*/
/**
- * @file rv34.c
+ * @file libavcodec/rv34.c
* RV30/40 decoder common data
*/
#include "dsputil.h"
#include "mpegvideo.h"
#include "golomb.h"
+#include "mathops.h"
#include "rectangle.h"
#include "rv34vlc.h"
* Generate VLC from codeword lengths.
* @param bits codeword lengths (zeroes are accepted)
* @param size length of input data
+ * @param vlc output VLC
* @param insyms symbols for input codes (NULL for default ones)
*/
static void rv34_gen_vlc(const uint8_t *bits, int size, VLC *vlc, const uint8_t *insyms)
/**
* Initialize all tables.
*/
-static av_cold void rv34_init_tables()
+static av_cold void rv34_init_tables(void)
{
int i, j, k;
int ones;
static const int cbp_masks[3] = {0x100000, 0x010000, 0x110000};
static const int shifts[4] = { 0, 2, 8, 10 };
- int *curshift = shifts;
+ const int *curshift = shifts;
int i, t, mask;
code = get_vlc2(gb, vlc->cbppattern[table].table, 9, 2);
{
int i;
for(i = 0; i < 5; i++)
- if(rv34_mb_max_sizes[i] > mb_size)
+ if(rv34_mb_max_sizes[i] >= mb_size - 1)
break;
return rv34_mb_bits_sizes[i];
}
B[1] = cur_pic->motion_val[dir][mv_pos - s->b8_stride][1];
has_B = 1;
}
- if((r->avail_cache[5-2] & type) & mask){
+ if(r->avail_cache[5-4] && (r->avail_cache[5-2] & type) & mask){
C[0] = cur_pic->motion_val[dir][mv_pos - s->b8_stride + 2][0];
C[1] = cur_pic->motion_val[dir][mv_pos - s->b8_stride + 2][1];
has_C = 1;
MpegEncContext *s = &r->s;
int mv_pos = s->mb_x * 2 + s->mb_y * 2 * s->b8_stride;
int A[2] = {0}, B[2], C[2];
- int i, j;
+ int i, j, k;
int mx, my;
int avail_index = avail_indexes[0];
my += r->dmv[0][1];
for(j = 0; j < 2; j++){
for(i = 0; i < 2; i++){
- s->current_picture_ptr->motion_val[0][mv_pos + i + j*s->b8_stride][0] = mx;
- s->current_picture_ptr->motion_val[0][mv_pos + i + j*s->b8_stride][1] = my;
+ for(k = 0; k < 2; k++){
+ s->current_picture_ptr->motion_val[k][mv_pos + i + j*s->b8_stride][0] = mx;
+ s->current_picture_ptr->motion_val[k][mv_pos + i + j*s->b8_stride][1] = my;
+ }
}
}
- if(block_type == RV34_MB_B_BACKWARD || block_type == RV34_MB_B_FORWARD)
- fill_rectangle(s->current_picture_ptr->motion_val[!dir][mv_pos], 2, 2, s->b8_stride, 0, 4);
}
static const int chroma_coeffs[3] = { 0, 3, 5 };
* @param mv_off offset to the motion vector information
* @param width width of the current partition in 8x8 blocks
* @param height height of the current partition in 8x8 blocks
+ * @param dir motion compensation direction (i.e. from the last or the next reference frame)
+ * @param thirdpel motion vectors are specified in 1/3 of pixel
+ * @param qpel_mc a set of functions used to perform luma motion compensation
+ * @param chroma_mc a set of functions used to perform chroma motion compensation
*/
static inline void rv34_mc(RV34DecContext *r, const int block_type,
const int xoff, const int yoff, int mv_off,
uvmx = chroma_coeffs[(chroma_mx + (3 << 24)) % 3];
uvmy = chroma_coeffs[(chroma_my + (3 << 24)) % 3];
}else{
+ int cx, cy;
mx = s->current_picture_ptr->motion_val[dir][mv_pos][0] >> 2;
my = s->current_picture_ptr->motion_val[dir][mv_pos][1] >> 2;
lx = s->current_picture_ptr->motion_val[dir][mv_pos][0] & 3;
ly = s->current_picture_ptr->motion_val[dir][mv_pos][1] & 3;
- umx = mx >> 1;
