* RV30/40 decoder common data
*/
+#include "libavutil/internal.h"
+
#include "avcodec.h"
#include "dsputil.h"
#include "mpegvideo.h"
#include "golomb.h"
+#include "internal.h"
#include "mathops.h"
#include "rectangle.h"
+#include "thread.h"
#include "rv34vlc.h"
#include "rv34data.h"
/** @} */ // vlc group
-
-/**
- * @name RV30/40 inverse transform functions
- * @{
- */
-
-static av_always_inline void rv34_row_transform(int temp[16], DCTELEM *block)
-{
- int i;
-
- for(i=0; i<4; i++){
- const int z0= 13*(block[i+8*0] + block[i+8*2]);
- const int z1= 13*(block[i+8*0] - block[i+8*2]);
- const int z2= 7* block[i+8*1] - 17*block[i+8*3];
- const int z3= 17* block[i+8*1] + 7*block[i+8*3];
-
- temp[4*i+0]= z0+z3;
- temp[4*i+1]= z1+z2;
- temp[4*i+2]= z1-z2;
- temp[4*i+3]= z0-z3;
- }
-}
-
-/**
- * Real Video 3.0/4.0 inverse transform
- * Code is almost the same as in SVQ3, only scaling is different.
- */
-static void rv34_inv_transform(DCTELEM *block){
- int temp[16];
- int i;
-
- rv34_row_transform(temp, block);
-
- for(i=0; i<4; i++){
- const int z0= 13*(temp[4*0+i] + temp[4*2+i]) + 0x200;
- const int z1= 13*(temp[4*0+i] - temp[4*2+i]) + 0x200;
- const int z2= 7* temp[4*1+i] - 17*temp[4*3+i];
- const int z3= 17* temp[4*1+i] + 7*temp[4*3+i];
-
- block[i*8+0]= (z0 + z3)>>10;
- block[i*8+1]= (z1 + z2)>>10;
- block[i*8+2]= (z1 - z2)>>10;
- block[i*8+3]= (z0 - z3)>>10;
- }
-
-}
-
-/**
- * RealVideo 3.0/4.0 inverse transform for DC block
- *
- * Code is almost the same as rv34_inv_transform()
- * but final coefficients are multiplied by 1.5 and have no rounding.
- */
-static void rv34_inv_transform_noround(DCTELEM *block){
- int temp[16];
- int i;
-
- rv34_row_transform(temp, block);
-
- for(i=0; i<4; i++){
- const int z0= 13*(temp[4*0+i] + temp[4*2+i]);
- const int z1= 13*(temp[4*0+i] - temp[4*2+i]);
- const int z2= 7* temp[4*1+i] - 17*temp[4*3+i];
- const int z3= 17* temp[4*1+i] + 7*temp[4*3+i];
-
- block[i*8+0]= ((z0 + z3)*3)>>11;
- block[i*8+1]= ((z1 + z2)*3)>>11;
- block[i*8+2]= ((z1 - z2)*3)>>11;
- block[i*8+3]= ((z0 - z3)*3)>>11;
- }
-
-}
-
-/** @} */ // transform
-
-
/**
* @name RV30/40 4x4 block decoding functions
* @{
}
for(i = 0; i < 4; i++){
- t = modulo_three_table[code][i];
+ t = (modulo_three_table[code] >> (6 - 2*i)) & 3;
if(t == 1)
cbp |= cbp_masks[get_bits1(gb)] << i;
if(t == 2)
/**
* Get one coefficient value from the bistream and store it.
*/
-static inline void decode_coeff(DCTELEM *dst, int coef, int esc, GetBitContext *gb, VLC* vlc)
+static inline void decode_coeff(DCTELEM *dst, int coef, int esc, GetBitContext *gb, VLC* vlc, int q)
{
if(coef){
if(coef == esc){
}
if(get_bits1(gb))
coef = -coef;
- *dst = coef;
+ *dst = (coef*q + 8) >> 4;
}
}
/**
* Decode 2x2 subblock of coefficients.
*/
-static inline void decode_subblock(DCTELEM *dst, int code, const int is_block2, GetBitContext *gb, VLC *vlc)
+static inline void decode_subblock(DCTELEM *dst, int code, const int is_block2, GetBitContext *gb, VLC *vlc, int q)
+{
+ int flags = modulo_three_table[code];
+
+ decode_coeff( dst+0, (flags >> 6) , 3, gb, vlc, q);
+ if(is_block2){
+ decode_coeff(dst+8, (flags >> 4) & 3, 2, gb, vlc, q);
+ decode_coeff(dst+1, (flags >> 2) & 3, 2, gb, vlc, q);
+ }else{
+ decode_coeff(dst+1, (flags >> 4) & 3, 2, gb, vlc, q);
+ decode_coeff(dst+8, (flags >> 2) & 3, 2, gb, vlc, q);
+ }
+ decode_coeff( dst+9, (flags >> 0) & 3, 2, gb, vlc, q);
+}
+
+/**
+ * Decode a single coefficient.
