aomenc.c

/*
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
 *
 * This source code is subject to the terms of the BSD 2 Clause License and
 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
 * was not distributed with this source code in the LICENSE file, you can
 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
 * Media Patent License 1.0 was not distributed with this source code in the
 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
 */

#include "./aomenc.h"
#include "./aom_config.h"

#include <assert.h>
#include <limits.h>
#include <math.h>
#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

#if CONFIG_LIBYUV
#include "third_party/libyuv/include/libyuv/scale.h"
#endif

#include "aom/aom_encoder.h"
#if CONFIG_DECODERS
#include "aom/aom_decoder.h"
#endif

#include "./args.h"
#include "./ivfenc.h"
#include "./tools_common.h"
#include "examples/encoder_util.h"

#if CONFIG_AV1_ENCODER
#include "aom/aomcx.h"
#endif
#if CONFIG_AV1_DECODER
#include "aom/aomdx.h"
#endif

#include "./aomstats.h"
#include "./rate_hist.h"
#include "./warnings.h"
#include "aom/aom_integer.h"
#include "aom_dsp/aom_dsp_common.h"
#include "aom_ports/aom_timer.h"
#include "aom_ports/mem_ops.h"
#if CONFIG_WEBM_IO
#include "./webmenc.h"
#endif
#include "./y4minput.h"

/* Swallow warnings about unused results of fread/fwrite */
static size_t wrap_fread(void *ptr, size_t size, size_t nmemb, FILE *stream) {
  return fread(ptr, size, nmemb, stream);
}
#define fread wrap_fread

static size_t wrap_fwrite(const void *ptr, size_t size, size_t nmemb,
                          FILE *stream) {
  return fwrite(ptr, size, nmemb, stream);
}
#define fwrite wrap_fwrite

static const char *exec_name;

static void warn_or_exit_on_errorv(aom_codec_ctx_t *ctx, int fatal,
                                   const char *s, va_list ap) {
  if (ctx->err) {
    const char *detail = aom_codec_error_detail(ctx);

    vfprintf(stderr, s, ap);
    fprintf(stderr, ": %s\n", aom_codec_error(ctx));

    if (detail) fprintf(stderr, "    %s\n", detail);

    if (fatal) exit(EXIT_FAILURE);
  }
}

static void ctx_exit_on_error(aom_codec_ctx_t *ctx, const char *s, ...) {
  va_list ap;

  va_start(ap, s);
  warn_or_exit_on_errorv(ctx, 1, s, ap);
  va_end(ap);
}

static void warn_or_exit_on_error(aom_codec_ctx_t *ctx, int fatal,
                                  const char *s, ...) {
  va_list ap;

  va_start(ap, s);
  warn_or_exit_on_errorv(ctx, fatal, s, ap);
  va_end(ap);
}

static int read_frame(struct AvxInputContext *input_ctx, aom_image_t *img) {
  FILE *f = input_ctx->file;
  y4m_input *y4m = &input_ctx->y4m;
  int shortread = 0;

  if (input_ctx->file_type == FILE_TYPE_Y4M) {
    if (y4m_input_fetch_frame(y4m, f, img) < 1) return 0;
  } else {
    shortread = read_yuv_frame(input_ctx, img);
  }

  return !shortread;
}

static int file_is_y4m(const char detect[4]) {
  if (memcmp(detect, "YUV4", 4) == 0) {
    return 1;
  }
  return 0;
}

static int fourcc_is_ivf(const char detect[4]) {
  if (memcmp(detect, "DKIF", 4) == 0) {
    return 1;
  }
  return 0;
}