- umy = my >> 1;
- uvmx = mx & 6;
- uvmy = my & 6;
+ cx = s->current_picture_ptr->motion_val[dir][mv_pos][0] / 2;
+ cy = s->current_picture_ptr->motion_val[dir][mv_pos][1] / 2;
+ umx = cx >> 2;
+ umy = cy >> 2;
+ uvmx = (cx & 3) << 1;
+ uvmy = (cy & 3) << 1;
+ //due to some flaw RV40 uses the same MC compensation routine for H2V2 and H3V3
+ if(uvmx == 6 && uvmy == 6)
+ uvmx = uvmy = 4;
}
dxy = ly*4 + lx;
srcY = dir ? s->next_picture_ptr->data[0] : s->last_picture_ptr->data[0];
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) - 3
- || (unsigned)(src_y - !!ly*2) > s->v_edge_pos - !!ly*2 - (height<<3) - 3){
- uint8_t *uvbuf= s->edge_emu_buffer + 20 * s->linesize;
+ 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;
srcY -= 2 + 2*s->linesize;
- ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, (width<<3)+4, (height<<3)+4,
+ ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, (width<<3)+6, (height<<3)+6,
src_x - 2, src_y - 2, s->h_edge_pos, s->v_edge_pos);
srcY = s->edge_emu_buffer + 2 + 2*s->linesize;
ff_emulated_edge_mc(uvbuf , srcU, s->uvlinesize, (width<<2)+1, (height<<2)+1,
{
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_h264_qpel_pixels_tab,
- r->s.dsp.put_h264_chroma_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);
}
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_h264_qpel_pixels_tab,
- r->s.dsp.put_h264_chroma_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_h264_qpel_pixels_tab,
- r->s.dsp.avg_h264_chroma_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);
}
static void rv34_mc_2mv_skip(RV34DecContext *r)
{
- int i, j, k;
+ int i, j;
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_h264_qpel_pixels_tab,
- r->s.dsp.put_h264_chroma_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, 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_h264_qpel_pixels_tab,
- r->s.dsp.avg_h264_chroma_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);
}
}
case RV34_MB_B_DIRECT:
//surprisingly, it uses motion scheme from next reference frame
next_bt = s->next_picture_ptr->mb_type[s->mb_x + s->mb_y * s->mb_stride];
- for(j = 0; j < 2; j++)
- for(i = 0; i < 2; i++)
- for(k = 0; k < 2; k++)
- for(l = 0; l < 2; l++)
- s->current_picture_ptr->motion_val[l][mv_pos + i + j*s->b8_stride][k] = calc_add_mv(r, l, s->next_picture_ptr->motion_val[0][mv_pos + i + j*s->b8_stride][k]);
- if(IS_16X16(next_bt)) //we can use whole macroblock MC
+ if(IS_INTRA(next_bt) || IS_SKIP(next_bt)){
+ fill_rectangle(s->current_picture_ptr->motion_val[0][s->mb_x * 2 + s->mb_y * 2 * s->b8_stride], 2, 2, s->b8_stride, 0, 4);
+ fill_rectangle(s->current_picture_ptr->motion_val[1][s->mb_x * 2 + s->mb_y * 2 * s->b8_stride], 2, 2, s->b8_stride, 0, 4);
+ }else
+ for(j = 0; j < 2; j++)
+ for(i = 0; i < 2; i++)
+ for(k = 0; k < 2; k++)
+ for(l = 0; l < 2; l++)
+ s->current_picture_ptr->motion_val[l][mv_pos + i + j*s->b8_stride][k] = calc_add_mv(r, l, s->next_picture_ptr->motion_val[0][mv_pos + i + j*s->b8_stride][k]);
+ if(!(IS_16X8(next_bt) || IS_8X16(next_bt) || IS_8X8(next_bt))) //we can use whole macroblock MC
rv34_mc_2mv(r, block_type);
else
rv34_mc_2mv_skip(r);
}
if(!right && up){
topleft = dst[-stride + 3] * 0x01010101;
