ref: bbe5ddfdfbded0867a8e8d445a453fd0596f9dae
parent: 7751651ea130b6d49fa1c813e17af6ef577482bf
parent: 72d4890caf2c2b8b740c5bf9e7f8906cd5910d0e
author: Hui Su <huisu@google.com>
date: Wed Jun 15 19:52:18 EDT 2016
Merge "Add vp9 encoder API VP9E_GET_LEVEL to provide bitstream level"
--- a/test/level_test.cc
+++ b/test/level_test.cc
@@ -22,7 +22,9 @@
: EncoderTest(GET_PARAM(0)),
encoding_mode_(GET_PARAM(1)),
cpu_used_(GET_PARAM(2)),
- target_level_(0) {}+ min_gf_internal_(24),
+ target_level_(0),
+ level_(0) {} virtual ~LevelTest() {} virtual void SetUp() {@@ -47,6 +49,7 @@
if (video->frame() == 0) {encoder->Control(VP8E_SET_CPUUSED, cpu_used_);
encoder->Control(VP9E_SET_TARGET_LEVEL, target_level_);
+ encoder->Control(VP9E_SET_MIN_GF_INTERVAL, min_gf_internal_);
if (encoding_mode_ != ::libvpx_test::kRealTime) {encoder->Control(VP8E_SET_ENABLEAUTOALTREF, 1);
encoder->Control(VP8E_SET_ARNR_MAXFRAMES, 7);
@@ -54,20 +57,33 @@
encoder->Control(VP8E_SET_ARNR_TYPE, 3);
}
}
+ encoder->Control(VP9E_GET_LEVEL, &level_);
+ ASSERT_LE(level_, 51);
+ ASSERT_GE(level_, 0);
}
::libvpx_test::TestMode encoding_mode_;
int cpu_used_;
+ int min_gf_internal_;
int target_level_;
+ int level_;
};
+// Test for keeping level stats only
TEST_P(LevelTest, TestTargetLevel0) { ::libvpx_test::I420VideoSource video("hantro_odd.yuv", 208, 144, 30, 1, 0,- 30);
+ 40);
target_level_ = 0;
+ min_gf_internal_ = 4;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+ ASSERT_EQ(11, level_);
+
+ cfg_.rc_target_bitrate = 1600;
+ ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+ ASSERT_EQ(20, level_);
}
+// Test for level control being turned off
TEST_P(LevelTest, TestTargetLevel255) { ::libvpx_test::I420VideoSource video("hantro_odd.yuv", 208, 144, 30, 1, 0,30);
@@ -98,7 +114,6 @@
VP9_INSTANTIATE_TEST_CASE(LevelTest,
::testing::Values(::libvpx_test::kTwoPassGood,
- ::libvpx_test::kOnePassGood,
- ::libvpx_test::kRealTime),
+ ::libvpx_test::kOnePassGood),
::testing::Range(0, 9));
} // namespace
--- a/vp9/common/vp9_common.h
+++ b/vp9/common/vp9_common.h
@@ -67,25 +67,6 @@
#define VP9_FRAME_MARKER 0x2
-typedef enum {- LEVEL_UNKNOWN = 0,
- LEVEL_1 = 10,
- LEVEL_1_1 = 11,
- LEVEL_2 = 20,
- LEVEL_2_1 = 21,
- LEVEL_3 = 30,
- LEVEL_3_1 = 31,
- LEVEL_4 = 40,
- LEVEL_4_1 = 41,
- LEVEL_5 = 50,
- LEVEL_5_1 = 51,
- LEVEL_5_2 = 52,
- LEVEL_6 = 60,
- LEVEL_6_1 = 61,
- LEVEL_6_2 = 62,
- LEVEL_NOT_CARE = 255,
-} VP9_LEVEL;
-
#ifdef __cplusplus
} // extern "C"
#endif
--- a/vp9/common/vp9_onyxc_int.h
+++ b/vp9/common/vp9_onyxc_int.h
@@ -168,8 +168,6 @@
int allow_high_precision_mv;
- int keep_level_stats;
-
// Flag signaling that the frame context should be reset to default values.