static const arg_def_t debugmode =
    ARG_DEF("D", "debug", 0, "Debug mode (makes output deterministic)");
static const arg_def_t outputfile =
    ARG_DEF("o", "output", 1, "Output filename");
static const arg_def_t use_yv12 =
    ARG_DEF(NULL, "yv12", 0, "Input file is YV12 ");
static const arg_def_t use_i420 =
    ARG_DEF(NULL, "i420", 0, "Input file is I420 (default)");
static const arg_def_t use_i422 =
    ARG_DEF(NULL, "i422", 0, "Input file is I422");
static const arg_def_t use_i444 =
    ARG_DEF(NULL, "i444", 0, "Input file is I444");
static const arg_def_t use_i440 =
    ARG_DEF(NULL, "i440", 0, "Input file is I440");
static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1, "Codec to use");
static const arg_def_t passes =
    ARG_DEF("p", "passes", 1, "Number of passes (1/2)");
static const arg_def_t pass_arg =
    ARG_DEF(NULL, "pass", 1, "Pass to execute (1/2)");
static const arg_def_t fpf_name =
    ARG_DEF(NULL, "fpf", 1, "First pass statistics file name");
#if CONFIG_FP_MB_STATS
static const arg_def_t fpmbf_name =
    ARG_DEF(NULL, "fpmbf", 1, "First pass block statistics file name");
#endif
static const arg_def_t limit =
    ARG_DEF(NULL, "limit", 1, "Stop encoding after n input frames");
static const arg_def_t skip =
    ARG_DEF(NULL, "skip", 1, "Skip the first n input frames");
static const arg_def_t deadline =
    ARG_DEF("d", "deadline", 1, "Deadline per frame (usec)");
static const arg_def_t good_dl =
    ARG_DEF(NULL, "good", 0, "Use Good Quality Deadline");
static const arg_def_t quietarg =
    ARG_DEF("q", "quiet", 0, "Do not print encode progress");
static const arg_def_t verbosearg =
    ARG_DEF("v", "verbose", 0, "Show encoder parameters");
static const arg_def_t psnrarg =
    ARG_DEF(NULL, "psnr", 0, "Show PSNR in status line");

static const struct arg_enum_list test_decode_enum[] = {
  { "off", TEST_DECODE_OFF },
  { "fatal", TEST_DECODE_FATAL },
  { "warn", TEST_DECODE_WARN },
  { NULL, 0 }
};
static const arg_def_t recontest = ARG_DEF_ENUM(
    NULL, "test-decode", 1, "Test encode/decode mismatch", test_decode_enum);
static const arg_def_t framerate =
    ARG_DEF(NULL, "fps", 1, "Stream frame rate (rate/scale)");
static const arg_def_t use_webm =
    ARG_DEF(NULL, "webm", 0, "Output WebM (default when WebM IO is enabled)");
static const arg_def_t use_ivf = ARG_DEF(NULL, "ivf", 0, "Output IVF");
static const arg_def_t out_part =
    ARG_DEF("P", "output-partitions", 0,
            "Makes encoder output partitions. Requires IVF output!");
static const arg_def_t q_hist_n =
    ARG_DEF(NULL, "q-hist", 1, "Show quantizer histogram (n-buckets)");
static const arg_def_t rate_hist_n =
    ARG_DEF(NULL, "rate-hist", 1, "Show rate histogram (n-buckets)");
static const arg_def_t disable_warnings =
    ARG_DEF(NULL, "disable-warnings", 0,
            "Disable warnings about potentially incorrect encode settings.");
static const arg_def_t disable_warning_prompt =
    ARG_DEF("y", "disable-warning-prompt", 0,
            "Display warnings, but do not prompt user to continue.");

#if CONFIG_HIGHBITDEPTH
static const arg_def_t test16bitinternalarg = ARG_DEF(
    NULL, "test-16bit-internal", 0, "Force use of 16 bit internal buffer");

static const struct arg_enum_list bitdepth_enum[] = {
  { "8", AOM_BITS_8 }, { "10", AOM_BITS_10 }, { "12", AOM_BITS_12 }, { NULL, 0 }
};

static const arg_def_t bitdeptharg = ARG_DEF_ENUM(
    "b", "bit-depth", 1,
    "Bit depth for codec (8 for version <=1, 10 or 12 for version 2)",
    bitdepth_enum);
static const arg_def_t inbitdeptharg =
    ARG_DEF(NULL, "input-bit-depth", 1, "Bit depth of input");
#endif

static const arg_def_t *main_args[] = { &debugmode,
                                        &outputfile,
                                        &codecarg,
                                        &passes,
                                        &pass_arg,
                                        &fpf_name,
                                        &limit,
                                        &skip,
                                        &deadline,
                                        &good_dl,
                                        &quietarg,
                                        &verbosearg,
                                        &psnrarg,
                                        &use_webm,
                                        &use_ivf,
                                        &out_part,
                                        &q_hist_n,
                                        &rate_hist_n,
                                        &disable_warnings,
                                        &disable_warning_prompt,
                                        &recontest,
                                        NULL };

