ static void put_pixels_nonclamped_c(const DCTELEM *block, uint8_t *restrict pixe
> }
> }
>
> -static void add_pixels_clamped_c(const DCTELEM *block, uint8_t *restrict pixels,
> - int line_size)
> +void ff_add_pixels_clamped_c(const DCTELEM *block, uint8_t *restrict pixels,
> + int line_size)
> {
> int i;
> uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
> @@ -3961,22 +3961,22 @@ void ff_wmv2_idct_c(short * block){
> static void ff_wmv2_idct_put_c(uint8_t *dest, int line_size, DCTELEM *block)
> {
> ff_wmv2_idct_c(block);
> - put_pixels_clamped_c(block, dest, line_size);
> + ff_put_pixels_clamped_c(block, dest, line_size);
> }
> static void ff_wmv2_idct_add_c(uint8_t *dest, int line_size, DCTELEM *block)
> {
> ff_wmv2_idct_c(block);
> - add_pixels_clamped_c(block, dest, line_size);
> + ff_add_pixels_clamped_c(block, dest, line_size);
> }
> static void ff_jref_idct_put(uint8_t *dest, int line_size, DCTELEM *block)
> {
> j_rev_dct (block);
> - put_pixels_clamped_c(block, dest, line_size);
> + ff_put_pixels_clamped_c(block, dest, line_size);
> }
> static void ff_jref_idct_add(uint8_t *dest, int line_size, DCTELEM *block)
> {
> j_rev_dct (block);
> - add_pixels_clamped_c(block, dest, line_size);
> + ff_add_pixels_clamped_c(block, dest, line_size);
> }
>
> static void ff_jref_idct4_put(uint8_t *dest, int line_size, DCTELEM *block)
> @@ -4135,10 +4135,10 @@ av_cold void dsputil_init(DSPContext* c, AVCodecContext *avctx)
>
> c->get_pixels = get_pixels_c;
> c->diff_pixels = diff_pixels_c;
> - c->put_pixels_clamped = put_pixels_clamped_c;
> - c->put_signed_pixels_clamped = put_signed_pixels_clamped_c;
> + c->put_pixels_clamped = ff_put_pixels_clamped_c;
> + c->put_signed_pixels_clamped = ff_put_signed_pixels_clamped_c;
> c->put_pixels_nonclamped = put_pixels_nonclamped_c;
> - c->add_pixels_clamped = add_pixels_clamped_c;
> + c->add_pixels_clamped = ff_add_pixels_clamped_c;
> c->add_pixels8 = add_pixels8_c;
> c->add_pixels4 = add_pixels4_c;
> c->sum_abs_dctelem = sum_abs_dctelem_c;
> diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h
> index f97b2b5..ac5b7e0 100644
> --- a/libavcodec/dsputil.h
> +++ b/libavcodec/dsputil.h
> @@ -198,6 +198,14 @@ void ff_emulated_edge_mc(uint8_t *buf, const uint8_t *src, int linesize,
> int block_w, int block_h,
> int src_x, int src_y, int w, int h);
>
> +void ff_put_pixels_clamped_c(const DCTELEM *block, uint8_t *restrict ptr,
> + int linesize);
> +void ff_put_signed_pixels_clamped_c(const DCTELEM *block, uint8_t *restrict ptr,
> + int linesize);
> +void ff_add_pixels_clamped_c(const DCTELEM *block, uint8_t *restrict ptr,
> + int linesize);
> +
> +
> /**
> * DSPContext.
> */
this part looks quite independent, maybe nominate it for separate commit?