// 0 or 1 implies don't reset, 2 reset just the context specified in the
// frame header, 3 reset all contexts.
--- a/vp9/encoder/vp9_bitstream.c
+++ b/vp9/encoder/vp9_bitstream.c
@@ -891,7 +891,7 @@
vpx_wb_write_bit(wb, cm->log2_tile_rows != 1);
}
-static int get_refresh_mask(VP9_COMP *cpi) {+int vp9_get_refresh_mask(VP9_COMP *cpi) { if (vp9_preserve_existing_gf(cpi)) {// We have decided to preserve the previously existing golden frame as our
// new ARF frame. However, in the short term we leave it in the GF slot and,
@@ -1107,11 +1107,11 @@
write_bitdepth_colorspace_sampling(cm, wb);
}
- vpx_wb_write_literal(wb, get_refresh_mask(cpi), REF_FRAMES);
+ vpx_wb_write_literal(wb, vp9_get_refresh_mask(cpi), REF_FRAMES);
write_frame_size(cm, wb);
} else {MV_REFERENCE_FRAME ref_frame;
- vpx_wb_write_literal(wb, get_refresh_mask(cpi), REF_FRAMES);
+ vpx_wb_write_literal(wb, vp9_get_refresh_mask(cpi), REF_FRAMES);
for (ref_frame = LAST_FRAME; ref_frame <= ALTREF_FRAME; ++ref_frame) {assert(get_ref_frame_map_idx(cpi, ref_frame) != INVALID_IDX);
vpx_wb_write_literal(wb, get_ref_frame_map_idx(cpi, ref_frame),
--- a/vp9/encoder/vp9_bitstream.h
+++ b/vp9/encoder/vp9_bitstream.h
@@ -18,6 +18,8 @@
#include "vp9/encoder/vp9_encoder.h"
+int vp9_get_refresh_mask(VP9_COMP *cpi);
+
void vp9_pack_bitstream(VP9_COMP *cpi, uint8_t *dest, size_t *size);
static INLINE int vp9_preserve_existing_gf(VP9_COMP *cpi) {--- a/vp9/encoder/vp9_encoder.c
+++ b/vp9/encoder/vp9_encoder.c
@@ -86,6 +86,25 @@
FILE *keyfile;
#endif
+static const Vp9LevelSpec vp9_level_defs[VP9_LEVELS] = {+ {LEVEL_1, 829440, 36864, 200, 400, 2, 1, 4, 8},+ {LEVEL_1_1, 2764800, 73728, 800, 1000, 2, 1, 4, 8},+ {LEVEL_2, 4608000, 122880, 1800, 1500, 2, 1, 4, 8},+ {LEVEL_2_1, 9216000, 245760, 3600, 2800, 2, 2, 4, 8},+ {LEVEL_3, 20736000, 552960, 7200, 6000, 2, 4, 4, 8},+ {LEVEL_3_1, 36864000, 983040, 12000, 10000, 2, 4, 4, 8},+ {LEVEL_4, 83558400, 2228224, 18000, 16000, 4, 4, 4, 8},+ {LEVEL_4_1, 160432128, 2228224, 30000, 18000, 4, 4, 5, 6},+ {LEVEL_5, 311951360, 8912896, 60000, 36000, 6, 8, 6, 4},+ {LEVEL_5_1, 588251136, 8912896, 120000, 46000, 8, 8, 10, 4},+ // TODO(huisu): update max_cpb_size for level 5_2 ~ 6_2 when
+ // they are finalized (currently TBD).