static const arg_def_t usage =
    ARG_DEF("u", "usage", 1, "Usage profile number to use");
static const arg_def_t threads =
    ARG_DEF("t", "threads", 1, "Max number of threads to use");
static const arg_def_t profile =
    ARG_DEF(NULL, "profile", 1, "Bitstream profile number to use");
static const arg_def_t width = ARG_DEF("w", "width", 1, "Frame width");
static const arg_def_t height = ARG_DEF("h", "height", 1, "Frame height");
#if CONFIG_WEBM_IO
static const struct arg_enum_list stereo_mode_enum[] = {
  { "mono", STEREO_FORMAT_MONO },
  { "left-right", STEREO_FORMAT_LEFT_RIGHT },
  { "bottom-top", STEREO_FORMAT_BOTTOM_TOP },
  { "top-bottom", STEREO_FORMAT_TOP_BOTTOM },
  { "right-left", STEREO_FORMAT_RIGHT_LEFT },
  { NULL, 0 }
};
static const arg_def_t stereo_mode = ARG_DEF_ENUM(
    NULL, "stereo-mode", 1, "Stereo 3D video format", stereo_mode_enum);
#endif
static const arg_def_t timebase = ARG_DEF(
    NULL, "timebase", 1, "Output timestamp precision (fractional seconds)");
static const arg_def_t error_resilient =
    ARG_DEF(NULL, "error-resilient", 1, "Enable error resiliency features");
static const arg_def_t lag_in_frames =
    ARG_DEF(NULL, "lag-in-frames", 1, "Max number of frames to lag");

static const arg_def_t *global_args[] = { &use_yv12,
                                          &use_i420,
                                          &use_i422,
                                          &use_i444,
                                          &use_i440,
                                          &usage,
                                          &threads,
                                          &profile,
                                          &width,
                                          &height,
#if CONFIG_WEBM_IO
                                          &stereo_mode,
#endif
                                          &timebase,
                                          &framerate,
                                          &error_resilient,
#if CONFIG_HIGHBITDEPTH
                                          &test16bitinternalarg,
                                          &bitdeptharg,
#endif
                                          &lag_in_frames,
                                          NULL };

static const arg_def_t dropframe_thresh =
    ARG_DEF(NULL, "drop-frame", 1, "Temporal resampling threshold (buf %)");
static const arg_def_t resize_allowed =
    ARG_DEF(NULL, "resize-allowed", 1, "Spatial resampling enabled (bool)");
static const arg_def_t resize_width =
    ARG_DEF(NULL, "resize-width", 1, "Width of encoded frame");
static const arg_def_t resize_height =
    ARG_DEF(NULL, "resize-height", 1, "Height of encoded frame");
static const arg_def_t resize_up_thresh =
    ARG_DEF(NULL, "resize-up", 1, "Upscale threshold (buf %)");
static const arg_def_t resize_down_thresh =
    ARG_DEF(NULL, "resize-down", 1, "Downscale threshold (buf %)");
static const struct arg_enum_list end_usage_enum[] = { { "vbr", AOM_VBR },
                                                       { "cbr", AOM_CBR },
                                                       { "cq", AOM_CQ },
                                                       { "q", AOM_Q },
                                                       { NULL, 0 } };
static const arg_def_t end_usage =
    ARG_DEF_ENUM(NULL, "end-usage", 1, "Rate control mode", end_usage_enum);
static const arg_def_t target_bitrate =
    ARG_DEF(NULL, "target-bitrate", 1, "Bitrate (kbps)");
static const arg_def_t min_quantizer =
    ARG_DEF(NULL, "min-q", 1, "Minimum (best) quantizer");
static const arg_def_t max_quantizer =
    ARG_DEF(NULL, "max-q", 1, "Maximum (worst) quantizer");
static const arg_def_t undershoot_pct =
    ARG_DEF(NULL, "undershoot-pct", 1, "Datarate undershoot (min) target (%)");
static const arg_def_t overshoot_pct =
    ARG_DEF(NULL, "overshoot-pct", 1, "Datarate overshoot (max) target (%)");
static const arg_def_t buf_sz =
    ARG_DEF(NULL, "buf-sz", 1, "Client buffer size (ms)");
static const arg_def_t buf_initial_sz =
    ARG_DEF(NULL, "buf-initial-sz", 1, "Client initial buffer size (ms)");
static const arg_def_t buf_optimal_sz =
    ARG_DEF(NULL, "buf-optimal-sz", 1, "Client optimal buffer size (ms)");
static const arg_def_t *rc_args[] = {
  &dropframe_thresh, &resize_allowed,     &resize_width,   &resize_height,
  &resize_up_thresh, &resize_down_thresh, &end_usage,      &target_bitrate,
  &min_quantizer,    &max_quantizer,      &undershoot_pct, &overshoot_pct,
  &buf_sz,           &buf_initial_sz,     &buf_optimal_sz, NULL
};