+ */
+static inline void decode_subblock1(DCTELEM *dst, int code, GetBitContext *gb, VLC *vlc, int q)
{
- int coeffs[4];
+ int coeff = modulo_three_table[code] >> 6;
+ decode_coeff(dst, coeff, 3, gb, vlc, q);
+}
+
+static inline void decode_subblock3(DCTELEM *dst, int code, const int is_block2, GetBitContext *gb, VLC *vlc,
+ int q_dc, int q_ac1, int q_ac2)
+{
+ int flags = modulo_three_table[code];
- coeffs[0] = modulo_three_table[code][0];
- coeffs[1] = modulo_three_table[code][1];
- coeffs[2] = modulo_three_table[code][2];
- coeffs[3] = modulo_three_table[code][3];
- decode_coeff(dst , coeffs[0], 3, gb, vlc);
+ decode_coeff( dst+0, (flags >> 6) , 3, gb, vlc, q_dc);
if(is_block2){
- decode_coeff(dst+8, coeffs[1], 2, gb, vlc);
- decode_coeff(dst+1, coeffs[2], 2, gb, vlc);
+ decode_coeff(dst+8, (flags >> 4) & 3, 2, gb, vlc, q_ac1);
+ decode_coeff(dst+1, (flags >> 2) & 3, 2, gb, vlc, q_ac1);
}else{
- decode_coeff(dst+1, coeffs[1], 2, gb, vlc);
- decode_coeff(dst+8, coeffs[2], 2, gb, vlc);
+ decode_coeff(dst+1, (flags >> 4) & 3, 2, gb, vlc, q_ac1);
+ decode_coeff(dst+8, (flags >> 2) & 3, 2, gb, vlc, q_ac1);
}
- decode_coeff(dst+9, coeffs[3], 2, gb, vlc);
+ decode_coeff( dst+9, (flags >> 0) & 3, 2, gb, vlc, q_ac2);
}
/**
* o--o
*/
-static inline void rv34_decode_block(DCTELEM *dst, GetBitContext *gb, RV34VLC *rvlc, int fc, int sc)
+static inline int rv34_decode_block(DCTELEM *dst, GetBitContext *gb, RV34VLC *rvlc, int fc, int sc, int q_dc, int q_ac1, int q_ac2)
{
- int code, pattern;
+ int code, pattern, has_ac = 1;
code = get_vlc2(gb, rvlc->first_pattern[fc].table, 9, 2);
pattern = code & 0x7;
code >>= 3;
- decode_subblock(dst, code, 0, gb, &rvlc->coefficient);
+
+ if (modulo_three_table[code] & 0x3F) {
+ decode_subblock3(dst, code, 0, gb, &rvlc->coefficient, q_dc, q_ac1, q_ac2);
+ } else {
+ decode_subblock1(dst, code, gb, &rvlc->coefficient, q_dc);
+ if (!pattern)
+ return 0;
+ has_ac = 0;
+ }
if(pattern & 4){
code = get_vlc2(gb, rvlc->second_pattern[sc].table, 9, 2);
- decode_subblock(dst + 2, code, 0, gb, &rvlc->coefficient);
+ decode_subblock(dst + 2, code, 0, gb, &rvlc->coefficient, q_ac2);
}
if(pattern & 2){ // Looks like coefficients 1 and 2 are swapped for this block
code = get_vlc2(gb, rvlc->second_pattern[sc].table, 9, 2);
- decode_subblock(dst + 8*2, code, 1, gb, &rvlc->coefficient);
+ decode_subblock(dst + 8*2, code, 1, gb, &rvlc->coefficient, q_ac2);
}
if(pattern & 1){
code = get_vlc2(gb, rvlc->third_pattern[sc].table, 9, 2);
- decode_subblock(dst + 8*2+2, code, 0, gb, &rvlc->coefficient);
+ decode_subblock(dst + 8*2+2, code, 0, gb, &rvlc->coefficient, q_ac2);
}
-
-}
-
-/**
- * Dequantize ordinary 4x4 block.
- * @todo optimize
- */
-static inline void rv34_dequant4x4(DCTELEM *block, int Qdc, int Q)
-{
- int i, j;
-
- block[0] = (block[0] * Qdc + 8) >> 4;
- for(i = 0; i < 4; i++)
- for(j = !i; j < 4; j++)
- block[j + i*8] = (block[j + i*8] * Q + 8) >> 4;
+ return has_ac || pattern;
}
-/**
- * Dequantize 4x4 block of DC values for 16x16 macroblock.
