ext_tile: add 2 tile encoding modes
Added 2 tile encoding modes:
TILE_NORMAL mode supports the large-scale tile coding up to 1024 x 1024
tiles;
TILE_VR mode supports the large-scale tile coding and also the decoding
of a single tile.
TILE_NORMAL mode allows the non-vr content to be encoded in multiple tiles
without using the coding tool that are designed for vr applications.
Change-Id: Id804806a1bbcb916b7f5dc7e2c5509d88f3defcb
diff --git a/aom/aomcx.h b/aom/aomcx.h
index 9896854..9526015 100644
--- a/aom/aomcx.h
+++ b/aom/aomcx.h
@@ -502,6 +502,15 @@
* Experiment: LOOPFILTERING_ACROSS_TILES
*/
AV1E_SET_TILE_LOOPFILTER,
+
+ /*!\brief Codec control function to set the tile encoding mode to 0 or 1.
+ *
+ * 0 means that the tile encoding mode is TILE_NORMAL, and 1 means that the
+ * tile encoding mode is TILE_VR.
+ *
+ * Experiment: EXT_TILE
+ */
+ AV1E_SET_TILE_ENCODING_MODE,
};
/*!\brief aom 1-D scaling mode
@@ -703,6 +712,9 @@
AOM_CTRL_USE_TYPE(AV1E_SET_ANS_WINDOW_SIZE_LOG2, unsigned int)
#define AOM_CTRL_AV1E_SET_ANS_WINDOW_SIZE_LOG2
+
+AOM_CTRL_USE_TYPE(AV1E_SET_TILE_ENCODING_MODE, unsigned int)
+#define AOM_CTRL_AV1E_SET_TILE_ENCODING_MODE
/*!\endcond */
/*! @} - end defgroup aom_encoder */
#ifdef __cplusplus
diff --git a/aomenc.c b/aomenc.c
index fb43d09..c26d346 100644
--- a/aomenc.c
+++ b/aomenc.c
@@ -370,6 +370,12 @@
static const arg_def_t tile_rows =
ARG_DEF(NULL, "tile-rows", 1,
"Number of tile rows to use, log2 (set to 0 while threads > 1)");
+#if CONFIG_EXT_TILE
+static const arg_def_t tile_encoding_mode =
+ ARG_DEF(NULL, "tile-encoding-mode", 1,
+ "Tile encoding mode (0: normal"
+ " (default), 1: vr)");
+#endif
#if CONFIG_DEPENDENT_HORZTILES
static const arg_def_t tile_dependent_rows =
ARG_DEF(NULL, "tile-dependent-rows", 1, "Enable dependent Tile rows");
@@ -474,6 +480,9 @@
&static_thresh,
&tile_cols,
&tile_rows,
+#if CONFIG_EXT_TILE
+ &tile_encoding_mode,
+#endif
#if CONFIG_DEPENDENT_HORZTILES
&tile_dependent_rows,
#endif
@@ -522,6 +531,9 @@
AOME_SET_STATIC_THRESHOLD,
AV1E_SET_TILE_COLUMNS,
AV1E_SET_TILE_ROWS,
+#if CONFIG_EXT_TILE
+ AV1E_SET_TILE_ENCODING_MODE,
+#endif
#if CONFIG_DEPENDENT_HORZTILES
AV1E_SET_TILE_DEPENDENT_ROWS,
#endif
diff --git a/av1/av1_cx_iface.c b/av1/av1_cx_iface.c
index 4ec146e..6389572 100644
--- a/av1/av1_cx_iface.c
+++ b/av1/av1_cx_iface.c
@@ -75,6 +75,9 @@
#if CONFIG_ANS && ANS_MAX_SYMBOLS
int ans_window_size_log2;
#endif
+#if CONFIG_EXT_TILE
+ unsigned int tile_encoding_mode;
+#endif // CONFIG_EXT_TILE
};
static struct av1_extracfg default_extra_cfg = {
@@ -134,6 +137,9 @@
#if CONFIG_ANS && ANS_MAX_SYMBOLS
23, // ans_window_size_log2
#endif
+#if CONFIG_EXT_TILE
+ 0, // Tile encoding mode is TILE_NORMAL by default.