static const arg_def_t bias_pct =
    ARG_DEF(NULL, "bias-pct", 1, "CBR/VBR bias (0=CBR, 100=VBR)");
static const arg_def_t minsection_pct =
    ARG_DEF(NULL, "minsection-pct", 1, "GOP min bitrate (% of target)");
static const arg_def_t maxsection_pct =
    ARG_DEF(NULL, "maxsection-pct", 1, "GOP max bitrate (% of target)");
static const arg_def_t *rc_twopass_args[] = { &bias_pct, &minsection_pct,
                                              &maxsection_pct, NULL };

static const arg_def_t kf_min_dist =
    ARG_DEF(NULL, "kf-min-dist", 1, "Minimum keyframe interval (frames)");
static const arg_def_t kf_max_dist =
    ARG_DEF(NULL, "kf-max-dist", 1, "Maximum keyframe interval (frames)");
static const arg_def_t kf_disabled =
    ARG_DEF(NULL, "disable-kf", 0, "Disable keyframe placement");
static const arg_def_t *kf_args[] = { &kf_min_dist, &kf_max_dist, &kf_disabled,
                                      NULL };

static const arg_def_t noise_sens =
    ARG_DEF(NULL, "noise-sensitivity", 1, "Noise sensitivity (frames to blur)");
static const arg_def_t sharpness =
    ARG_DEF(NULL, "sharpness", 1, "Loop filter sharpness (0..7)");
static const arg_def_t static_thresh =
    ARG_DEF(NULL, "static-thresh", 1, "Motion detection threshold");
static const arg_def_t auto_altref =
    ARG_DEF(NULL, "auto-alt-ref", 1, "Enable automatic alt reference frames");
static const arg_def_t arnr_maxframes =
    ARG_DEF(NULL, "arnr-maxframes", 1, "AltRef max frames (0..15)");
static const arg_def_t arnr_strength =
    ARG_DEF(NULL, "arnr-strength", 1, "AltRef filter strength (0..6)");
static const struct arg_enum_list tuning_enum[] = {
  { "psnr", AOM_TUNE_PSNR }, { "ssim", AOM_TUNE_SSIM }, { NULL, 0 }
};
static const arg_def_t tune_ssim =
    ARG_DEF_ENUM(NULL, "tune", 1, "Material to favor", tuning_enum);
static const arg_def_t cq_level =
    ARG_DEF(NULL, "cq-level", 1, "Constant/Constrained Quality level");
static const arg_def_t max_intra_rate_pct =
    ARG_DEF(NULL, "max-intra-rate", 1, "Max I-frame bitrate (pct)");