+ {LEVEL_5_2, 1176502272, 8912896, 180000, 0, 8, 8, 10, 4},+ {LEVEL_6, 1176502272, 35651584, 180000, 0, 8, 16, 10, 4},+ {LEVEL_6_1, 2353004544u, 35651584, 240000, 0, 8, 16, 10, 4},+ {LEVEL_6_2, 4706009088u, 35651584, 480000, 0, 8, 16, 10, 4},+};
+
static INLINE void Scale2Ratio(VPX_SCALING mode, int *hr, int *hs) { switch (mode) {case NORMAL:
@@ -159,6 +178,39 @@
}
}
+static void init_level_info(Vp9LevelInfo *level_info) {+ Vp9LevelStats *const level_stats = &level_info->level_stats;
+ Vp9LevelSpec *const level_spec = &level_info->level_spec;
+
+ memset(level_stats, 0, sizeof(*level_stats));
+ memset(level_spec, 0, sizeof(*level_spec));
+ level_spec->level = LEVEL_UNKNOWN;
+ level_spec->min_altref_distance = INT_MAX;
+}
+
+VP9_LEVEL vp9_get_level(const Vp9LevelSpec * const level_spec) {+ int i;
+ const Vp9LevelSpec *this_level;
+
+ vpx_clear_system_state();
+
+ for (i = 0; i < VP9_LEVELS; ++i) {+ this_level = &vp9_level_defs[i];
+ if ((double)level_spec->max_luma_sample_rate * (1 + SAMPLE_RATE_GRACE_P) >
+ (double)this_level->max_luma_sample_rate ||
+ level_spec->max_luma_picture_size > this_level->max_luma_picture_size ||
+ level_spec->average_bitrate > this_level->average_bitrate ||
+ level_spec->max_cpb_size > this_level->max_cpb_size ||
+ level_spec->compression_ratio < this_level->compression_ratio ||
+ level_spec->max_col_tiles > this_level->max_col_tiles ||
+ level_spec->min_altref_distance < this_level->min_altref_distance ||
+ level_spec->max_ref_frame_buffers > this_level->max_ref_frame_buffers)
+ continue;
+ break;
+ }
+ return (i == VP9_LEVELS) ? LEVEL_UNKNOWN : vp9_level_defs[i].level;
+}
+
int vp9_set_active_map(VP9_COMP* cpi,
unsigned char* new_map_16x16,
int rows,
@@ -783,7 +835,7 @@
cm->color_range = oxcf->color_range;
cpi->target_level = oxcf->target_level;
- cm->keep_level_stats = oxcf->target_level != LEVEL_NOT_CARE;
+ cpi->keep_level_stats = oxcf->target_level != LEVEL_MAX;
cm->width = oxcf->width;
cm->height = oxcf->height;
@@ -1476,7 +1528,7 @@
cm->color_range = oxcf->color_range;
cpi->target_level = oxcf->target_level;
- cm->keep_level_stats = oxcf->target_level != LEVEL_NOT_CARE;
+ cpi->keep_level_stats = oxcf->target_level != LEVEL_MAX;
if (cm->profile <= PROFILE_1)
assert(cm->bit_depth == VPX_BITS_8);
@@ -1660,7 +1712,6 @@
} while (++i <= MV_MAX);
}
-
VP9_COMP *vp9_create_compressor(VP9EncoderConfig *oxcf,
BufferPool *const pool) {unsigned int i;
@@ -1749,6 +1800,9 @@
cpi->multi_arf_last_grp_enabled = 0;
cpi->b_calculate_psnr = CONFIG_INTERNAL_STATS;
+
+ init_level_info(&cpi->level_info);
+
#if CONFIG_INTERNAL_STATS
cpi->b_calculate_ssimg = 0;
cpi->b_calculate_blockiness = 1;
@@ -2798,7 +2852,7 @@
} else if (vp9_preserve_existing_gf(cpi)) {// We have decided to preserve the previously existing golden frame as our
// new ARF frame. However, in the short term in function
- // vp9_bitstream.c::get_refresh_mask() we left it in the GF slot and, if
+ // vp9_get_refresh_mask() we left it in the GF slot and, if
// we're updating the GF with the current decoded frame, we save it to the
// ARF slot instead.