+#endif // CONFIG_EXT_TILE
};
struct aom_codec_alg_priv {
@@ -270,6 +276,7 @@
if (extra_cfg->tile_rows != UINT_MAX)
RANGE_CHECK(extra_cfg, tile_rows, 1, 64);
}
+ RANGE_CHECK_HI(extra_cfg, tile_encoding_mode, 1);
#else
RANGE_CHECK_HI(extra_cfg, tile_columns, 6);
RANGE_CHECK_HI(extra_cfg, tile_rows, 2);
@@ -520,6 +527,7 @@
#endif // CONFIG_EXT_PARTITION
oxcf->tile_columns = AOMMIN(extra_cfg->tile_columns, max);
oxcf->tile_rows = AOMMIN(extra_cfg->tile_rows, max);
+ oxcf->tile_encoding_mode = extra_cfg->tile_encoding_mode;
}
#else
oxcf->tile_columns = extra_cfg->tile_columns;
@@ -822,6 +830,15 @@
return update_extra_cfg(ctx, &extra_cfg);
}
+#if CONFIG_EXT_TILE
+static aom_codec_err_t ctrl_set_tile_encoding_mode(aom_codec_alg_priv_t *ctx,
+ va_list args) {
+ struct av1_extracfg extra_cfg = ctx->extra_cfg;
+ extra_cfg.tile_encoding_mode = CAST(AV1E_SET_TILE_ENCODING_MODE, args);
+ return update_extra_cfg(ctx, &extra_cfg);
+}
+#endif // CONFIG_EXT_TILE
+
static aom_codec_err_t ctrl_set_aq_mode(aom_codec_alg_priv_t *ctx,
va_list args) {
struct av1_extracfg extra_cfg = ctx->extra_cfg;
@@ -1449,6 +1466,9 @@
#if CONFIG_ANS && ANS_MAX_SYMBOLS
{ AV1E_SET_ANS_WINDOW_SIZE_LOG2, ctrl_set_ans_window_size_log2 },
#endif
+#if CONFIG_EXT_TILE
+ { AV1E_SET_TILE_ENCODING_MODE, ctrl_set_tile_encoding_mode },
+#endif // CONFIG_EXT_TILE
// Getters
{ AOME_GET_LAST_QUANTIZER, ctrl_get_quantizer },
diff --git a/av1/av1_dx_iface.c b/av1/av1_dx_iface.c
index 8cd40a3..f20ea48 100644
--- a/av1/av1_dx_iface.c
+++ b/av1/av1_dx_iface.c
@@ -759,7 +759,8 @@
yuvconfig2image(&ctx->img, &sd, frame_worker_data->user_priv);
#if CONFIG_EXT_TILE
- if (frame_worker_data->pbi->dec_tile_row >= 0) {
+ if (cm->tile_encoding_mode &&
+ frame_worker_data->pbi->dec_tile_row >= 0) {
const int tile_row =
AOMMIN(frame_worker_data->pbi->dec_tile_row, cm->tile_rows - 1);
const int mi_row = tile_row * cm->tile_height;
@@ -774,7 +775,8 @@
AOMMIN(cm->tile_height, cm->mi_rows - mi_row) * MI_SIZE;
}
- if (frame_worker_data->pbi->dec_tile_col >= 0) {
+ if (cm->tile_encoding_mode &&
+ frame_worker_data->pbi->dec_tile_col >= 0) {
const int tile_col =
AOMMIN(frame_worker_data->pbi->dec_tile_col, cm->tile_cols - 1);
const int mi_col = tile_col * cm->tile_width;
diff --git a/av1/common/blockd.h b/av1/common/blockd.h
index eae726c..b2be442 100644
--- a/av1/common/blockd.h
+++ b/av1/common/blockd.h
@@ -861,11 +861,7 @@
#endif // CONFIG_INTRA_INTERP
#endif // CONFIG_EXT_INTRA
-#if CONFIG_EXT_TILE
-#define FIXED_TX_TYPE 1
-#else
#define FIXED_TX_TYPE 0
-#endif
// Converts block_index for given transform size to index of the block in raster
// order.