> diff --git a/libavcodec/vc1.c b/libavcodec/vc1.c
> index 8bd6647..d73a2c9 100644
> --- a/libavcodec/vc1.c
> +++ b/libavcodec/vc1.c
> @@ -30,6 +30,7 @@
> #include "avcodec.h"
> #include "mpegvideo.h"
> #include "vc1.h"
> +#include "vc1dsp.h"
> #include "vc1data.h"
> #include "msmpeg4data.h"
> #include "unary.h"
> @@ -280,6 +281,36 @@ static int vop_dquant_decoding(VC1Context *v)
>
> static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb);
>
> +static void simple_idct_put_signed(uint8_t *dest, int line_size, DCTELEM *block)
> +{
> + ff_simple_idct(block);
> + ff_put_signed_pixels_clamped_c(block, dest, line_size);
> +}
> +
> +static void simple_idct_put_signed_rangered(uint8_t *dest, int line_size, DCTELEM *block)
> +{
> + int i;
> +
> + ff_simple_idct(block);
> + for (i = 0; i < 64; i++) block[i] <<= 1;
> + ff_put_signed_pixels_clamped_c(block, dest, line_size);
> +}
> +
> +static void simple_idct_put(uint8_t *dest, int line_size, DCTELEM *block)
> +{
> + ff_simple_idct(block);
> + ff_put_pixels_clamped_c(block, dest, line_size);
> +}
> +
> +static void simple_idct_put_rangered(uint8_t *dest, int line_size, DCTELEM *block)
> +{
> + int i;
> +
> + ff_simple_idct(block);
> + for (i = 0; i < 64; i++) block[i] <<= 1;
> + ff_put_pixels_clamped_c(block, dest, line_size);
> +}
> +
> /**
> * Decode Simple/Main Profiles sequence header
> * @see Figure 7-8, p16-17
> @@ -337,7 +368,11 @@ int vc1_decode_sequence_header(AVCodecContext *avctx, VC1Context *v, GetBitConte
> v->res_fasttx = get_bits1(gb);
> if (!v->res_fasttx)
> {
> - v->vc1dsp.vc1_inv_trans_8x8 = ff_simple_idct;
> + v->vc1dsp.vc1_inv_trans_8x8_add = ff_simple_idct_add;
> + v->vc1dsp.vc1_inv_trans_8x8_put[0] = simple_idct_put_signed;
> + v->vc1dsp.vc1_inv_trans_8x8_put[1] = simple_idct_put_signed_rangered;
> + v->vc1dsp.vc1_inv_trans_8x8_put[2] = simple_idct_put;
> + v->vc1dsp.vc1_inv_trans_8x8_put[3] = simple_idct_put_rangered;
> v->vc1dsp.vc1_inv_trans_8x4 = ff_simple_idct84_add;
> v->vc1dsp.vc1_inv_trans_4x8 = ff_simple_idct48_add;
> v->vc1dsp.vc1_inv_trans_4x4 = ff_simple_idct44_add;
> diff --git a/libavcodec/vc1dec.c b/libavcodec/vc1dec.c
> index 7eb9576..1704bcd 100644
> --- a/libavcodec/vc1dec.c
> +++ b/libavcodec/vc1dec.c
> @@ -187,39 +187,6 @@ static void vc1_loop_filter_iblk(VC1Context *v, int pq)
> }
> }
>
> -/** Put block onto picture
> - */
> -static void vc1_put_block(VC1Context *v, DCTELEM block[6][64])
> -{
> - uint8_t *Y;
> - int ys, us, vs;
> - DSPContext *dsp = &v->s.dsp;
> -
> - if(v->rangeredfrm) {
> - int i, j, k;
> - for(k = 0; k < 6; k++)
> - for(j = 0; j < 8; j++)
> - for(i = 0; i < 8; i++)
> - block[k][i + j*8] = (block[k][i + j*8] - 64) << 1;
> -
> - }
> - ys = v->s.current_picture.linesize[0];
> - us = v->s.current_picture.linesize[1];
> - vs = v->s.current_picture.linesize[2];
> - Y = v->s.dest[0];
> -
> - dsp->put_pixels_clamped(block[0], Y, ys);
> - dsp->put_pixels_clamped(block[1], Y + 8, ys);
> - Y += ys * 8;
> - dsp->put_pixels_clamped(block[2], Y, ys);