- * @todo optimize
- */
-static inline void rv34_dequant4x4_16x16(DCTELEM *block, int Qdc, int Q)
-{
- int i;
-
- for(i = 0; i < 3; i++)
- block[rv34_dezigzag[i]] = (block[rv34_dezigzag[i]] * Qdc + 8) >> 4;
- for(; i < 16; i++)
- block[rv34_dezigzag[i]] = (block[rv34_dezigzag[i]] * Q + 8) >> 4;
-}
-/** @} */ //block functions
-
-
/**
* @name RV30/40 bitstream parsing
* @{
*/
static int calc_add_mv(RV34DecContext *r, int dir, int val)
{
- int refdist = GET_PTS_DIFF(r->next_pts, r->last_pts);
- int dist = dir ? -GET_PTS_DIFF(r->next_pts, r->cur_pts) : GET_PTS_DIFF(r->cur_pts, r->last_pts);
- int mul;
+ int mul = dir ? -r->weight2 : r->weight1;
- if(!refdist) return 0;
- mul = (dist << 14) / refdist;
return (val * mul + 0x2000) >> 14;
}
static inline void rv34_mc(RV34DecContext *r, const int block_type,
const int xoff, const int yoff, int mv_off,
const int width, const int height, int dir,
- const int thirdpel,
+ const int thirdpel, int weighted,
qpel_mc_func (*qpel_mc)[16],
h264_chroma_mc_func (*chroma_mc))
{
my = (s->current_picture_ptr->f.motion_val[dir][mv_pos][1] + (3 << 24)) / 3 - (1 << 24);
lx = (s->current_picture_ptr->f.motion_val[dir][mv_pos][0] + (3 << 24)) % 3;
ly = (s->current_picture_ptr->f.motion_val[dir][mv_pos][1] + (3 << 24)) % 3;
- chroma_mx = (s->current_picture_ptr->f.motion_val[dir][mv_pos][0] + 1) >> 1;
- chroma_my = (s->current_picture_ptr->f.motion_val[dir][mv_pos][1] + 1) >> 1;
+ chroma_mx = s->current_picture_ptr->f.motion_val[dir][mv_pos][0] / 2;
+ chroma_my = s->current_picture_ptr->f.motion_val[dir][mv_pos][1] / 2;
umx = (chroma_mx + (3 << 24)) / 3 - (1 << 24);
umy = (chroma_my + (3 << 24)) / 3 - (1 << 24);
uvmx = chroma_coeffs[(chroma_mx + (3 << 24)) % 3];
if(uvmx == 6 && uvmy == 6)
uvmx = uvmy = 4;
}
+
+ if (HAVE_THREADS && (s->avctx->active_thread_type & FF_THREAD_FRAME)) {
+ /* wait for the referenced mb row to be finished */
+ int mb_row = FFMIN(s->mb_height - 1,
+ s->mb_y + ((yoff + my + 5 + 8 * height) >> 4));
+ AVFrame *f = dir ? &s->next_picture_ptr->f : &s->last_picture_ptr->f;
+ ff_thread_await_progress(f, mb_row, 0);
+ }
+
dxy = ly*4 + lx;
srcY = dir ? s->next_picture_ptr->f.data[0] : s->last_picture_ptr->f.data[0];
srcU = dir ? s->next_picture_ptr->f.data[1] : s->last_picture_ptr->f.data[1];
srcY += src_y * s->linesize + src_x;
srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
srcV += uvsrc_y * s->uvlinesize + uvsrc_x;
- if( (unsigned)(src_x - !!lx*2) > s->h_edge_pos - !!lx*2 - (width <<3) - 4
- || (unsigned)(src_y - !!ly*2) > s->v_edge_pos - !!ly*2 - (height<<3) - 4){
- uint8_t *uvbuf= s->edge_emu_buffer + 22 * s->linesize;
+ if(s->h_edge_pos - (width << 3) < 6 || s->v_edge_pos - (height << 3) < 6 ||
+ (unsigned)(src_x - !!lx*2) > s->h_edge_pos - !!lx*2 - (width <<3) - 4 ||
+ (unsigned)(src_y - !!ly*2) > s->v_edge_pos - !!ly*2 - (height<<3) - 4) {
+ uint8_t *uvbuf = s->edge_emu_buffer + 22 * s->linesize;
srcY -= 2 + 2*s->linesize;
s->dsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, (width<<3)+6, (height<<3)+6,
srcU = uvbuf;
srcV = uvbuf + 16;
}
- Y = s->dest[0] + xoff + yoff *s->linesize;
- U = s->dest[1] + (xoff>>1) + (yoff>>1)*s->uvlinesize;
- V = s->dest[2] + (xoff>>1) + (yoff>>1)*s->uvlinesize;
+ if(!weighted){
+ Y = s->dest[0] + xoff + yoff *s->linesize;