diff --git a/av1/common/enums.h b/av1/common/enums.h
index b42a326..acfcafd 100644
--- a/av1/common/enums.h
+++ b/av1/common/enums.h
@@ -55,6 +55,9 @@
// Maximum number of tile rows and tile columns
#if CONFIG_EXT_TILE
+#define TILE_NORMAL 0
+#define TILE_VR 1
+
#define MAX_TILE_ROWS 1024
#define MAX_TILE_COLS 1024
#else
diff --git a/av1/common/onyxc_int.h b/av1/common/onyxc_int.h
index 6c28678..5fc92f0 100644
--- a/av1/common/onyxc_int.h
+++ b/av1/common/onyxc_int.h
@@ -356,6 +356,9 @@
#endif // !CONFIG_EXT_TILE
int tile_cols, tile_rows;
int tile_width, tile_height; // In MI units
+#if CONFIG_EXT_TILE
+ unsigned int tile_encoding_mode;
+#endif // CONFIG_EXT_TILE
#if CONFIG_DEPENDENT_HORZTILES
int dependent_horz_tiles;
diff --git a/av1/decoder/decodeframe.c b/av1/decoder/decodeframe.c
index 61fc82f..91b140d 100644
--- a/av1/decoder/decodeframe.c
+++ b/av1/decoder/decodeframe.c
@@ -2934,6 +2934,7 @@
struct aom_read_bit_buffer *const rb) {
AV1_COMMON *const cm = &pbi->common;
#if CONFIG_EXT_TILE
+ cm->tile_encoding_mode = aom_rb_read_literal(rb, 1);
// Read the tile width/height
#if CONFIG_EXT_PARTITION
if (cm->sb_size == BLOCK_128X128) {
@@ -3051,7 +3052,8 @@
const uint8_t **data, aom_decrypt_cb decrypt_cb,
void *decrypt_state,
TileBufferDec (*const tile_buffers)[MAX_TILE_COLS],
- int tile_size_bytes, int col, int row) {
+ int tile_size_bytes, int col, int row,
+ unsigned int tile_encoding_mode) {
size_t size;
size_t copy_size = 0;
@@ -3070,8 +3072,9 @@
size = mem_get_varsize(*data, tile_size_bytes);
}
- // The top bit indicates copy mode
- if ((size >> (tile_size_bytes * 8 - 1)) == 1) {
+ // If cm->tile_encoding_mode = 1 (i.e. TILE_VR), then the top bit of the tile
+ // header indicates copy mode.
+ if (tile_encoding_mode && (size >> (tile_size_bytes * 8 - 1)) == 1) {
// The remaining bits in the top byte signal the row offset
int offset = (size >> (tile_size_bytes - 1) * 8) & 0x7f;
@@ -3167,7 +3170,7 @@
get_tile_buffer(tile_col_data_end[c], &pbi->common.error, &data,
pbi->decrypt_cb, pbi->decrypt_state, tile_buffers,
- tile_size_bytes, c, r);
+ tile_size_bytes, c, r, cm->tile_encoding_mode);
}
}
@@ -3182,7 +3185,7 @@
get_tile_buffer(tile_col_data_end[c], &pbi->common.error, &data,
pbi->decrypt_cb, pbi->decrypt_state, tile_buffers,
- tile_size_bytes, c, r);
+ tile_size_bytes, c, r, cm->tile_encoding_mode);
}
}
}
@@ -3548,7 +3551,9 @@
// after the entire frame is decoded.