- prev = &topleft;
+ prev = (uint8_t*)&topleft;
}
r->h.pred4x4[itype](dst, prev, stride);
}
itype = ittrans16[intra_types[0]];
itype = adjust_pred16(itype, r->avail_cache[5-4], r->avail_cache[5-1]);
r->h.pred16x16[itype](Y, s->linesize);
- dsp->add_pixels_clamped(s->block[0], Y, s->current_picture.linesize[0]);
- dsp->add_pixels_clamped(s->block[1], Y + 8, s->current_picture.linesize[0]);
- Y += s->current_picture.linesize[0] * 8;
- dsp->add_pixels_clamped(s->block[2], Y, s->current_picture.linesize[0]);
- dsp->add_pixels_clamped(s->block[3], Y + 8, s->current_picture.linesize[0]);
+ dsp->add_pixels_clamped(s->block[0], Y, s->linesize);
+ dsp->add_pixels_clamped(s->block[1], Y + 8, s->linesize);
+ Y += s->linesize * 8;
+ dsp->add_pixels_clamped(s->block[2], Y, s->linesize);
+ dsp->add_pixels_clamped(s->block[3], Y + 8, s->linesize);
itype = ittrans16[intra_types[0]];
if(itype == PLANE_PRED8x8) itype = DC_PRED8x8;
* mask for retrieving all bits in coded block pattern
* corresponding to one 8x8 block
*/
-#define LUMA_CBP_BLOCK_MASK 0x303
+#define LUMA_CBP_BLOCK_MASK 0x33
#define U_CBP_MASK 0x0F0000
#define V_CBP_MASK 0xF00000
int i;
for(i = 0; i < 4; i++)
- if(cbp & (LUMA_CBP_BLOCK_MASK << shifts[i]))
+ if((cbp & (LUMA_CBP_BLOCK_MASK << shifts[i])) || r->block_type == RV34_MB_P_MIX16x16)
s->dsp.add_pixels_clamped(s->block[i], s->dest[0] + (i & 1)*8 + (i&2)*4*s->linesize, s->linesize);
if(cbp & U_CBP_MASK)
s->dsp.add_pixels_clamped(s->block[4], s->dest[1], s->uvlinesize);
static int rv34_set_deblock_coef(RV34DecContext *r)
{
MpegEncContext *s = &r->s;
- int mvmask = 0, i, j;
+ int hmvmask = 0, vmvmask = 0, i, j;
int midx = s->mb_x * 2 + s->mb_y * 2 * s->b8_stride;
int16_t (*motion_val)[2] = s->current_picture_ptr->motion_val[0][midx];
for(j = 0; j < 16; j += 8){
for(i = 0; i < 2; i++){
if(is_mv_diff_gt_3(motion_val + i, 1))
- mvmask |= 0x11 << (j + i*2);
- if(is_mv_diff_gt_3(motion_val + i, s->b8_stride))
- mvmask |= 0x03 << (j + i*2);
+ vmvmask |= 0x11 << (j + i*2);
+ if((j || s->mb_y) && is_mv_diff_gt_3(motion_val + i, s->b8_stride))
+ hmvmask |= 0x03 << (j + i*2);
}
motion_val += s->b8_stride;
}
- return mvmask;
+ if(s->first_slice_line)
+ hmvmask &= ~0x000F;
+ if(!s->mb_x)
+ vmvmask &= ~0x1111;
+ if(r->rv30){ //RV30 marks both subblocks on the edge for filtering
+ vmvmask |= (vmvmask & 0x4444) >> 1;
+ hmvmask |= (hmvmask & 0x0F00) >> 4;
+ if(s->mb_x)
+ r->deblock_coefs[s->mb_x - 1 + s->mb_y*s->mb_stride] |= (vmvmask & 0x1111) << 3;
+ if(!s->first_slice_line)
+ r->deblock_coefs[s->mb_x + (s->mb_y - 1)*s->mb_stride] |= (hmvmask & 0xF) << 12;
+ }
+ return hmvmask | vmvmask;
}
static int rv34_decode_macroblock(RV34DecContext *r, int8_t *intra_types)
s->qscale = r->si.quant;
cbp = cbp2 = rv34_decode_mb_header(r, intra_types);
- r->cbp_luma [s->mb_x + s->mb_y * s->mb_stride] = cbp;
- r->cbp_chroma[s->mb_x + s->mb_y * s->mb_stride] = cbp >> 16;
+ r->cbp_luma [mb_pos] = cbp;
+ r->cbp_chroma[mb_pos] = cbp >> 16;
if(s->pict_type == FF_I_TYPE)
- r->deblock_coefs[mb_pos] = 0;
+ r->deblock_coefs[mb_pos] = 0xFFFF;
else
- r->deblock_coefs[mb_pos] = rv34_set_deblock_coef(r);
- s->current_picture.qscale_table[s->mb_x + s->mb_y * s->mb_stride] = s->qscale;