+ U = s->dest[1] + (xoff>>1) + (yoff>>1)*s->uvlinesize;
+ V = s->dest[2] + (xoff>>1) + (yoff>>1)*s->uvlinesize;
+ }else{
+ Y = r->tmp_b_block_y [dir] + xoff + yoff *s->linesize;
+ U = r->tmp_b_block_uv[dir*2] + (xoff>>1) + (yoff>>1)*s->uvlinesize;
+ V = r->tmp_b_block_uv[dir*2+1] + (xoff>>1) + (yoff>>1)*s->uvlinesize;
+ }
if(block_type == RV34_MB_P_16x8){
qpel_mc[1][dxy](Y, srcY, s->linesize);
const int xoff, const int yoff, int mv_off,
const int width, const int height, int dir)
{
- rv34_mc(r, block_type, xoff, yoff, mv_off, width, height, dir, r->rv30,
- r->rv30 ? r->s.dsp.put_rv30_tpel_pixels_tab
- : r->s.dsp.put_rv40_qpel_pixels_tab,
- r->rv30 ? r->s.dsp.put_h264_chroma_pixels_tab
- : r->s.dsp.put_rv40_chroma_pixels_tab);
+ rv34_mc(r, block_type, xoff, yoff, mv_off, width, height, dir, r->rv30, 0,
+ r->rdsp.put_pixels_tab,
+ r->rdsp.put_chroma_pixels_tab);
+}
+
+static void rv4_weight(RV34DecContext *r)
+{
+ r->rdsp.rv40_weight_pixels_tab[0](r->s.dest[0],
+ r->tmp_b_block_y[0],
+ r->tmp_b_block_y[1],
+ r->weight1,
+ r->weight2,
+ r->s.linesize);
+ r->rdsp.rv40_weight_pixels_tab[1](r->s.dest[1],
+ r->tmp_b_block_uv[0],
+ r->tmp_b_block_uv[2],
+ r->weight1,
+ r->weight2,
+ r->s.uvlinesize);
+ r->rdsp.rv40_weight_pixels_tab[1](r->s.dest[2],
+ r->tmp_b_block_uv[1],
+ r->tmp_b_block_uv[3],
+ r->weight1,
+ r->weight2,
+ r->s.uvlinesize);
}
static void rv34_mc_2mv(RV34DecContext *r, const int block_type)
{
- rv34_mc(r, block_type, 0, 0, 0, 2, 2, 0, r->rv30,
- r->rv30 ? r->s.dsp.put_rv30_tpel_pixels_tab
- : r->s.dsp.put_rv40_qpel_pixels_tab,
- r->rv30 ? r->s.dsp.put_h264_chroma_pixels_tab
- : r->s.dsp.put_rv40_chroma_pixels_tab);
- rv34_mc(r, block_type, 0, 0, 0, 2, 2, 1, r->rv30,
- r->rv30 ? r->s.dsp.avg_rv30_tpel_pixels_tab
- : r->s.dsp.avg_rv40_qpel_pixels_tab,
- r->rv30 ? r->s.dsp.avg_h264_chroma_pixels_tab
- : r->s.dsp.avg_rv40_chroma_pixels_tab);
+ int weighted = !r->rv30 && block_type != RV34_MB_B_BIDIR && r->weight1 != 8192;
+
+ rv34_mc(r, block_type, 0, 0, 0, 2, 2, 0, r->rv30, weighted,
+ r->rdsp.put_pixels_tab,
+ r->rdsp.put_chroma_pixels_tab);
+ if(!weighted){
+ rv34_mc(r, block_type, 0, 0, 0, 2, 2, 1, r->rv30, 0,
+ r->rdsp.avg_pixels_tab,
+ r->rdsp.avg_chroma_pixels_tab);
+ }else{
+ rv34_mc(r, block_type, 0, 0, 0, 2, 2, 1, r->rv30, 1,
+ r->rdsp.put_pixels_tab,
+ r->rdsp.put_chroma_pixels_tab);
+ rv4_weight(r);
+ }
}
static void rv34_mc_2mv_skip(RV34DecContext *r)
{
int i, j;
+ int weighted = !r->rv30 && r->weight1 != 8192;
+
for(j = 0; j < 2; j++)
for(i = 0; i < 2; i++){
rv34_mc(r, RV34_MB_P_8x8, i*8, j*8, i+j*r->s.b8_stride, 1, 1, 0, r->rv30,
- r->rv30 ? r->s.dsp.put_rv30_tpel_pixels_tab
- : r->s.dsp.put_rv40_qpel_pixels_tab,
- r->rv30 ? r->s.dsp.put_h264_chroma_pixels_tab
- : r->s.dsp.put_rv40_chroma_pixels_tab);
+ weighted,
+ r->rdsp.put_pixels_tab,
+ r->rdsp.put_chroma_pixels_tab);
rv34_mc(r, RV34_MB_P_8x8, i*8, j*8, i+j*r->s.b8_stride, 1, 1, 1, r->rv30,
- r->rv30 ? r->s.dsp.avg_rv30_tpel_pixels_tab
- : r->s.dsp.avg_rv40_qpel_pixels_tab,
- r->rv30 ? r->s.dsp.avg_h264_chroma_pixels_tab
- : r->s.dsp.avg_rv40_chroma_pixels_tab);
+ weighted,
+ weighted ? r->rdsp.put_pixels_tab : r->rdsp.avg_pixels_tab,
+ weighted ? r->rdsp.put_chroma_pixels_tab : r->rdsp.avg_chroma_pixels_tab);
}
+ if(weighted)
+ rv4_weight(r);
}