> - dsp->put_pixels_clamped(block[3], Y + 8, ys);
> -
> - if(!(v->s.flags & CODEC_FLAG_GRAY)) {
> - dsp->put_pixels_clamped(block[4], v->s.dest[1], us);
> - dsp->put_pixels_clamped(block[5], v->s.dest[2], vs);
> - }
> -}
> -
> /** Do motion compensation over 1 macroblock
> * Mostly adapted hpel_motion and qpel_motion from mpegvideo.c
> */
> @@ -2042,8 +2009,7 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan
> if(i==1)
> v->vc1dsp.vc1_inv_trans_8x8_dc(dst, linesize, block);
> else{
> - v->vc1dsp.vc1_inv_trans_8x8(block);
> - s->dsp.add_pixels_clamped(block, dst, linesize);
> + v->vc1dsp.vc1_inv_trans_8x8_add(dst, linesize, block);
> }
> if(apply_filter && cbp_top & 0xC)
> v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);
> @@ -2150,7 +2116,7 @@ static int vc1_decode_p_mb(VC1Context *v)
> {
> MpegEncContext *s = &v->s;
> GetBitContext *gb = &s->gb;
> - int i, j;
> + int i;
> int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
> int cbp; /* cbp decoding stuff */
> int mqdiff, mquant; /* MB quantization */
> @@ -2165,6 +2131,7 @@ static int vc1_decode_p_mb(VC1Context *v)
> int skipped, fourmv;
> int block_cbp = 0, pat;
> int apply_loop_filter;
> + vc1_idct_func idct8x8_fn;
>
> mquant = v->pq; /* Loosy initialization */
>
> @@ -2216,6 +2183,7 @@ static int vc1_decode_p_mb(VC1Context *v)
> VC1_TTMB_VLC_BITS, 2);
> if(!s->mb_intra) vc1_mc_1mv(v, 0);
> dst_idx = 0;
> + idct8x8_fn = v->vc1dsp.vc1_inv_trans_8x8_put[!!v->rangeredfrm];
> for (i=0; i<6; i++)
> {
> s->dc_val[0][s->block_index[i]] = 0;
> @@ -2233,9 +2201,9 @@ static int vc1_decode_p_mb(VC1Context *v)
>
> vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
> if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
> - v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);
> - if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
> - s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);
> + idct8x8_fn(s->dest[dst_idx] + off,
> + i & 4 ? s->uvlinesize : s->linesize,
> + s->block[i]);
> if(v->pq >= 9 && v->overlap) {
> if(v->c_avail)
> v->vc1dsp.vc1_h_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);
> @@ -2349,6 +2317,7 @@ static int vc1_decode_p_mb(VC1Context *v)
> }
> if (!v->ttmbf && coded_inter)
> ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);
> + idct8x8_fn = v->vc1dsp.vc1_inv_trans_8x8_put[!!v->rangeredfrm];
> for (i=0; i<6; i++)
> {
> dst_idx += i >> 2;
> @@ -2364,9 +2333,9 @@ static int vc1_decode_p_mb(VC1Context *v)
>
> vc1_decode_intra_block(v, s->block[i], i, is_coded[i], mquant, (i&4)?v->codingset2:v->codingset);
> if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
> - v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);
> - if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
> - s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize);
> + idct8x8_fn(s->dest[dst_idx] + off,
> + i & 4 ? s->uvlinesize : s->linesize,
> + s->block[i]);
> if(v->pq >= 9 && v->overlap) {
> if(v->c_avail)
