/**
* @file
- * H.264 / AVC / MPEG4 part10 macroblock decoding
+ * H.264 / AVC / MPEG-4 part10 macroblock decoding
*/
#include <stdint.h>
#include "libavutil/common.h"
#include "libavutil/intreadwrite.h"
#include "avcodec.h"
-#include "h264.h"
+#include "h264dec.h"
+#include "h264_ps.h"
#include "qpeldsp.h"
-#include "svq3.h"
#include "thread.h"
-static inline int get_lowest_part_list_y(H264Context *h, H264SliceContext *sl,
- H264Picture *pic, int n,
- int height, int y_offset, int list)
+static inline int get_lowest_part_list_y(H264SliceContext *sl,
+ int n, int height, int y_offset, int list)
{
int raw_my = sl->mv_cache[list][scan8[n]][1];
int filter_height_up = (raw_my & 3) ? 2 : 0;
return FFMAX(abs(top), bottom);
}
-static inline void get_lowest_part_y(H264Context *h, H264SliceContext *sl,
+static inline void get_lowest_part_y(const H264Context *h, H264SliceContext *sl,
int refs[2][48], int n,
int height, int y_offset, int list0,
int list1, int *nrefs)
{
int my;
- y_offset += 16 * (h->mb_y >> MB_FIELD(h));
+ y_offset += 16 * (sl->mb_y >> MB_FIELD(sl));
if (list0) {
int ref_n = sl->ref_cache[0][scan8[n]];
- H264Picture *ref = &sl->ref_list[0][ref_n];
+ H264Ref *ref = &sl->ref_list[0][ref_n];
// Error resilience puts the current picture in the ref list.
// Don't try to wait on these as it will cause a deadlock.
// Fields can wait on each other, though.
- if (ref->tf.progress->data != h->cur_pic.tf.progress->data ||
+ if (ref->parent->tf.progress->data != h->cur_pic.tf.progress->data ||
(ref->reference & 3) != h->picture_structure) {
- my = get_lowest_part_list_y(h, sl, ref, n, height, y_offset, 0);
+ my = get_lowest_part_list_y(sl, n, height, y_offset, 0);
if (refs[0][ref_n] < 0)
nrefs[0] += 1;
refs[0][ref_n] = FFMAX(refs[0][ref_n], my);
if (list1) {
int ref_n = sl->ref_cache[1][scan8[n]];
- H264Picture *ref = &sl->ref_list[1][ref_n];
+ H264Ref *ref = &sl->ref_list[1][ref_n];
- if (ref->tf.progress->data != h->cur_pic.tf.progress->data ||
+ if (ref->parent->tf.progress->data != h->cur_pic.tf.progress->data ||
(ref->reference & 3) != h->picture_structure) {
- my = get_lowest_part_list_y(h, sl, ref, n, height, y_offset, 1);
+ my = get_lowest_part_list_y(sl, n, height, y_offset, 1);
if (refs[1][ref_n] < 0)
nrefs[1] += 1;
refs[1][ref_n] = FFMAX(refs[1][ref_n], my);
/**
* Wait until all reference frames are available for MC operations.
*
- * @param h the H264 context
+ * @param h the H.264 context
*/
-static void await_references(H264Context *h, H264SliceContext *sl)
+static void await_references(const H264Context *h, H264SliceContext *sl)
{
- const int mb_xy = h->mb_xy;
+ const int mb_xy = sl->mb_xy;
const int mb_type = h->cur_pic.mb_type[mb_xy];
int refs[2][48];
int nrefs[2] = { 0 };
for (ref = 0; ref < 48 && nrefs[list]; ref++) {
int row = refs[list][ref];
if (row >= 0) {
- H264Picture *ref_pic = &sl->ref_list[list][ref];
+ H264Ref *ref_pic = &sl->ref_list[list][ref];
int ref_field = ref_pic->reference - 1;
- int ref_field_picture = ref_pic->field_picture;
+ int ref_field_picture = ref_pic->parent->field_picture;
int pic_height = 16 * h->mb_height >> ref_field_picture;
- row <<= MB_MBAFF(h);
+ row <<= MB_MBAFF(sl);
nrefs[list]--;
if (!FIELD_PICTURE(h) && ref_field_picture) { // frame referencing two fields
- ff_thread_await_progress(&ref_pic->tf,
+ ff_thread_await_progress(&ref_pic->parent->tf,
FFMIN((row >> 1) - !(row & 1),
pic_height - 1),
1);
- ff_thread_await_progress(&ref_pic->tf,
+ ff_thread_await_progress(&ref_pic->parent->tf,
FFMIN((row >> 1), pic_height - 1),
0);