// We now have to update the ARF with the current frame and swap gld_fb_idx
@@ -4420,6 +4474,124 @@
}
#endif // CONFIG_INTERNAL_STATS
+static void update_level_info(VP9_COMP *cpi, size_t *size, int arf_src_index) {+ VP9_COMMON *const cm = &cpi->common;
+ Vp9LevelInfo *const level_info = &cpi->level_info;
+ Vp9LevelSpec *const level_spec = &level_info->level_spec;
+ Vp9LevelStats *const level_stats = &level_info->level_stats;
+ int i, idx;
+ uint64_t luma_samples, dur_end;
+ const uint32_t luma_pic_size = cm->width * cm->height;
+ double cpb_data_size;
+
+ vpx_clear_system_state();
+
+ // update level_stats
+ level_stats->total_compressed_size += *size;
+ if (cm->show_frame) {+ level_stats->total_uncompressed_size +=
+ luma_pic_size +
+ 2 * (luma_pic_size >> (cm->subsampling_x + cm->subsampling_y));
+ level_stats->time_encoded =
+ (cpi->last_end_time_stamp_seen - cpi->first_time_stamp_ever) /
+ (double)TICKS_PER_SEC;
+ }
+
+ if (arf_src_index > 0) {+ if (!level_stats->seen_first_altref) {+ level_stats->seen_first_altref = 1;
+ } else if (level_stats->frames_since_last_altref <
+ level_spec->min_altref_distance) {+ level_spec->min_altref_distance = level_stats->frames_since_last_altref;
+ }
+ level_stats->frames_since_last_altref = 0;
+ } else {+ ++level_stats->frames_since_last_altref;
+ }
+
+ if (level_stats->frame_window_buffer.len < FRAME_WINDOW_SIZE - 1) {+ idx = (level_stats->frame_window_buffer.start +
+ level_stats->frame_window_buffer.len++) % FRAME_WINDOW_SIZE;
+ } else {+ idx = level_stats->frame_window_buffer.start;
+ level_stats->frame_window_buffer.start = (idx + 1) % FRAME_WINDOW_SIZE;
+ }
+ level_stats->frame_window_buffer.buf[idx].ts = cpi->last_time_stamp_seen;
+ level_stats->frame_window_buffer.buf[idx].size = (uint32_t)(*size);
+ level_stats->frame_window_buffer.buf[idx].luma_samples = luma_pic_size;
+
+ if (cm->frame_type == KEY_FRAME) {+ level_stats->ref_refresh_map = 0;
+ } else {+ int count = 0;
+ level_stats->ref_refresh_map |= vp9_get_refresh_mask(cpi);
+ // Also need to consider the case where the encoder refers to a buffer
+ // that has been implicitly refreshed after encoding a keyframe.
+ if (!cm->intra_only) {+ level_stats->ref_refresh_map |= (1 << cpi->lst_fb_idx);
+ level_stats->ref_refresh_map |= (1 << cpi->gld_fb_idx);
+ level_stats->ref_refresh_map |= (1 << cpi->alt_fb_idx);
+ }
+ for (i = 0; i < REF_FRAMES; ++i) {+ count += (level_stats->ref_refresh_map >> i) & 1;
+ }
+ if (count > level_spec->max_ref_frame_buffers) {+ level_spec->max_ref_frame_buffers = count;
+ }
+ }
+
+ // update average_bitrate
+ level_spec->average_bitrate =
+ (double)level_stats->total_compressed_size / 125.0 /
+ level_stats->time_encoded;
+
+ // update max_luma_sample_rate
+ luma_samples = 0;
+ for (i = 0; i < level_stats->frame_window_buffer.len; ++i) {+ idx = (level_stats->frame_window_buffer.start +
+ level_stats->frame_window_buffer.len - 1 - i) % FRAME_WINDOW_SIZE;