> v->vc1dsp.vc1_h_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);
> @@ -2442,7 +2411,7 @@ static void vc1_decode_b_mb(VC1Context *v)
> {
> MpegEncContext *s = &v->s;
> GetBitContext *gb = &s->gb;
> - int i, j;
> + int i;
> int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
> int cbp = 0; /* cbp decoding stuff */
> int mqdiff, mquant; /* MB quantization */
> @@ -2455,6 +2424,7 @@ static void vc1_decode_b_mb(VC1Context *v)
> int skipped, direct;
> int dmv_x[2], dmv_y[2];
> int bmvtype = BMV_TYPE_BACKWARD;
> + vc1_idct_func idct8x8_fn;
>
> mquant = v->pq; /* Loosy initialization */
> s->mb_intra = 0;
> @@ -2552,6 +2522,7 @@ static void vc1_decode_b_mb(VC1Context *v)
> }
> }
> dst_idx = 0;
> + idct8x8_fn = v->vc1dsp.vc1_inv_trans_8x8_put[!!v->rangeredfrm];
> for (i=0; i<6; i++)
> {
> s->dc_val[0][s->block_index[i]] = 0;
> @@ -2569,9 +2540,9 @@ static void vc1_decode_b_mb(VC1Context *v)
>
> vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
> if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;
> - v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);
> - if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;
> - s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);
> + idct8x8_fn(s->dest[dst_idx] + off,
> + i & 4 ? s->uvlinesize : s->linesize,
> + s->block[i]);
> } else if(val) {
> vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), 0, 0, 0);
> if(!v->ttmbf && ttmb < 8) ttmb = -1;
> @@ -2584,11 +2555,12 @@ static void vc1_decode_b_mb(VC1Context *v)
> */
> static void vc1_decode_i_blocks(VC1Context *v)
> {
> - int k, j;
> + int k;
> MpegEncContext *s = &v->s;
> int cbp, val;
> uint8_t *coded_val;
> int mb_pos;
> + vc1_idct_func idct8x8_fn;
>
> /* select codingmode used for VLC tables selection */
> switch(v->y_ac_table_index){
> @@ -2623,11 +2595,19 @@ static void vc1_decode_i_blocks(VC1Context *v)
> s->mb_x = s->mb_y = 0;
> s->mb_intra = 1;
> s->first_slice_line = 1;
> + idct8x8_fn = v->vc1dsp.vc1_inv_trans_8x8_put[(!!v->rangeredfrm) | ((v->pq < 9 || !v->overlap) << 1)];
> for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
> s->mb_x = 0;
> ff_init_block_index(s);
> for(; s->mb_x < s->mb_width; s->mb_x++) {
> + uint8_t *dst[6];
> ff_update_block_index(s);
> + dst[0] = s->dest[0];
> + dst[1] = dst[0] + 8;
> + dst[2] = dst[0] + s->linesize * 8;
> + dst[3] = dst[2] + 8;
> + dst[4] = s->dest[1];
> + dst[5] = s->dest[2];
> s->dsp.clear_blocks(s->block[0]);
> mb_pos = s->mb_x + s->mb_y * s->mb_width;
> s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
> @@ -2651,13 +2631,12 @@ static void vc1_decode_i_blocks(VC1Context *v)
>
> vc1_decode_i_block(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2);
>
> - v->vc1dsp.vc1_inv_trans_8x8(s->block[k]);
> - if(v->pq >= 9 && v->overlap) {
> - for(j = 0; j < 64; j++) s->block[k][j] += 128;
> - }
> + if(k > 3 && (v->s.flags & CODEC_FLAG_GRAY)) continue;
> + idct8x8_fn(dst[k],
> + k & 4 ? s->uvlinesize : s->linesize,
> + s->block[k]);
> }
>
> - vc1_put_block(v, s->block);