+ r->deblock_coefs[mb_pos] = rv34_set_deblock_coef(r) | r->cbp_luma[mb_pos];
+ s->current_picture_ptr->qscale_table[mb_pos] = s->qscale;
if(cbp == -1)
return -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);
}
- if(IS_INTRA(s->current_picture_ptr->mb_type[s->mb_x + s->mb_y*s->mb_stride]))
+ if(IS_INTRA(s->current_picture_ptr->mb_type[mb_pos]))
rv34_output_macroblock(r, intra_types, cbp2, r->is16);
else
rv34_apply_differences(r, cbp2);
si1->pts != si2->pts;
}
-static int rv34_decode_slice(RV34DecContext *r, int end, uint8_t* buf, int buf_size)
+static int rv34_decode_slice(RV34DecContext *r, int end, const uint8_t* buf, int buf_size)
{
MpegEncContext *s = &r->s;
GetBitContext *gb = &s->gb;
if(MPV_frame_start(s, s->avctx) < 0)
return -1;
ff_er_frame_start(s);
- s->current_picture_ptr = &s->current_picture;
r->cur_pts = r->si.pts;
if(s->pict_type != FF_B_TYPE){
r->last_pts = r->next_pts;
memmove(r->intra_types_hist, r->intra_types, s->b4_stride * 4 * sizeof(*r->intra_types_hist));
memset(r->intra_types, -1, s->b4_stride * 4 * sizeof(*r->intra_types_hist));
+
+ if(r->loop_filter && s->mb_y >= 2)
+ r->loop_filter(r, s->mb_y - 2);
}
if(s->mb_x == s->resync_mb_x)
s->first_slice_line=0;
return 0;
}
-static int get_slice_offset(AVCodecContext *avctx, uint8_t *buf, int n)
+static int get_slice_offset(AVCodecContext *avctx, const uint8_t *buf, int n)
{
if(avctx->slice_count) return avctx->slice_offset[n];
else return AV_RL32(buf + n*8 - 4) == 1 ? AV_RL32(buf + n*8) : AV_RB32(buf + n*8);
int ff_rv34_decode_frame(AVCodecContext *avctx,
void *data, int *data_size,
- uint8_t *buf, int buf_size)
+ AVPacket *avpkt)
{
+ const uint8_t *buf = avpkt->data;
+ int buf_size = avpkt->size;
RV34DecContext *r = avctx->priv_data;
MpegEncContext *s = &r->s;
AVFrame *pict = data;
SliceInfo si;
int i;
int slice_count;
- uint8_t *slices_hdr = NULL;
+ const uint8_t *slices_hdr = NULL;
int last = 0;
/* no supplementary picture */
}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");
+ return -1;
+ }
+ init_get_bits(&s->gb, buf+get_slice_offset(avctx, slices_hdr, 0), buf_size-get_slice_offset(avctx, slices_hdr, 0));
+ 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((!s->last_picture_ptr || !s->last_picture_ptr->data[0]) && si.type == FF_B_TYPE)
+ return -1;
+ /* skip b frames if we are in a hurry */
+ if(avctx->hurry_up && si.type==FF_B_TYPE) return buf_size;
+ if( (avctx->skip_frame >= AVDISCARD_NONREF && si.type==FF_B_TYPE)
+ || (avctx->skip_frame >= AVDISCARD_NONKEY && si.type!=FF_I_TYPE)
+ || avctx->skip_frame >= AVDISCARD_ALL)
+ return buf_size;
+ /* skip everything if we are in a hurry>=5 */
+ if(avctx->hurry_up>=5)
+ return buf_size;
+
for(i=0; i<slice_count; i++){
int offset= get_slice_offset(avctx, slices_hdr, i);
int size;
else
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");
+ break;
+ }
+
r->si.end = s->mb_width * s->mb_height;
if(i+1 < slice_count){
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(last){
if(r->loop_filter)
- r->loop_filter(r);
+ r->loop_filter(r, s->mb_height - 1);
ff_er_frame_end(s);
MPV_frame_end(s);
if (s->pict_type == FF_B_TYPE || s->low_delay) {
projects / ffmpeg / blobdiff