/** number of motion vectors in each macroblock type */
}
case RV34_MB_B_DIRECT:
//surprisingly, it uses motion scheme from next reference frame
+ /* wait for the current mb row to be finished */
+ if (HAVE_THREADS && (s->avctx->active_thread_type & FF_THREAD_FRAME))
+ ff_thread_await_progress(&s->next_picture_ptr->f, FFMAX(0, s->mb_y-1), 0);
+
next_bt = s->next_picture_ptr->f.mb_type[s->mb_x + s->mb_y * s->mb_stride];
if(IS_INTRA(next_bt) || IS_SKIP(next_bt)){
ZERO8x2(s->current_picture_ptr->f.motion_val[0][s->mb_x * 2 + s->mb_y * 2 * s->b8_stride], s->b8_stride);
if(itype == VERT_LEFT_PRED) itype = VERT_LEFT_PRED_RV40_NODOWN;
}
if(!right && up){
- topleft = dst[-stride + 3] * 0x01010101;
+ topleft = dst[-stride + 3] * 0x01010101u;
prev = (uint8_t*)&topleft;
}
r->h.pred4x4[itype](dst, prev, stride);
MpegEncContext *s = &r->s;
GetBitContext *gb = &s->gb;
int cbp, cbp2;
+ int q_dc, q_ac, has_ac;
int i, blknum, blkoff;
- DCTELEM block16[64];
+ LOCAL_ALIGNED_16(DCTELEM, block16, [64]);
int luma_dc_quant;
int dist;
int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
luma_dc_quant = r->block_type == RV34_MB_P_MIX16x16 ? r->luma_dc_quant_p[s->qscale] : r->luma_dc_quant_i[s->qscale];
if(r->is16){
- memset(block16, 0, sizeof(block16));
- rv34_decode_block(block16, gb, r->cur_vlcs, 3, 0);
- rv34_dequant4x4_16x16(block16, rv34_qscale_tab[luma_dc_quant],rv34_qscale_tab[s->qscale]);
- rv34_inv_transform_noround(block16);
+ q_dc = rv34_qscale_tab[luma_dc_quant];
+ q_ac = rv34_qscale_tab[s->qscale];
+ s->dsp.clear_block(block16);
+ if (rv34_decode_block(block16, gb, r->cur_vlcs, 3, 0, q_dc, q_dc, q_ac))
+ r->rdsp.rv34_inv_transform_tab[1](block16);
+ else
+ r->rdsp.rv34_inv_transform_dc_tab[1](block16);
}
+ q_ac = rv34_qscale_tab[s->qscale];
for(i = 0; i < 16; i++, cbp >>= 1){
+ DCTELEM *ptr;
if(!r->is16 && !(cbp & 1)) continue;
blknum = ((i & 2) >> 1) + ((i & 8) >> 2);
blkoff = ((i & 1) << 2) + ((i & 4) << 3);
+ ptr = s->block[blknum] + blkoff;
if(cbp & 1)
- rv34_decode_block(s->block[blknum] + blkoff, gb, r->cur_vlcs, r->luma_vlc, 0);
- rv34_dequant4x4(s->block[blknum] + blkoff, rv34_qscale_tab[s->qscale],rv34_qscale_tab[s->qscale]);
+ has_ac = rv34_decode_block(ptr, gb, r->cur_vlcs, r->luma_vlc, 0, q_ac, q_ac, q_ac);
+ else
+ has_ac = 0;
if(r->is16) //FIXME: optimize
- s->block[blknum][blkoff] = block16[(i & 3) | ((i & 0xC) << 1)];
- rv34_inv_transform(s->block[blknum] + blkoff);
+ ptr[0] = block16[(i & 3) | ((i & 0xC) << 1)];
+ if(has_ac)
+ r->rdsp.rv34_inv_transform_tab[0](ptr);
+ else
+ r->rdsp.rv34_inv_transform_dc_tab[0](ptr);
}
if(r->block_type == RV34_MB_P_MIX16x16)
r->cur_vlcs = choose_vlc_set(r->si.quant, r->si.vlc_set, 1);
+ q_dc = rv34_qscale_tab[rv34_chroma_quant[1][s->qscale]];
+ q_ac = rv34_qscale_tab[rv34_chroma_quant[0][s->qscale]];
for(; i < 24; i++, cbp >>= 1){
+ DCTELEM *ptr;
if(!(cbp & 1)) continue;
blknum = ((i & 4) >> 2) + 4;
blkoff = ((i & 1) << 2) + ((i & 2) << 4);
- rv34_decode_block(s->block[blknum] + blkoff, gb, r->cur_vlcs, r->chroma_vlc, 1);
- rv34_dequant4x4(s->block[blknum] + blkoff, rv34_qscale_tab[rv34_chroma_quant[1][s->qscale]],rv34_qscale_tab[rv34_chroma_quant[0][s->qscale]]);
- rv34_inv_transform(s->block[blknum] + blkoff);
+ ptr = s->block[blknum] + blkoff;