} else if (FIELD_PICTURE(h) && !ref_field_picture) { // field referencing one field of a frame
- ff_thread_await_progress(&ref_pic->tf,
+ ff_thread_await_progress(&ref_pic->parent->tf,
FFMIN(row * 2 + ref_field,
pic_height - 1),
0);
} else if (FIELD_PICTURE(h)) {
- ff_thread_await_progress(&ref_pic->tf,
+ ff_thread_await_progress(&ref_pic->parent->tf,
FFMIN(row, pic_height - 1),
ref_field);
} else {
- ff_thread_await_progress(&ref_pic->tf,
+ ff_thread_await_progress(&ref_pic->parent->tf,
FFMIN(row, pic_height - 1),
0);
}
}
}
-static av_always_inline void mc_dir_part(H264Context *h, H264SliceContext *sl,
- H264Picture *pic,
+static av_always_inline void mc_dir_part(const H264Context *h, H264SliceContext *sl,
+ H264Ref *pic,
int n, int square, int height,
int delta, int list,
uint8_t *dest_y, uint8_t *dest_cb,
uint8_t *dest_cr,
int src_x_offset, int src_y_offset,
- qpel_mc_func *qpix_op,
+ const qpel_mc_func *qpix_op,
h264_chroma_mc_func chroma_op,
int pixel_shift, int chroma_idc)
{
int my = sl->mv_cache[list][scan8[n]][1] + src_y_offset * 8;
const int luma_xy = (mx & 3) + ((my & 3) << 2);
ptrdiff_t offset = ((mx >> 2) << pixel_shift) + (my >> 2) * sl->mb_linesize;
- uint8_t *src_y = pic->f.data[0] + offset;
+ uint8_t *src_y = pic->data[0] + offset;
uint8_t *src_cb, *src_cr;
int extra_width = 0;
int extra_height = 0;
const int full_mx = mx >> 2;
const int full_my = my >> 2;
const int pic_width = 16 * h->mb_width;
- const int pic_height = 16 * h->mb_height >> MB_FIELD(h);
+ const int pic_height = 16 * h->mb_height >> MB_FIELD(sl);
int ysh;
if (mx & 7)
full_my < 0 - extra_height ||
full_mx + 16 /*FIXME*/ > pic_width + extra_width ||
full_my + 16 /*FIXME*/ > pic_height + extra_height) {
- h->vdsp.emulated_edge_mc(h->edge_emu_buffer,
+ h->vdsp.emulated_edge_mc(sl->edge_emu_buffer,
src_y - (2 << pixel_shift) - 2 * sl->mb_linesize,
sl->mb_linesize, sl->mb_linesize,
16 + 5, 16 + 5 /*FIXME*/, full_mx - 2,
full_my - 2, pic_width, pic_height);
- src_y = h->edge_emu_buffer + (2 << pixel_shift) + 2 * sl->mb_linesize;
+ src_y = sl->edge_emu_buffer + (2 << pixel_shift) + 2 * sl->mb_linesize;
emu = 1;
}
if (!square)
qpix_op[luma_xy](dest_y + delta, src_y + delta, sl->mb_linesize);
- if (CONFIG_GRAY && h->flags & CODEC_FLAG_GRAY)
+ if (CONFIG_GRAY && h->flags & AV_CODEC_FLAG_GRAY)
return;
if (chroma_idc == 3 /* yuv444 */) {
- src_cb = pic->f.data[1] + offset;
+ src_cb = pic->data[1] + offset;
if (emu) {
- h->vdsp.emulated_edge_mc(h->edge_emu_buffer,
+ h->vdsp.emulated_edge_mc(sl->edge_emu_buffer,
src_cb - (2 << pixel_shift) - 2 * sl->mb_linesize,
sl->mb_linesize, sl->mb_linesize,
16 + 5, 16 + 5 /*FIXME*/,
full_mx - 2, full_my - 2,
pic_width, pic_height);
- src_cb = h->edge_emu_buffer + (2 << pixel_shift) + 2 * sl->mb_linesize;
+ src_cb = sl->edge_emu_buffer + (2 << pixel_shift) + 2 * sl->mb_linesize;
}
qpix_op[luma_xy](dest_cb, src_cb, sl->mb_linesize); // FIXME try variable height perhaps?
if (!square)
qpix_op[luma_xy](dest_cb + delta, src_cb + delta, sl->mb_linesize);
- src_cr = pic->f.data[2] + offset;
+ src_cr = pic->data[2] + offset;
if (emu) {
- h->vdsp.emulated_edge_mc(h->edge_emu_buffer,
+ h->vdsp.emulated_edge_mc(sl->edge_emu_buffer,
src_cr - (2 << pixel_shift) - 2 * sl->mb_linesize,
sl->mb_linesize, sl->mb_linesize,
16 + 5, 16 + 5 /*FIXME*/,
full_mx - 2, full_my - 2,
pic_width, pic_height);
- src_cr = h->edge_emu_buffer + (2 << pixel_shift) + 2 * sl->mb_linesize;
+ src_cr = sl->edge_emu_buffer + (2 << pixel_shift) + 2 * sl->mb_linesize;
}
qpix_op[luma_xy](dest_cr, src_cr, sl->mb_linesize); // FIXME try variable height perhaps?