#if CONFIG_AV1_ENCODER
static const arg_def_t cpu_used_av1 =
    ARG_DEF(NULL, "cpu-used", 1, "CPU Used (0..8)");
static const arg_def_t tile_cols =
    ARG_DEF(NULL, "tile-columns", 1, "Number of tile columns to use, log2");
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");
#endif
#if CONFIG_LOOPFILTERING_ACROSS_TILES
static const arg_def_t tile_loopfilter = ARG_DEF(
    NULL, "tile-loopfilter", 1, "Enable loop filter across tile boundary");
#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES
static const arg_def_t lossless =
    ARG_DEF(NULL, "lossless", 1, "Lossless mode (0: false (default), 1: true)");
#if CONFIG_AOM_QM
static const arg_def_t enable_qm =
    ARG_DEF(NULL, "enable-qm", 1,
            "Enable quantisation matrices (0: false (default), 1: true)");
static const arg_def_t qm_min = ARG_DEF(
    NULL, "qm-min", 1, "Min quant matrix flatness (0..15), default is 8");
static const arg_def_t qm_max = ARG_DEF(
    NULL, "qm-max", 1, "Max quant matrix flatness (0..15), default is 16");
#endif
#if CONFIG_TILE_GROUPS
static const arg_def_t num_tg = ARG_DEF(
    NULL, "num-tile-groups", 1, "Maximum number of tile groups, default is 1");
static const arg_def_t mtu_size =
    ARG_DEF(NULL, "mtu-size", 1,
            "MTU size for a tile group, default is 0 (no MTU targeting), "
            "overrides maximum number of tile groups");
#endif
#if CONFIG_TEMPMV_SIGNALING
static const arg_def_t disable_tempmv = ARG_DEF(
    NULL, "disable-tempmv", 1, "Disable temporal mv prediction (default is 0)");
#endif
static const arg_def_t frame_parallel_decoding =
    ARG_DEF(NULL, "frame-parallel", 1,
            "Enable frame parallel decodability features "
            "(0: false (default), 1: true)");
#if CONFIG_DELTA_Q && !CONFIG_EXT_DELTA_Q
static const arg_def_t aq_mode = ARG_DEF(
    NULL, "aq-mode", 1,
    "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
    "3: cyclic refresh, 4: delta quant)");
#else
static const arg_def_t aq_mode = ARG_DEF(
    NULL, "aq-mode", 1,
    "Adaptive quantization mode (0: off (default), 1: variance 2: complexity, "
    "3: cyclic refresh)");
#endif
#if CONFIG_EXT_DELTA_Q
static const arg_def_t deltaq_mode = ARG_DEF(
    NULL, "deltaq-mode", 1,
    "Delta qindex mode (0: off (default), 1: deltaq 2: deltaq + deltalf)");
#endif
static const arg_def_t frame_periodic_boost =
    ARG_DEF(NULL, "frame-boost", 1,
            "Enable frame periodic boost (0: off (default), 1: on)");
static const arg_def_t gf_cbr_boost_pct = ARG_DEF(
    NULL, "gf-cbr-boost", 1, "Boost for Golden Frame in CBR mode (pct)");
static const arg_def_t max_inter_rate_pct =
    ARG_DEF(NULL, "max-inter-rate", 1, "Max P-frame bitrate (pct)");
static const arg_def_t min_gf_interval = ARG_DEF(
    NULL, "min-gf-interval", 1,
    "min gf/arf frame interval (default 0, indicating in-built behavior)");
static const arg_def_t max_gf_interval = ARG_DEF(
    NULL, "max-gf-interval", 1,
    "max gf/arf frame interval (default 0, indicating in-built behavior)");

static const struct arg_enum_list color_space_enum[] = {
  { "unknown", AOM_CS_UNKNOWN },
  { "bt601", AOM_CS_BT_601 },
  { "bt709", AOM_CS_BT_709 },
  { "smpte170", AOM_CS_SMPTE_170 },
  { "smpte240", AOM_CS_SMPTE_240 },
  { "bt2020", AOM_CS_BT_2020 },
  { "reserved", AOM_CS_RESERVED },
  { "sRGB", AOM_CS_SRGB },
  { NULL, 0 }
};

static const arg_def_t input_color_space =
    ARG_DEF_ENUM(NULL, "color-space", 1, "The color space of input content:",
                 color_space_enum);

static const struct arg_enum_list tune_content_enum[] = {
  { "default", AOM_CONTENT_DEFAULT },
  { "screen", AOM_CONTENT_SCREEN },
  { NULL, 0 }
};

static const arg_def_t tune_content = ARG_DEF_ENUM(
    NULL, "tune-content", 1, "Tune content type", tune_content_enum);
#endif