#if !CONFIG_VAR_TX && !CONFIG_PARALLEL_DEBLOCKING && !CONFIG_CB4X4
// Loopfilter one tile row.
- if (cm->lf.filter_level && !cm->skip_loop_filter) {
+ // Note: If out-of-order tile decoding is used(for example, inv_row_order
+ // = 1), the loopfiltering has be done after all tile rows are decoded.
+ if (!inv_row_order && cm->lf.filter_level && !cm->skip_loop_filter) {
LFWorkerData *const lf_data = (LFWorkerData *)pbi->lf_worker.data1;
const int lf_start = AOMMAX(0, tile_info.mi_row_start - cm->mib_size);
const int lf_end = tile_info.mi_row_end - cm->mib_size;
@@ -4842,6 +4847,17 @@
first_partition_size = read_uncompressed_header(
pbi, init_read_bit_buffer(pbi, &rb, data, data_end, clear_data));
+
+#if CONFIG_EXT_TILE
+ // If cm->tile_encoding_mode == TILE_NORMAL, the independent decoding of a
+ // single tile or a section of a frame is not allowed.
+ if (!cm->tile_encoding_mode &&
+ (pbi->dec_tile_row >= 0 || pbi->dec_tile_col >= 0)) {
+ pbi->dec_tile_row = -1;
+ pbi->dec_tile_col = -1;
+ }
+#endif // CONFIG_EXT_TILE
+
#if CONFIG_TILE_GROUPS
pbi->first_partition_size = first_partition_size;
pbi->uncomp_hdr_size = aom_rb_bytes_read(&rb);
diff --git a/av1/encoder/bitstream.c b/av1/encoder/bitstream.c
index 993ac99..ce518e6 100644
--- a/av1/encoder/bitstream.c
+++ b/av1/encoder/bitstream.c
@@ -3731,6 +3731,8 @@
assert(tile_width > 0);
assert(tile_height > 0);
+ aom_wb_write_literal(wb, cm->tile_encoding_mode, 1);
+
// Write the tile sizes
#if CONFIG_EXT_PARTITION
if (cm->sb_size == BLOCK_128X128) {
@@ -3990,10 +3992,11 @@
// tile header: size of this tile, or copy offset
uint32_t tile_header = tile_size;
- // Check if this tile is a copy tile.
+ // If the tile_encoding_mode is 1 (i.e. TILE_VR), check if this tile is
+ // a copy tile.
// Very low chances to have copy tiles on the key frames, so don't
// search on key frames to reduce unnecessary search.
- if (cm->frame_type != KEY_FRAME) {
+ if (cm->frame_type != KEY_FRAME && cm->tile_encoding_mode) {
const int idendical_tile_offset =
find_identical_tile(tile_row, tile_col, tile_buffers);
diff --git a/av1/encoder/encoder.c b/av1/encoder/encoder.c
index e430ed4..27a6801 100644
--- a/av1/encoder/encoder.c
+++ b/av1/encoder/encoder.c
@@ -4824,6 +4824,10 @@
}
#endif
+#if CONFIG_EXT_TILE
+ cm->tile_encoding_mode = cpi->oxcf.tile_encoding_mode;
+#endif // CONFIG_EXT_TILE
+
#if CONFIG_XIPHRC
if (drop_this_frame) {
av1_rc_postencode_update_drop_frame(cpi);
diff --git a/av1/encoder/encoder.h b/av1/encoder/encoder.h
index 2299f01..76edaa7 100644
--- a/av1/encoder/encoder.h
+++ b/av1/encoder/encoder.h
@@ -267,6 +267,9 @@
#if CONFIG_ANS && ANS_MAX_SYMBOLS
int ans_window_size_log2;
#endif // CONFIG_ANS && ANS_MAX_SYMBOLS
+#if CONFIG_EXT_TILE
+ unsigned int tile_encoding_mode;
+#endif // CONFIG_EXT_TILE
} AV1EncoderConfig;
static INLINE int is_lossless_requested(const AV1EncoderConfig *cfg) {
diff --git a/av1/encoder/picklpf.c b/av1/encoder/picklpf.c
index b925451..fc0ea48 100644
--- a/av1/encoder/picklpf.c
+++ b/av1/encoder/picklpf.c
@@ -206,6 +206,6 @@
// TODO(any): 0 loopfilter level is only necessary if individual tile
// decoding is required. We need to communicate this requirement to this
// code and force loop filter level 0 only if required.