if (!square)
}
ysh = 3 - (chroma_idc == 2 /* yuv422 */);
- if (chroma_idc == 1 /* yuv420 */ && MB_FIELD(h)) {
+ if (chroma_idc == 1 /* yuv420 */ && MB_FIELD(sl)) {
// chroma offset when predicting from a field of opposite parity
- my += 2 * ((h->mb_y & 1) - (pic->reference - 1));
+ my += 2 * ((sl->mb_y & 1) - (pic->reference - 1));
emu |= (my >> 3) < 0 || (my >> 3) + 8 >= (pic_height >> 1);
}
- src_cb = pic->f.data[1] + ((mx >> 3) << pixel_shift) +
+ src_cb = pic->data[1] + ((mx >> 3) << pixel_shift) +
(my >> ysh) * sl->mb_uvlinesize;
- src_cr = pic->f.data[2] + ((mx >> 3) << pixel_shift) +
+ src_cr = pic->data[2] + ((mx >> 3) << pixel_shift) +
(my >> ysh) * sl->mb_uvlinesize;
if (emu) {
- h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src_cb,
+ h->vdsp.emulated_edge_mc(sl->edge_emu_buffer, src_cb,
sl->mb_uvlinesize, sl->mb_uvlinesize,
9, 8 * chroma_idc + 1, (mx >> 3), (my >> ysh),
pic_width >> 1, pic_height >> (chroma_idc == 1 /* yuv420 */));
- src_cb = h->edge_emu_buffer;
+ src_cb = sl->edge_emu_buffer;
}
chroma_op(dest_cb, src_cb, sl->mb_uvlinesize,
height >> (chroma_idc == 1 /* yuv420 */),
mx & 7, (my << (chroma_idc == 2 /* yuv422 */)) & 7);
if (emu) {
- h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src_cr,
+ h->vdsp.emulated_edge_mc(sl->edge_emu_buffer, src_cr,
sl->mb_uvlinesize, sl->mb_uvlinesize,
9, 8 * chroma_idc + 1, (mx >> 3), (my >> ysh),
pic_width >> 1, pic_height >> (chroma_idc == 1 /* yuv420 */));
- src_cr = h->edge_emu_buffer;
+ src_cr = sl->edge_emu_buffer;
}
chroma_op(dest_cr, src_cr, sl->mb_uvlinesize, height >> (chroma_idc == 1 /* yuv420 */),
mx & 7, (my << (chroma_idc == 2 /* yuv422 */)) & 7);
}
-static av_always_inline void mc_part_std(H264Context *h, H264SliceContext *sl,
+static av_always_inline void mc_part_std(const H264Context *h, H264SliceContext *sl,
int n, int square,
int height, int delta,
uint8_t *dest_y, uint8_t *dest_cb,
uint8_t *dest_cr,
int x_offset, int y_offset,
- qpel_mc_func *qpix_put,
+ const qpel_mc_func *qpix_put,
h264_chroma_mc_func chroma_put,
- qpel_mc_func *qpix_avg,
+ const qpel_mc_func *qpix_avg,
h264_chroma_mc_func chroma_avg,
int list0, int list1,
int pixel_shift, int chroma_idc)
{
- qpel_mc_func *qpix_op = qpix_put;
+ const qpel_mc_func *qpix_op = qpix_put;
h264_chroma_mc_func chroma_op = chroma_put;
dest_y += (2 * x_offset << pixel_shift) + 2 * y_offset * sl->mb_linesize;
dest_cb += (x_offset << pixel_shift) + y_offset * sl->mb_uvlinesize;
dest_cr += (x_offset << pixel_shift) + y_offset * sl->mb_uvlinesize;
}
- x_offset += 8 * h->mb_x;
- y_offset += 8 * (h->mb_y >> MB_FIELD(h));
+ x_offset += 8 * sl->mb_x;
+ y_offset += 8 * (sl->mb_y >> MB_FIELD(sl));
if (list0) {
- H264Picture *ref = &sl->ref_list[0][sl->ref_cache[0][scan8[n]]];
+ H264Ref *ref = &sl->ref_list[0][sl->ref_cache[0][scan8[n]]];
mc_dir_part(h, sl, ref, n, square, height, delta, 0,
dest_y, dest_cb, dest_cr, x_offset, y_offset,
qpix_op, chroma_op, pixel_shift, chroma_idc);
}
if (list1) {
- H264Picture *ref = &sl->ref_list[1][sl->ref_cache[1][scan8[n]]];
+ H264Ref *ref = &sl->ref_list[1][sl->ref_cache[1][scan8[n]]];
mc_dir_part(h, sl, ref, n, square, height, delta, 1,
dest_y, dest_cb, dest_cr, x_offset, y_offset,