+ if (rv34_decode_block(ptr, gb, r->cur_vlcs, r->chroma_vlc, 1, q_dc, q_ac, q_ac))
+ r->rdsp.rv34_inv_transform_tab[0](ptr);
+ else
+ r->rdsp.rv34_inv_transform_dc_tab[0](ptr);
}
if (IS_INTRA(s->current_picture_ptr->f.mb_type[mb_pos]))
rv34_output_macroblock(r, intra_types, cbp2, r->is16);
return 1;
if(r->s.mb_skip_run > 1)
return 0;
- bits = r->bits - get_bits_count(&s->gb);
+ bits = get_bits_left(&s->gb);
if(bits < 0 || (bits < 8 && !show_bits(&s->gb, bits)))
return 1;
return 0;
r->cbp_luma = av_realloc(r->cbp_luma, r->s.mb_stride * r->s.mb_height * sizeof(*r->cbp_luma));
r->cbp_chroma = av_realloc(r->cbp_chroma, r->s.mb_stride * r->s.mb_height * sizeof(*r->cbp_chroma));
r->deblock_coefs = av_realloc(r->deblock_coefs, r->s.mb_stride * r->s.mb_height * sizeof(*r->deblock_coefs));
+ av_freep(&r->tmp_b_block_base);
}
s->pict_type = r->si.type ? r->si.type : AV_PICTURE_TYPE_I;
if(MPV_frame_start(s, s->avctx) < 0)
return -1;
ff_er_frame_start(s);
+ if (!r->tmp_b_block_base) {
+ int i;
+
+ r->tmp_b_block_base = av_malloc(s->linesize * 48);
+ for (i = 0; i < 2; i++)
+ r->tmp_b_block_y[i] = r->tmp_b_block_base + i * 16 * s->linesize;
+ for (i = 0; i < 4; i++)
+ r->tmp_b_block_uv[i] = r->tmp_b_block_base + 32 * s->linesize
+ + (i >> 1) * 8 * s->uvlinesize + (i & 1) * 16;
+ }
r->cur_pts = r->si.pts;
if(s->pict_type != AV_PICTURE_TYPE_B){
r->last_pts = r->next_pts;
r->next_pts = r->cur_pts;
+ }else{
+ int refdist = GET_PTS_DIFF(r->next_pts, r->last_pts);
+ int dist0 = GET_PTS_DIFF(r->cur_pts, r->last_pts);
+ int dist1 = GET_PTS_DIFF(r->next_pts, r->cur_pts);
+
+ if(!refdist){
+ r->weight1 = r->weight2 = 8192;
+ }else{
+ r->weight1 = (dist0 << 14) / refdist;
+ r->weight2 = (dist1 << 14) / refdist;
+ }
}
s->mb_x = s->mb_y = 0;
+ ff_thread_finish_setup(s->avctx);
+ } else {
+ int slice_type = r->si.type ? r->si.type : AV_PICTURE_TYPE_I;
+
+ if (slice_type != s->pict_type) {
+ av_log(s->avctx, AV_LOG_ERROR, "Slice type mismatch\n");
+ return AVERROR_INVALIDDATA;
+ }
}
r->si.end = end;
s->qscale = r->si.quant;
- r->bits = buf_size*8;
s->mb_num_left = r->si.end - r->si.start;
r->s.mb_skip_run = 0;
}
memset(r->intra_types_hist, -1, r->intra_types_stride * 4 * 2 * sizeof(*r->intra_types_hist));
s->first_slice_line = 1;
- s->resync_mb_x= s->mb_x;
- s->resync_mb_y= s->mb_y;
+ s->resync_mb_x = s->mb_x;
+ s->resync_mb_y = s->mb_y;
ff_init_block_index(s);
while(!check_slice_end(r, s)) {
s->dsp.clear_blocks(s->block[0]);
if(rv34_decode_macroblock(r, r->intra_types + s->mb_x * 4 + 4) < 0){
- ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, AC_ERROR|DC_ERROR|MV_ERROR);
+ ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, ER_MB_ERROR);
return -1;
}
if (++s->mb_x == s->mb_width) {
if(r->loop_filter && s->mb_y >= 2)
r->loop_filter(r, s->mb_y - 2);
+
+ if (HAVE_THREADS && (s->avctx->active_thread_type & FF_THREAD_FRAME))
+ ff_thread_report_progress(&s->current_picture_ptr->f,
+ s->mb_y - 2, 0);
+
}
if(s->mb_x == s->resync_mb_x)
s->first_slice_line=0;
s->mb_num_left--;
}
- ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, AC_END|DC_END|MV_END);
+ ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, ER_MB_END);
return s->mb_y == s->mb_height;
}
MpegEncContext *s = &r->s;
MPV_decode_defaults(s);
- s->avctx= avctx;
+ s->avctx = avctx;
s->out_format = FMT_H263;