+ if (i == 0) {+ dur_end = level_stats->frame_window_buffer.buf[idx].ts;
+ }
+ if (dur_end - level_stats->frame_window_buffer.buf[idx].ts >=
+ TICKS_PER_SEC) {+ break;
+ }
+ luma_samples += level_stats->frame_window_buffer.buf[idx].luma_samples;
+ }
+ if (luma_samples > level_spec->max_luma_sample_rate) {+ level_spec->max_luma_sample_rate = luma_samples;
+ }
+
+ // update max_cpb_size
+ cpb_data_size = 0;
+ for (i = 0; i < CPB_WINDOW_SIZE; ++i) {+ if (i >= level_stats->frame_window_buffer.len) break;
+ idx = (level_stats->frame_window_buffer.start +
+ level_stats->frame_window_buffer.len - 1 - i) % FRAME_WINDOW_SIZE;
+ cpb_data_size += level_stats->frame_window_buffer.buf[idx].size;
+ }
+ cpb_data_size = cpb_data_size / 125.0;
+ if (cpb_data_size > level_spec->max_cpb_size) {+ level_spec->max_cpb_size = cpb_data_size;
+ }
+
+ // update max_luma_picture_size
+ if (luma_pic_size > level_spec->max_luma_picture_size) {+ level_spec->max_luma_picture_size = luma_pic_size;
+ }
+
+ // update compression_ratio
+ level_spec->compression_ratio =
+ (double)level_stats->total_uncompressed_size * cm->bit_depth /
+ level_stats->total_compressed_size / 8.0;
+
+ // update max_col_tiles
+ if (level_spec->max_col_tiles < (1 << cm->log2_tile_cols)) {+ level_spec->max_col_tiles = (1 << cm->log2_tile_cols);
+ }
+}
+
int vp9_get_compressed_data(VP9_COMP *cpi, unsigned int *frame_flags,
size_t *size, uint8_t *dest,
int64_t *time_stamp, int64_t *time_end, int flush) {@@ -4689,6 +4861,9 @@
if (cpi->b_calculate_psnr && oxcf->pass != 1 && cm->show_frame)
generate_psnr_packet(cpi);
+
+ if (cpi->keep_level_stats && oxcf->pass != 1)
+ update_level_info(cpi, size, arf_src_index);
#if CONFIG_INTERNAL_STATS
--- a/vp9/encoder/vp9_encoder.h
+++ b/vp9/encoder/vp9_encoder.h
@@ -20,6 +20,7 @@
#include "vpx_dsp/ssim.h"
#endif
#include "vpx_dsp/variance.h"
+#include "vpx_ports/system_state.h"
#include "vpx_util/vpx_thread.h"
#include "vp9/common/vp9_alloccommon.h"
@@ -51,6 +52,9 @@
extern "C" {#endif
+// vp9 uses 10,000,000 ticks/second as time stamp
+#define TICKS_PER_SEC 10000000
+
typedef struct {int nmvjointcost[MV_JOINTS];
int nmvcosts[2][MV_VALS];
@@ -297,6 +301,69 @@
double worst;
} ImageStat;
+#define CPB_WINDOW_SIZE 4
+#define FRAME_WINDOW_SIZE 128
+#define SAMPLE_RATE_GRACE_P 0.015
+#define VP9_LEVELS 14
+
+typedef enum {+ LEVEL_UNKNOWN = 0,
+ LEVEL_1 = 10,
+ LEVEL_1_1 = 11,
+ LEVEL_2 = 20,
+ LEVEL_2_1 = 21,
+ LEVEL_3 = 30,
+ LEVEL_3_1 = 31,
+ LEVEL_4 = 40,
+ LEVEL_4_1 = 41,
+ LEVEL_5 = 50,
+ LEVEL_5_1 = 51,
+ LEVEL_5_2 = 52,
+ LEVEL_6 = 60,
+ LEVEL_6_1 = 61,
+ LEVEL_6_2 = 62,
+ LEVEL_MAX = 255
+} VP9_LEVEL;
+
+typedef struct {+ VP9_LEVEL level;
+ uint64_t max_luma_sample_rate;
+ uint32_t max_luma_picture_size;
+ double average_bitrate; // in kilobits per second
+ double max_cpb_size; // in kilobits
+ double compression_ratio;
+ uint8_t max_col_tiles;
+ uint32_t min_altref_distance;
+ uint8_t max_ref_frame_buffers;
+} Vp9LevelSpec;
+
+typedef struct {+ int64_t ts; // timestamp