qpix_op, chroma_op, pixel_shift, chroma_idc);
}
}
-static av_always_inline void mc_part_weighted(H264Context *h, H264SliceContext *sl,
+static av_always_inline void mc_part_weighted(const H264Context *h, H264SliceContext *sl,
int n, int square,
int height, int delta,
uint8_t *dest_y, uint8_t *dest_cb,
uint8_t *dest_cr,
int x_offset, int y_offset,
- qpel_mc_func *qpix_put,
+ const qpel_mc_func *qpix_put,
h264_chroma_mc_func chroma_put,
h264_weight_func luma_weight_op,
h264_weight_func chroma_weight_op,
dest_cb += (x_offset << pixel_shift) + y_offset * sl->mb_uvlinesize;
dest_cr += (x_offset << pixel_shift) + y_offset * sl->mb_uvlinesize;
}
- x_offset += 8 * h->mb_x;
- y_offset += 8 * (h->mb_y >> MB_FIELD(h));
+ x_offset += 8 * sl->mb_x;
+ y_offset += 8 * (sl->mb_y >> MB_FIELD(sl));
if (list0 && list1) {
/* don't optimize for luma-only case, since B-frames usually
* use implicit weights => chroma too. */
- uint8_t *tmp_cb = h->bipred_scratchpad;
- uint8_t *tmp_cr = h->bipred_scratchpad + (16 << pixel_shift);
- uint8_t *tmp_y = h->bipred_scratchpad + 16 * sl->mb_uvlinesize;
+ uint8_t *tmp_cb = sl->bipred_scratchpad;
+ uint8_t *tmp_cr = sl->bipred_scratchpad + (16 << pixel_shift);
+ uint8_t *tmp_y = sl->bipred_scratchpad + 16 * sl->mb_uvlinesize;
int refn0 = sl->ref_cache[0][scan8[n]];
int refn1 = sl->ref_cache[1][scan8[n]];
x_offset, y_offset, qpix_put, chroma_put,
pixel_shift, chroma_idc);
- if (sl->use_weight == 2) {
- int weight0 = sl->implicit_weight[refn0][refn1][h->mb_y & 1];
+ if (sl->pwt.use_weight == 2) {
+ int weight0 = sl->pwt.implicit_weight[refn0][refn1][sl->mb_y & 1];
int weight1 = 64 - weight0;
luma_weight_avg(dest_y, tmp_y, sl->mb_linesize,
height, 5, weight0, weight1, 0);
chroma_height, 5, weight0, weight1, 0);
} else {
luma_weight_avg(dest_y, tmp_y, sl->mb_linesize, height,
- sl->luma_log2_weight_denom,
- sl->luma_weight[refn0][0][0],
- sl->luma_weight[refn1][1][0],
- sl->luma_weight[refn0][0][1] +
- sl->luma_weight[refn1][1][1]);
+ sl->pwt.luma_log2_weight_denom,
+ sl->pwt.luma_weight[refn0][0][0],
+ sl->pwt.luma_weight[refn1][1][0],
+ sl->pwt.luma_weight[refn0][0][1] +
+ sl->pwt.luma_weight[refn1][1][1]);
chroma_weight_avg(dest_cb, tmp_cb, sl->mb_uvlinesize, chroma_height,
- sl->chroma_log2_weight_denom,
- sl->chroma_weight[refn0][0][0][0],
- sl->chroma_weight[refn1][1][0][0],
- sl->chroma_weight[refn0][0][0][1] +
- sl->chroma_weight[refn1][1][0][1]);
+ sl->pwt.chroma_log2_weight_denom,
+ sl->pwt.chroma_weight[refn0][0][0][0],
+ sl->pwt.chroma_weight[refn1][1][0][0],
+ sl->pwt.chroma_weight[refn0][0][0][1] +
+ sl->pwt.chroma_weight[refn1][1][0][1]);
chroma_weight_avg(dest_cr, tmp_cr, sl->mb_uvlinesize, chroma_height,
- sl->chroma_log2_weight_denom,
- sl->chroma_weight[refn0][0][1][0],
- sl->chroma_weight[refn1][1][1][0],
- sl->chroma_weight[refn0][0][1][1] +
- sl->chroma_weight[refn1][1][1][1]);
+ sl->pwt.chroma_log2_weight_denom,
+ sl->pwt.chroma_weight[refn0][0][1][0],
+ sl->pwt.chroma_weight[refn1][1][1][0],
+ sl->pwt.chroma_weight[refn0][0][1][1] +
+ sl->pwt.chroma_weight[refn1][1][1][1]);
}
} else {
int list = list1 ? 1 : 0;
int refn = sl->ref_cache[list][scan8[n]];
- H264Picture *ref = &sl->ref_list[list][refn];
+ H264Ref *ref = &sl->ref_list[list][refn];