- s->codec_id= avctx->codec_id;
+ s->codec_id = avctx->codec_id;
- s->width = avctx->width;
+ s->width = avctx->width;
s->height = avctx->height;
r->s.avctx = avctx;
if (MPV_common_init(s) < 0)
return -1;
- ff_h264_pred_init(&r->h, CODEC_ID_RV40, 8);
+ ff_h264_pred_init(&r->h, CODEC_ID_RV40, 8, 1);
+
+#if CONFIG_RV30_DECODER
+ if (avctx->codec_id == CODEC_ID_RV30)
+ ff_rv30dsp_init(&r->rdsp, &r->s.dsp);
+#endif
+#if CONFIG_RV40_DECODER
+ if (avctx->codec_id == CODEC_ID_RV40)
+ ff_rv40dsp_init(&r->rdsp, &r->s.dsp);
+#endif
r->intra_types_stride = 4*s->mb_stride + 4;
r->intra_types_hist = av_malloc(r->intra_types_stride * 4 * 2 * sizeof(*r->intra_types_hist));
return 0;
}
+int ff_rv34_decode_init_thread_copy(AVCodecContext *avctx)
+{
+ RV34DecContext *r = avctx->priv_data;
+
+ r->s.avctx = avctx;
+
+ if (avctx->internal->is_copy) {
+ r->cbp_chroma = av_malloc(r->s.mb_stride * r->s.mb_height *
+ sizeof(*r->cbp_chroma));
+ r->cbp_luma = av_malloc(r->s.mb_stride * r->s.mb_height *
+ sizeof(*r->cbp_luma));
+ r->deblock_coefs = av_malloc(r->s.mb_stride * r->s.mb_height *
+ sizeof(*r->deblock_coefs));
+ r->intra_types_hist = av_malloc(r->intra_types_stride * 4 * 2 *
+ sizeof(*r->intra_types_hist));
+ r->mb_type = av_malloc(r->s.mb_stride * r->s.mb_height *
+ sizeof(*r->mb_type));
+
+ if (!(r->cbp_chroma && r->cbp_luma && r->deblock_coefs &&
+ r->intra_types_hist && r->mb_type)) {
+ av_freep(&r->cbp_chroma);
+ av_freep(&r->cbp_luma);
+ av_freep(&r->deblock_coefs);
+ av_freep(&r->intra_types_hist);
+ av_freep(&r->mb_type);
+ r->intra_types = NULL;
+ return AVERROR(ENOMEM);
+ }
+
+ r->intra_types = r->intra_types_hist + r->intra_types_stride * 4;
+ r->tmp_b_block_base = NULL;
+
+ memset(r->mb_type, 0, r->s.mb_stride * r->s.mb_height *
+ sizeof(*r->mb_type));
+
+ MPV_common_init(&r->s);
+ }
+ return 0;
+}
+
+int ff_rv34_decode_update_thread_context(AVCodecContext *dst, const AVCodecContext *src)
+{
+ RV34DecContext *r = dst->priv_data, *r1 = src->priv_data;
+ MpegEncContext * const s = &r->s, * const s1 = &r1->s;
+ int err;
+
+ if (dst == src || !s1->context_initialized)
+ return 0;
+
+ if ((err = ff_mpeg_update_thread_context(dst, src)))
+ return err;
+
+ r->cur_pts = r1->cur_pts;
+ r->last_pts = r1->last_pts;
+ r->next_pts = r1->next_pts;
+
+ memset(&r->si, 0, sizeof(r->si));
+
+ /* necessary since it is it the condition checked for in decode_slice
+ * to call MPV_frame_start. cmp. comment at the end of decode_frame */
+ s->current_picture_ptr = NULL;
+
+ return 0;
+}
+
static int get_slice_offset(AVCodecContext *avctx, const uint8_t *buf, int n)
{
if(avctx->slice_count) return avctx->slice_offset[n];
if (buf_size == 0) {
/* special case for last picture */
if (s->low_delay==0 && s->next_picture_ptr) {
- *pict= *(AVFrame*)s->next_picture_ptr;
- s->next_picture_ptr= NULL;
+ *pict = *(AVFrame*)s->next_picture_ptr;
+ s->next_picture_ptr = NULL;
*data_size = sizeof(AVFrame);
}
slice_count = (*buf++) + 1;
slices_hdr = buf + 4;
buf += 8 * slice_count;
+ buf_size -= 1 + 8 * slice_count;
}else
slice_count = avctx->slice_count;
//parse first slice header to check whether this frame can be decoded
- if(get_slice_offset(avctx, slices_hdr, 0) > buf_size){
- av_log(avctx, AV_LOG_ERROR, "Slice offset is greater than frame size\n");