+ uint32_t luma_samples;
+ uint32_t size; // in bytes
+} FrameRecord;
+
+typedef struct {+ FrameRecord buf[FRAME_WINDOW_SIZE];
+ uint8_t start;
+ uint8_t len;
+} FrameWindowBuffer;
+
+typedef struct {+ uint8_t seen_first_altref;
+ uint32_t frames_since_last_altref;
+ uint64_t total_compressed_size;
+ uint64_t total_uncompressed_size;
+ double time_encoded; // in seconds
+ FrameWindowBuffer frame_window_buffer;
+ int ref_refresh_map;
+} Vp9LevelStats;
+
+typedef struct {+ Vp9LevelStats level_stats;
+ Vp9LevelSpec level_spec;
+} Vp9LevelInfo;
+
typedef struct VP9_COMP {QUANTS quants;
ThreadData td;
@@ -519,6 +586,9 @@
VPxWorker *workers;
struct EncWorkerData *tile_thr_data;
VP9LfSync lf_row_sync;
+
+ int keep_level_stats;
+ Vp9LevelInfo level_info;
} VP9_COMP;
void vp9_initialize_enc(void);
@@ -673,6 +743,8 @@
static INLINE int *cond_cost_list(const struct VP9_COMP *cpi, int *cost_list) {return cpi->sf.mv.subpel_search_method != SUBPEL_TREE ? cost_list : NULL;
}
+
+VP9_LEVEL vp9_get_level(const Vp9LevelSpec *const level_spec);
void vp9_new_framerate(VP9_COMP *cpi, double framerate);
--- a/vp9/vp9_cx_iface.c
+++ b/vp9/vp9_cx_iface.c
@@ -205,7 +205,7 @@
level != LEVEL_4 && level != LEVEL_4_1 && level != LEVEL_5 &&
level != LEVEL_5_1 && level != LEVEL_5_2 && level != LEVEL_6 &&
level != LEVEL_6_1 && level != LEVEL_6_2 &&
- level != LEVEL_UNKNOWN && level != LEVEL_NOT_CARE)
+ level != LEVEL_UNKNOWN && level != LEVEL_MAX)
ERROR("target_level is invalid");}
@@ -807,6 +807,13 @@
return update_extra_cfg(ctx, &extra_cfg);
}
+static vpx_codec_err_t ctrl_get_level(vpx_codec_alg_priv_t *ctx, va_list args) {+ int *const arg = va_arg(args, int *);
+ if (arg == NULL) return VPX_CODEC_INVALID_PARAM;
+ *arg = (int)vp9_get_level(&ctx->cpi->level_info.level_spec);
+ return VPX_CODEC_OK;
+}
+
static vpx_codec_err_t encoder_init(vpx_codec_ctx_t *ctx,
vpx_codec_priv_enc_mr_cfg_t *data) {vpx_codec_err_t res = VPX_CODEC_OK;
@@ -964,9 +971,6 @@
return index_sz;
}
-// vp9 uses 10,000,000 ticks/second as time stamp
-#define TICKS_PER_SEC 10000000LL
-
static int64_t timebase_units_to_ticks(const vpx_rational_t *timebase,
int64_t n) {return n * TICKS_PER_SEC * timebase->num / timebase->den;
@@ -1547,6 +1551,7 @@
{VP9_GET_REFERENCE, ctrl_get_reference}, {VP9E_GET_SVC_LAYER_ID, ctrl_get_svc_layer_id}, {VP9E_GET_ACTIVEMAP, ctrl_get_active_map},+ {VP9E_GET_LEVEL, ctrl_get_level}, { -1, NULL},};
--- a/vpx/vp8cx.h
+++ b/vpx/vp8cx.h
@@ -562,7 +562,13 @@
*
* Supported in codecs: VP9
*/
- VP9E_SET_TARGET_LEVEL
+ VP9E_SET_TARGET_LEVEL,
+
+ /*!\brief Codec control function to get bitstream level.
+ *
+ * Supported in codecs: VP9
+ */
+ VP9E_GET_LEVEL
};
/*!\brief vpx 1-D scaling mode
@@ -820,6 +826,9 @@
VPX_CTRL_USE_TYPE(VP9E_SET_TARGET_LEVEL, unsigned int)
#define VPX_CTRL_VP9E_SET_TARGET_LEVEL
+
+VPX_CTRL_USE_TYPE(VP9E_GET_LEVEL, int *)
+#define VPX_CTRL_VP9E_GET_LEVEL
/*!\endcond */
/*! @} - end defgroup vp8_encoder */
--
⑨