mc_dir_part(h, sl, ref, n, square, height, delta, list,
dest_y, dest_cb, dest_cr, x_offset, y_offset,
qpix_put, chroma_put, pixel_shift, chroma_idc);
luma_weight_op(dest_y, sl->mb_linesize, height,
- sl->luma_log2_weight_denom,
- sl->luma_weight[refn][list][0],
- sl->luma_weight[refn][list][1]);
- if (sl->use_weight_chroma) {
+ sl->pwt.luma_log2_weight_denom,
+ sl->pwt.luma_weight[refn][list][0],
+ sl->pwt.luma_weight[refn][list][1]);
+ if (sl->pwt.use_weight_chroma) {
chroma_weight_op(dest_cb, sl->mb_uvlinesize, chroma_height,
- sl->chroma_log2_weight_denom,
- sl->chroma_weight[refn][list][0][0],
- sl->chroma_weight[refn][list][0][1]);
+ sl->pwt.chroma_log2_weight_denom,
+ sl->pwt.chroma_weight[refn][list][0][0],
+ sl->pwt.chroma_weight[refn][list][0][1]);
chroma_weight_op(dest_cr, sl->mb_uvlinesize, chroma_height,
- sl->chroma_log2_weight_denom,
- sl->chroma_weight[refn][list][1][0],
- sl->chroma_weight[refn][list][1][1]);
+ sl->pwt.chroma_log2_weight_denom,
+ sl->pwt.chroma_weight[refn][list][1][0],
+ sl->pwt.chroma_weight[refn][list][1][1]);
}
}
}
-static av_always_inline void prefetch_motion(H264Context *h, H264SliceContext *sl,
+static av_always_inline void prefetch_motion(const H264Context *h, H264SliceContext *sl,
int list, int pixel_shift,
int chroma_idc)
{
* optimized for 64byte cache lines */
const int refn = sl->ref_cache[list][scan8[0]];
if (refn >= 0) {
- const int mx = (sl->mv_cache[list][scan8[0]][0] >> 2) + 16 * h->mb_x + 8;
- const int my = (sl->mv_cache[list][scan8[0]][1] >> 2) + 16 * h->mb_y;
- uint8_t **src = sl->ref_list[list][refn].f.data;
+ const int mx = (sl->mv_cache[list][scan8[0]][0] >> 2) + 16 * sl->mb_x + 8;
+ const int my = (sl->mv_cache[list][scan8[0]][1] >> 2) + 16 * sl->mb_y;
+ uint8_t **src = sl->ref_list[list][refn].data;
int off = (mx << pixel_shift) +
- (my + (h->mb_x & 3) * 4) * sl->mb_linesize +
+ (my + (sl->mb_x & 3) * 4) * sl->mb_linesize +
(64 << pixel_shift);
- h->vdsp.prefetch(src[0] + off, h->linesize, 4);
+ h->vdsp.prefetch(src[0] + off, sl->linesize, 4);
if (chroma_idc == 3 /* yuv444 */) {
- h->vdsp.prefetch(src[1] + off, h->linesize, 4);
- h->vdsp.prefetch(src[2] + off, h->linesize, 4);
+ h->vdsp.prefetch(src[1] + off, sl->linesize, 4);
+ h->vdsp.prefetch(src[2] + off, sl->linesize, 4);
} else {
off = ((mx >> 1) << pixel_shift) +
- ((my >> 1) + (h->mb_x & 7)) * h->uvlinesize +
+ ((my >> 1) + (sl->mb_x & 7)) * sl->uvlinesize +
(64 << pixel_shift);
h->vdsp.prefetch(src[1] + off, src[2] - src[1], 2);
}
}
}
-static av_always_inline void xchg_mb_border(H264Context *h, H264SliceContext *sl,
+static av_always_inline void xchg_mb_border(const H264Context *h, H264SliceContext *sl,
uint8_t *src_y,
uint8_t *src_cb, uint8_t *src_cr,
int linesize, int uvlinesize,
uint8_t *top_border;
if (!simple && FRAME_MBAFF(h)) {
- if (h->mb_y & 1) {
- if (!MB_MBAFF(h))
+ if (sl->mb_y & 1) {
+ if (!MB_MBAFF(sl))
return;
} else {
- top_idx = MB_MBAFF(h) ? 0 : 1;
+ top_idx = MB_MBAFF(sl) ? 0 : 1;
}
}
- if (h->deblocking_filter == 2) {
- deblock_topleft = h->slice_table[h->mb_xy - 1 - h->mb_stride] == sl->slice_num;
+ if (sl->deblocking_filter == 2) {
+ deblock_topleft = h->slice_table[sl->mb_xy - 1 - h->mb_stride] == sl->slice_num;