+ if(get_slice_offset(avctx, slices_hdr, 0) < 0 ||
+ get_slice_offset(avctx, slices_hdr, 0) > buf_size){
+ av_log(avctx, AV_LOG_ERROR, "Slice offset is invalid\n");
return -1;
}
- init_get_bits(&s->gb, buf+get_slice_offset(avctx, slices_hdr, 0), buf_size-get_slice_offset(avctx, slices_hdr, 0));
+ init_get_bits(&s->gb, buf+get_slice_offset(avctx, slices_hdr, 0), (buf_size-get_slice_offset(avctx, slices_hdr, 0))*8);
if(r->parse_slice_header(r, &r->s.gb, &si) < 0 || si.start){
av_log(avctx, AV_LOG_ERROR, "First slice header is incorrect\n");
return -1;
if( (avctx->skip_frame >= AVDISCARD_NONREF && si.type==AV_PICTURE_TYPE_B)
|| (avctx->skip_frame >= AVDISCARD_NONKEY && si.type!=AV_PICTURE_TYPE_I)
|| avctx->skip_frame >= AVDISCARD_ALL)
- return buf_size;
+ return avpkt->size;
- for(i=0; i<slice_count; i++){
- int offset= get_slice_offset(avctx, slices_hdr, i);
+ for(i = 0; i < slice_count; i++){
+ int offset = get_slice_offset(avctx, slices_hdr, i);
int size;
if(i+1 == slice_count)
- size= buf_size - offset;
+ size = buf_size - offset;
else
- size= get_slice_offset(avctx, slices_hdr, i+1) - offset;
+ size = get_slice_offset(avctx, slices_hdr, i+1) - offset;
- if(offset > buf_size){
- av_log(avctx, AV_LOG_ERROR, "Slice offset is greater than frame size\n");
+ if(offset < 0 || offset > buf_size){
+ av_log(avctx, AV_LOG_ERROR, "Slice offset is invalid\n");
break;
}
r->si.end = s->mb_width * s->mb_height;
if(i+1 < slice_count){
+ if (get_slice_offset(avctx, slices_hdr, i+1) < 0 ||
+ get_slice_offset(avctx, slices_hdr, i+1) > buf_size) {
+ av_log(avctx, AV_LOG_ERROR, "Slice offset is invalid\n");
+ break;
+ }
init_get_bits(&s->gb, buf+get_slice_offset(avctx, slices_hdr, i+1), (buf_size-get_slice_offset(avctx, slices_hdr, i+1))*8);
if(r->parse_slice_header(r, &r->s.gb, &si) < 0){
if(i+2 < slice_count)
}else
r->si.end = si.start;
}
+ if (size < 0 || size > buf_size - offset) {
+ av_log(avctx, AV_LOG_ERROR, "Slice size is invalid\n");
+ break;
+ }
last = rv34_decode_slice(r, r->si.end, buf + offset, size);
s->mb_num_left = r->s.mb_x + r->s.mb_y*r->s.mb_width - r->si.start;
if(last)
break;
}
- if(last){
+ if(last && s->current_picture_ptr){
if(r->loop_filter)
r->loop_filter(r, s->mb_height - 1);
+ if (HAVE_THREADS && (s->avctx->active_thread_type & FF_THREAD_FRAME))
+ ff_thread_report_progress(&s->current_picture_ptr->f,
+ s->mb_height - 1, 0);
ff_er_frame_end(s);
MPV_frame_end(s);
if (s->pict_type == AV_PICTURE_TYPE_B || s->low_delay) {
- *pict= *(AVFrame*)s->current_picture_ptr;
+ *pict = *(AVFrame*)s->current_picture_ptr;
} else if (s->last_picture_ptr != NULL) {
- *pict= *(AVFrame*)s->last_picture_ptr;
+ *pict = *(AVFrame*)s->last_picture_ptr;
}
if(s->last_picture_ptr || s->low_delay){
*data_size = sizeof(AVFrame);
ff_print_debug_info(s, pict);
}
- s->current_picture_ptr= NULL; //so we can detect if frame_end wasnt called (find some nicer solution...)
+ s->current_picture_ptr = NULL; //so we can detect if frame_end wasnt called (find some nicer solution...)
}
- return buf_size;
+ return avpkt->size;
}
av_cold int ff_rv34_decode_end(AVCodecContext *avctx)
av_freep(&r->intra_types_hist);
r->intra_types = NULL;
+ av_freep(&r->tmp_b_block_base);
av_freep(&r->mb_type);
av_freep(&r->cbp_luma);
av_freep(&r->cbp_chroma);
projects / ffmpeg / blobdiff