> if(v->pq >= 9 && v->overlap) {
> if(s->mb_x) {
> v->vc1dsp.vc1_h_overlap(s->dest[0], s->linesize);
> @@ -2704,7 +2683,7 @@ static void vc1_decode_i_blocks(VC1Context *v)
> */
> static void vc1_decode_i_blocks_adv(VC1Context *v)
> {
> - int k, j;
> + int k;
> MpegEncContext *s = &v->s;
> int cbp, val;
> uint8_t *coded_val;
> @@ -2713,6 +2692,7 @@ static void vc1_decode_i_blocks_adv(VC1Context *v)
> int mqdiff;
> int overlap;
> GetBitContext *gb = &s->gb;
> + vc1_idct_func idct8x8_fn;
>
> /* select codingmode used for VLC tables selection */
> switch(v->y_ac_table_index){
> @@ -2743,11 +2723,19 @@ static void vc1_decode_i_blocks_adv(VC1Context *v)
> s->mb_x = s->mb_y = 0;
> s->mb_intra = 1;
> s->first_slice_line = 1;
> + idct8x8_fn = v->vc1dsp.vc1_inv_trans_8x8_put[!!v->rangeredfrm];
> for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
> s->mb_x = 0;
> ff_init_block_index(s);
> for(;s->mb_x < s->mb_width; s->mb_x++) {
> + uint8_t *dst[6];
> ff_update_block_index(s);
> + dst[0] = s->dest[0];
> + dst[1] = dst[0] + 8;
> + dst[2] = dst[0] + s->linesize * 8;
> + dst[3] = dst[2] + 8;
> + dst[4] = s->dest[1];
> + dst[5] = s->dest[2];
> s->dsp.clear_blocks(s->block[0]);
> mb_pos = s->mb_x + s->mb_y * s->mb_stride;
> s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;
> @@ -2791,11 +2779,12 @@ static void vc1_decode_i_blocks_adv(VC1Context *v)
>
> vc1_decode_i_block_adv(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2, mquant);
>
> - v->vc1dsp.vc1_inv_trans_8x8(s->block[k]);
> - for(j = 0; j < 64; j++) s->block[k][j] += 128;
> + if(k > 3 && (v->s.flags & CODEC_FLAG_GRAY)) continue;
> + idct8x8_fn(dst[k],
> + k & 4 ? s->uvlinesize : s->linesize,
> + s->block[k]);
> }
>
> - vc1_put_block(v, s->block);
> if(overlap) {
> if(s->mb_x) {
> v->vc1dsp.vc1_h_overlap(s->dest[0], s->linesize);
> diff --git a/libavcodec/vc1dsp.c b/libavcodec/vc1dsp.c
> index 000dad7..ea1c840 100644
> --- a/libavcodec/vc1dsp.c
> +++ b/libavcodec/vc1dsp.c
> @@ -199,7 +199,7 @@ static void vc1_inv_trans_8x8_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
> }
> }
>
> -static void vc1_inv_trans_8x8_c(DCTELEM block[64])
> +static av_always_inline void vc1_inv_trans_8x8_c(DCTELEM block[64], int shl)
> {
> int i;
> register int t1,t2,t3,t4,t5,t6,t7,t8;
> @@ -254,20 +254,54 @@ static void vc1_inv_trans_8x8_c(DCTELEM block[64])
> t3 = 9 * src[ 8] - 16 * src[24] + 4 * src[40] + 15 * src[56];
> t4 = 4 * src[ 8] - 9 * src[24] + 15 * src[40] - 16 * src[56];
>
> - dst[ 0] = (t5 + t1) >> 7;
> - dst[ 8] = (t6 + t2) >> 7;
> - dst[16] = (t7 + t3) >> 7;
> - dst[24] = (t8 + t4) >> 7;
> - dst[32] = (t8 - t4 + 1) >> 7;
> - dst[40] = (t7 - t3 + 1) >> 7;
> - dst[48] = (t6 - t2 + 1) >> 7;
> - dst[56] = (t5 - t1 + 1) >> 7;
> + dst[ 0] = ((t5 + t1 ) >> 7) << shl;
> + dst[ 8] = ((t6 + t2 ) >> 7) << shl;
> + dst[16] = ((t7 + t3 ) >> 7) << shl;
> + dst[24] = ((t8 + t4 ) >> 7) << shl;
> + dst[32] = ((t8 - t4 + 1) >> 7) << shl;
> + dst[40] = ((t7 - t3 + 1) >> 7) << shl;