deblock_top = sl->top_type;
} else {
- deblock_topleft = (h->mb_x > 0);
- deblock_top = (h->mb_y > !!MB_FIELD(h));
+ deblock_topleft = (sl->mb_x > 0);
+ deblock_top = (sl->mb_y > !!MB_FIELD(sl));
}
src_y -= linesize + 1 + pixel_shift;
src_cb -= uvlinesize + 1 + pixel_shift;
src_cr -= uvlinesize + 1 + pixel_shift;
- top_border_m1 = h->top_borders[top_idx][h->mb_x - 1];
- top_border = h->top_borders[top_idx][h->mb_x];
+ top_border_m1 = sl->top_borders[top_idx][sl->mb_x - 1];
+ top_border = sl->top_borders[top_idx][sl->mb_x];
#define XCHG(a, b, xchg) \
if (pixel_shift) { \
}
XCHG(top_border + (0 << pixel_shift), src_y + (1 << pixel_shift), xchg);
XCHG(top_border + (8 << pixel_shift), src_y + (9 << pixel_shift), 1);
- if (h->mb_x + 1 < h->mb_width) {
- XCHG(h->top_borders[top_idx][h->mb_x + 1],
+ if (sl->mb_x + 1 < h->mb_width) {
+ XCHG(sl->top_borders[top_idx][sl->mb_x + 1],
src_y + (17 << pixel_shift), 1);
}
}
- if (simple || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) {
+ if (simple || !CONFIG_GRAY || !(h->flags & AV_CODEC_FLAG_GRAY)) {
if (chroma444) {
if (deblock_top) {
if (deblock_topleft) {
XCHG(top_border + (24 << pixel_shift), src_cb + (9 << pixel_shift), 1);
XCHG(top_border + (32 << pixel_shift), src_cr + (1 << pixel_shift), xchg);
XCHG(top_border + (40 << pixel_shift), src_cr + (9 << pixel_shift), 1);
- if (h->mb_x + 1 < h->mb_width) {
- XCHG(h->top_borders[top_idx][h->mb_x + 1] + (16 << pixel_shift), src_cb + (17 << pixel_shift), 1);
- XCHG(h->top_borders[top_idx][h->mb_x + 1] + (32 << pixel_shift), src_cr + (17 << pixel_shift), 1);
+ if (sl->mb_x + 1 < h->mb_width) {
+ XCHG(sl->top_borders[top_idx][sl->mb_x + 1] + (16 << pixel_shift), src_cb + (17 << pixel_shift), 1);
+ XCHG(sl->top_borders[top_idx][sl->mb_x + 1] + (32 << pixel_shift), src_cr + (17 << pixel_shift), 1);
}
}
} else {
AV_WN16A(mb + index, value);
}
-static av_always_inline void hl_decode_mb_predict_luma(H264Context *h,
+static av_always_inline void hl_decode_mb_predict_luma(const H264Context *h,
H264SliceContext *sl,
- int mb_type, int is_h264,
- int simple,
+ int mb_type, int simple,
int transform_bypass,
int pixel_shift,
- int *block_offset,
+ const int *block_offset,
int linesize,
uint8_t *dest_y, int p)
{
for (i = 0; i < 16; i += 4) {
uint8_t *const ptr = dest_y + block_offset[i];
const int dir = sl->intra4x4_pred_mode_cache[scan8[i]];
- if (transform_bypass && h->sps.profile_idc == 244 && dir <= 1) {
+ if (transform_bypass && h->ps.sps->profile_idc == 244 && dir <= 1) {
h->hpc.pred8x8l_add[dir](ptr, sl->mb + (i * 16 + p * 256 << pixel_shift), linesize);
} else {
const int nnz = sl->non_zero_count_cache[scan8[i + p * 16]];
uint8_t *const ptr = dest_y + block_offset[i];
const int dir = sl->intra4x4_pred_mode_cache[scan8[i]];
- if (transform_bypass && h->sps.profile_idc == 244 && dir <= 1) {
+ if (transform_bypass && h->ps.sps->profile_idc == 244 && dir <= 1) {
h->hpc.pred4x4_add[dir](ptr, sl->mb + (i * 16 + p * 256 << pixel_shift), linesize);
} else {
uint8_t *topright;
uint64_t tr_high;
if (dir == DIAG_DOWN_LEFT_PRED || dir == VERT_LEFT_PRED) {
const int topright_avail = (sl->topright_samples_available << i) & 0x8000;