- lf->filter_level = 0;
+ if (cm->tile_encoding_mode) lf->filter_level = 0;
#endif // CONFIG_EXT_TILE
}
diff --git a/av1/encoder/speed_features.c b/av1/encoder/speed_features.c
index bf99c5d..0f9ea75 100644
--- a/av1/encoder/speed_features.c
+++ b/av1/encoder/speed_features.c
@@ -371,11 +371,7 @@
sf->disable_filter_search_var_thresh = 0;
sf->adaptive_interp_filter_search = 0;
sf->allow_partition_search_skip = 0;
-#if CONFIG_EXT_TILE
- sf->use_upsampled_references = 0;
-#else
sf->use_upsampled_references = 1;
-#endif // CONFIG_EXT_TILE
#if CONFIG_EXT_INTER
sf->disable_wedge_search_var_thresh = 0;
sf->fast_wedge_sign_estimate = 0;
@@ -406,12 +402,8 @@
sf->partition_search_breakout_rate_thr = 0;
sf->simple_model_rd_from_var = 0;
-// Set this at the appropriate speed levels
-#if CONFIG_EXT_TILE
- sf->use_transform_domain_distortion = 1;
-#else
+ // Set this at the appropriate speed levels
sf->use_transform_domain_distortion = 0;
-#endif // CONFIG_EXT_TILE
if (oxcf->mode == GOOD
#if CONFIG_XIPHRC
diff --git a/test/av1_ext_tile_test.cc b/test/av1_ext_tile_test.cc
index cf70734..f964479 100644
--- a/test/av1_ext_tile_test.cc
+++ b/test/av1_ext_tile_test.cc
@@ -30,6 +30,8 @@
const int kImgWidth = 704;
const int kImgHeight = 576;
+// This test tests "tile_encoding_mode = TILE_VR" case. The TILE_NORMAL case is
+// tested by the tile_independence test.
class AV1ExtTileTest
: public ::libaom_test::EncoderTest,
public ::libaom_test::CodecTestWith2Params<libaom_test::TestMode, int> {
@@ -81,6 +83,7 @@
// The tile size is 64x64.
encoder->Control(AV1E_SET_TILE_COLUMNS, kTileSize);
encoder->Control(AV1E_SET_TILE_ROWS, kTileSize);
+ encoder->Control(AV1E_SET_TILE_ENCODING_MODE, 1); // TILE_VR
#if CONFIG_EXT_PARTITION
// Always use 64x64 max partition.
encoder->Control(AV1E_SET_SUPERBLOCK_SIZE, AOM_SUPERBLOCK_SIZE_64X64);
diff --git a/test/tile_independence_test.cc b/test/tile_independence_test.cc
index a98faff..a29051f 100644
--- a/test/tile_independence_test.cc
+++ b/test/tile_independence_test.cc
@@ -62,6 +62,9 @@
if (video->frame() == 1) {
encoder->Control(AV1E_SET_TILE_COLUMNS, n_tile_cols_);
encoder->Control(AV1E_SET_TILE_ROWS, n_tile_rows_);
+#if CONFIG_EXT_TILE
+ encoder->Control(AV1E_SET_TILE_ENCODING_MODE, 0); // TILE_NORMAL
+#endif // CONFIG_EXT_TILE
#if CONFIG_LOOPFILTERING_ACROSS_TILES
encoder->Control(AV1E_SET_TILE_LOOPFILTER, 0);
#endif // CONFIG_LOOPFILTERING_ACROSS_TILES