> + dst[48] = ((t6 - t2 + 1) >> 7) << shl;
> + dst[56] = ((t5 - t1 + 1) >> 7) << shl;
>
> src++;
> dst++;
> }
> }
>
> +static void vc1_inv_trans_8x8_add_c(uint8_t *dst, int linesize, DCTELEM block[64])
> +{
> + vc1_inv_trans_8x8_c(block, 0);
> + ff_add_pixels_clamped_c(block, dst, linesize);
> +}
> +
> +static void vc1_inv_trans_8x8_put_c(uint8_t *dst, int linesize,
> + DCTELEM block[64])
> +{
> + vc1_inv_trans_8x8_c(block, 0);
> + ff_put_pixels_clamped_c(block, dst, linesize);
> +}
> +
> +static void vc1_inv_trans_8x8_put_rangered_c(uint8_t *dst, int linesize,
> + DCTELEM block[64])
> +{
> + vc1_inv_trans_8x8_c(block, 1);
> + ff_put_pixels_clamped_c(block, dst, linesize);
> +}
> +
> +static void vc1_inv_trans_8x8_put_signed_c(uint8_t *dst, int linesize,
> + DCTELEM block[64])
> +{
> + vc1_inv_trans_8x8_c(block, 0);
> + ff_put_signed_pixels_clamped_c(block, dst, linesize);
> +}
> +
> +static void vc1_inv_trans_8x8_put_signed_rangered_c(uint8_t *dst, int linesize,
> + DCTELEM block[64])
> +{
> + vc1_inv_trans_8x8_c(block, 1);
> + ff_put_signed_pixels_clamped_c(block, dst, linesize);
> +}
> +
> /** Do inverse transform on 8x4 part of block
> */
> static void vc1_inv_trans_8x4_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
> @@ -662,7 +696,11 @@ static void avg_no_rnd_vc1_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*a
> }
>
> av_cold void ff_vc1dsp_init(VC1DSPContext* dsp) {
> - dsp->vc1_inv_trans_8x8 = vc1_inv_trans_8x8_c;
> + dsp->vc1_inv_trans_8x8_put[0] = vc1_inv_trans_8x8_put_signed_c;
> + dsp->vc1_inv_trans_8x8_put[1] = vc1_inv_trans_8x8_put_signed_rangered_c;
> + dsp->vc1_inv_trans_8x8_put[2] = vc1_inv_trans_8x8_put_c;
> + dsp->vc1_inv_trans_8x8_put[3] = vc1_inv_trans_8x8_put_rangered_c;
> + dsp->vc1_inv_trans_8x8_add = vc1_inv_trans_8x8_add_c;
> dsp->vc1_inv_trans_4x8 = vc1_inv_trans_4x8_c;
> dsp->vc1_inv_trans_8x4 = vc1_inv_trans_8x4_c;
> dsp->vc1_inv_trans_4x4 = vc1_inv_trans_4x4_c;
> diff --git a/libavcodec/vc1dsp.h b/libavcodec/vc1dsp.h
> index a1f3d90..aac03ac 100644
> --- a/libavcodec/vc1dsp.h
> +++ b/libavcodec/vc1dsp.h
> @@ -30,9 +30,15 @@
>
> #include "dsputil.h"
>
> +typedef void (*vc1_idct_func)(uint8_t *dst, int linesize, DCTELEM *block);
> +
> typedef struct VC1DSPContext {
> /* vc1 functions */
> - void (*vc1_inv_trans_8x8)(DCTELEM *b);
> + void (*vc1_inv_trans_8x8_add)(uint8_t *dest, int line_size, DCTELEM *b);
> + /* [0] is default idct, [1] is range reduction which does <<= 1 for each
> + * coeff after the idct itself. [2]/[3] are the same, but use
> + * put_pixels_clamped() instead of put_signed_pixels_clamped(). */
> + vc1_idct_func vc1_inv_trans_8x8_put[4];
> void (*vc1_inv_trans_8x4)(uint8_t *dest, int line_size, DCTELEM *block);
> void (*vc1_inv_trans_4x8)(uint8_t *dest, int line_size, DCTELEM *block);
> void (*vc1_inv_trans_4x4)(uint8_t *dest, int line_size, DCTELEM *block);
> --
> 1.7.2.1
overall looks except for that that put_signed/unsigned part
More information about the ffmpeg-devel
mailing list