- assert(h->mb_y || linesize <= block_offset[i]);
+ assert(sl->mb_y || linesize <= block_offset[i]);
if (!topright_avail) {
if (pixel_shift) {
tr_high = ((uint16_t *)ptr)[3 - linesize / 2] * 0x0001000100010001ULL;
h->hpc.pred4x4[dir](ptr, topright, linesize);
nnz = sl->non_zero_count_cache[scan8[i + p * 16]];
if (nnz) {
- if (is_h264) {
- if (nnz == 1 && dctcoef_get(sl->mb, pixel_shift, i * 16 + p * 256))
- idct_dc_add(ptr, sl->mb + (i * 16 + p * 256 << pixel_shift), linesize);
- else
- idct_add(ptr, sl->mb + (i * 16 + p * 256 << pixel_shift), linesize);
- } else if (CONFIG_SVQ3_DECODER)
- ff_svq3_add_idct_c(ptr, sl->mb + i * 16 + p * 256, linesize, qscale, 0);
+ if (nnz == 1 && dctcoef_get(sl->mb, pixel_shift, i * 16 + p * 256))
+ idct_dc_add(ptr, sl->mb + (i * 16 + p * 256 << pixel_shift), linesize);
+ else
+ idct_add(ptr, sl->mb + (i * 16 + p * 256 << pixel_shift), linesize);
}
}
}
}
} else {
h->hpc.pred16x16[sl->intra16x16_pred_mode](dest_y, linesize);
- if (is_h264) {
- if (sl->non_zero_count_cache[scan8[LUMA_DC_BLOCK_INDEX + p]]) {
- if (!transform_bypass)
- h->h264dsp.h264_luma_dc_dequant_idct(sl->mb + (p * 256 << pixel_shift),
- sl->mb_luma_dc[p],
- h->dequant4_coeff[p][qscale][0]);
- else {
- static const uint8_t dc_mapping[16] = {
- 0 * 16, 1 * 16, 4 * 16, 5 * 16,
- 2 * 16, 3 * 16, 6 * 16, 7 * 16,
- 8 * 16, 9 * 16, 12 * 16, 13 * 16,
- 10 * 16, 11 * 16, 14 * 16, 15 * 16
- };
- for (i = 0; i < 16; i++)
- dctcoef_set(sl->mb + (p * 256 << pixel_shift),
- pixel_shift, dc_mapping[i],
- dctcoef_get(sl->mb_luma_dc[p],
- pixel_shift, i));
- }
+ if (sl->non_zero_count_cache[scan8[LUMA_DC_BLOCK_INDEX + p]]) {
+ if (!transform_bypass)
+ h->h264dsp.h264_luma_dc_dequant_idct(sl->mb + (p * 256 << pixel_shift),
+ sl->mb_luma_dc[p],
+ h->ps.pps->dequant4_coeff[p][qscale][0]);
+ else {
+ static const uint8_t dc_mapping[16] = {
+ 0 * 16, 1 * 16, 4 * 16, 5 * 16,
+ 2 * 16, 3 * 16, 6 * 16, 7 * 16,
+ 8 * 16, 9 * 16, 12 * 16, 13 * 16,
+ 10 * 16, 11 * 16, 14 * 16, 15 * 16
+ };
+ for (i = 0; i < 16; i++)
+ dctcoef_set(sl->mb + (p * 256 << pixel_shift),
+ pixel_shift, dc_mapping[i],
+ dctcoef_get(sl->mb_luma_dc[p],
+ pixel_shift, i));
}
- } else if (CONFIG_SVQ3_DECODER)
- ff_svq3_luma_dc_dequant_idct_c(sl->mb + p * 256,
- sl->mb_luma_dc[p], qscale);
+ }
}
}
-static av_always_inline void hl_decode_mb_idct_luma(H264Context *h, H264SliceContext *sl,
- int mb_type,
- int is_h264, int simple,
+static av_always_inline void hl_decode_mb_idct_luma(const H264Context *h, H264SliceContext *sl,
+ int mb_type, int simple,
int transform_bypass,
int pixel_shift,
- int *block_offset,
+ const int *block_offset,
int linesize,
uint8_t *dest_y, int p)
{
int i;
block_offset += 16 * p;
if (!IS_INTRA4x4(mb_type)) {
- if (is_h264) {
- if (IS_INTRA16x16(mb_type)) {
- if (transform_bypass) {
- if (h->sps.profile_idc == 244 &&
- (sl->intra16x16_pred_mode == VERT_PRED8x8 ||
- sl->intra16x16_pred_mode == HOR_PRED8x8)) {
- h->hpc.pred16x16_add[sl->intra16x16_pred_mode](dest_y, block_offset,
- sl->mb + (p * 256 << pixel_shift),
- linesize);
- } else {
- for (i = 0; i < 16; i++)
- if (sl->non_zero_count_cache[scan8[i + p * 16]] ||
- dctcoef_get(sl->mb, pixel_shift, i * 16 + p * 256))
- h->h264dsp.h264_add_pixels4_clear(dest_y + block_offset[i],
- sl->mb + (i * 16 + p * 256 << pixel_shift),
- linesize);
- }
- } else {
- h->h264dsp.h264_idct_add16intra(dest_y, block_offset,
- sl->mb + (p * 256 << pixel_shift),
- linesize,
- sl->non_zero_count_cache + p * 5 * 8);
- }
- } else if (sl->cbp & 15) {
- if (transform_bypass) {
- const int di = IS_8x8DCT(mb_type) ? 4 : 1;
- idct_add = IS_8x8DCT(mb_type) ? h->h264dsp.h264_add_pixels8_clear
- : h->h264dsp.h264_add_pixels4_clear;
- for (i = 0; i < 16; i += di)
- if (sl->non_zero_count_cache[scan8[i + p * 16]])
- idct_add(dest_y + block_offset[i],
- sl->mb + (i * 16 + p * 256 << pixel_shift),
- linesize);
+ if (IS_INTRA16x16(mb_type)) {
+ if (transform_bypass) {
+ if (h->ps.sps->profile_idc == 244 &&
+ (sl->intra16x16_pred_mode == VERT_PRED8x8 ||
+ sl->intra16x16_pred_mode == HOR_PRED8x8)) {
+ h->hpc.pred16x16_add[sl->intra16x16_pred_mode](dest_y, block_offset,
+ sl->mb + (p * 256 << pixel_shift),
+ linesize);
} else {
- if (IS_8x8DCT(mb_type))
- h->h264dsp.h264_idct8_add4(dest_y, block_offset,
- sl->mb + (p * 256 << pixel_shift),
- linesize,
- sl->non_zero_count_cache + p * 5 * 8);
- else
- h->h264dsp.h264_idct_add16(dest_y, block_offset,
- sl->mb + (p * 256 << pixel_shift),
- linesize,
- sl->non_zero_count_cache + p * 5 * 8);
+ for (i = 0; i < 16; i++)
+ if (sl->non_zero_count_cache[scan8[i + p * 16]] ||
+ dctcoef_get(sl->mb, pixel_shift, i * 16 + p * 256))
+ h->h264dsp.h264_add_pixels4_clear(dest_y + block_offset[i],
+ sl->mb + (i * 16 + p * 256 << pixel_shift),
+ linesize);
}
+ } else {
+ h->h264dsp.h264_idct_add16intra(dest_y, block_offset,
+ sl->mb + (p * 256 << pixel_shift),
+ linesize,
+ sl->non_zero_count_cache + p * 5 * 8);
+ }
+ } else if (sl->cbp & 15) {
+ if (transform_bypass) {
+ const int di = IS_8x8DCT(mb_type) ? 4 : 1;
+ idct_add = IS_8x8DCT(mb_type) ? h->h264dsp.h264_add_pixels8_clear
+ : h->h264dsp.h264_add_pixels4_clear;
+ for (i = 0; i < 16; i += di)
+ if (sl->non_zero_count_cache[scan8[i + p * 16]])
+ idct_add(dest_y + block_offset[i],
+ sl->mb + (i * 16 + p * 256 << pixel_shift),
+ linesize);
+ } else {
+ if (IS_8x8DCT(mb_type))
+ h->h264dsp.h264_idct8_add4(dest_y, block_offset,
+ sl->mb + (p * 256 << pixel_shift),
+ linesize,
+ sl->non_zero_count_cache + p * 5 * 8);
+ else
+ h->h264dsp.h264_idct_add16(dest_y, block_offset,
+ sl->mb + (p * 256 << pixel_shift),
+ linesize,
+ sl->non_zero_count_cache + p * 5 * 8);
}
- } else if (CONFIG_SVQ3_DECODER) {
- for (i = 0; i < 16; i++)
- if (sl->non_zero_count_cache[scan8[i + p * 16]] || sl->mb[i * 16 + p * 256]) {
- // FIXME benchmark weird rule, & below
- uint8_t *const ptr = dest_y + block_offset[i];
- ff_svq3_add_idct_c(ptr, sl->mb + i * 16 + p * 256, linesize,
- sl->qscale, IS_INTRA(mb_type) ? 1 : 0);
- }
}
}
}
#define SIMPLE 0
#include "h264_mb_template.c"
-void ff_h264_hl_decode_mb(H264Context *h, H264SliceContext *sl)
+void ff_h264_hl_decode_mb(const H264Context *h, H264SliceContext *sl)
{
- const int mb_xy = h->mb_xy;
+ const int mb_xy = sl->mb_xy;
const int mb_type = h->cur_pic.mb_type[mb_xy];
- int is_complex = CONFIG_SMALL || h->is_complex ||
+ int is_complex = CONFIG_SMALL || sl->is_complex ||
IS_INTRA_PCM(mb_type) || sl->qscale == 0;
if (CHROMA444(h)) {