#if CONFIG_AV1_ENCODER
#if CONFIG_EXT_PARTITION
static const struct arg_enum_list superblock_size_enum[] = {
  { "dynamic", AOM_SUPERBLOCK_SIZE_DYNAMIC },
  { "64", AOM_SUPERBLOCK_SIZE_64X64 },
  { "128", AOM_SUPERBLOCK_SIZE_128X128 },
  { NULL, 0 }
};
static const arg_def_t superblock_size = ARG_DEF_ENUM(
    NULL, "sb-size", 1, "Superblock size to use", superblock_size_enum);
#endif  // CONFIG_EXT_PARTITION

static const arg_def_t *av1_args[] = { &cpu_used_av1,
                                       &auto_altref,
                                       &sharpness,
                                       &static_thresh,
                                       &tile_cols,
                                       &tile_rows,
#if CONFIG_EXT_TILE
                                       &tile_encoding_mode,
#endif
#if CONFIG_DEPENDENT_HORZTILES
                                       &tile_dependent_rows,
#endif
#if CONFIG_LOOPFILTERING_ACROSS_TILES
                                       &tile_loopfilter,
#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES
                                       &arnr_maxframes,
                                       &arnr_strength,
                                       &tune_ssim,
                                       &cq_level,
                                       &max_intra_rate_pct,
                                       &max_inter_rate_pct,
                                       &gf_cbr_boost_pct,
                                       &lossless,
#if CONFIG_AOM_QM
                                       &enable_qm,
                                       &qm_min,
                                       &qm_max,
#endif
                                       &frame_parallel_decoding,
                                       &aq_mode,
#if CONFIG_EXT_DELTA_Q
                                       &deltaq_mode,
#endif
                                       &frame_periodic_boost,
                                       &noise_sens,
                                       &tune_content,
                                       &input_color_space,
                                       &min_gf_interval,
                                       &max_gf_interval,
#if CONFIG_EXT_PARTITION
                                       &superblock_size,
#endif  // CONFIG_EXT_PARTITION
#if CONFIG_TILE_GROUPS
                                       &num_tg,
                                       &mtu_size,
#endif
#if CONFIG_TEMPMV_SIGNALING
                                       &disable_tempmv,
#endif
#if CONFIG_HIGHBITDEPTH
                                       &bitdeptharg,
                                       &inbitdeptharg,
#endif  // CONFIG_HIGHBITDEPTH
                                       NULL };
static const int av1_arg_ctrl_map[] = { AOME_SET_CPUUSED,
                                        AOME_SET_ENABLEAUTOALTREF,
                                        AOME_SET_SHARPNESS,
                                        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
#if CONFIG_LOOPFILTERING_ACROSS_TILES
                                        AV1E_SET_TILE_LOOPFILTER,
#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES
                                        AOME_SET_ARNR_MAXFRAMES,
                                        AOME_SET_ARNR_STRENGTH,
                                        AOME_SET_TUNING,
                                        AOME_SET_CQ_LEVEL,
                                        AOME_SET_MAX_INTRA_BITRATE_PCT,
                                        AV1E_SET_MAX_INTER_BITRATE_PCT,
                                        AV1E_SET_GF_CBR_BOOST_PCT,
                                        AV1E_SET_LOSSLESS,
#if CONFIG_AOM_QM
                                        AV1E_SET_ENABLE_QM,
                                        AV1E_SET_QM_MIN,
                                        AV1E_SET_QM_MAX,
#endif
                                        AV1E_SET_FRAME_PARALLEL_DECODING,
                                        AV1E_SET_AQ_MODE,
#if CONFIG_EXT_DELTA_Q
                                        AV1E_SET_DELTAQ_MODE,
#endif
                                        AV1E_SET_FRAME_PERIODIC_BOOST,
                                        AV1E_SET_NOISE_SENSITIVITY,
                                        AV1E_SET_TUNE_CONTENT,
                                        AV1E_SET_COLOR_SPACE,
                                        AV1E_SET_MIN_GF_INTERVAL,
                                        AV1E_SET_MAX_GF_INTERVAL,
#if CONFIG_EXT_PARTITION
                                        AV1E_SET_SUPERBLOCK_SIZE,
#endif  // CONFIG_EXT_PARTITION
#if CONFIG_TILE_GROUPS
                                        AV1E_SET_NUM_TG,
                                        AV1E_SET_MTU,
#endif
#if CONFIG_TEMPMV_SIGNALING
                                        AV1E_SET_DISABLE_TEMPMV,
#endif
                                        0 };
#endif

static const arg_def_t *no_args[] = { NULL };

void usage_exit(void) {
  int i;
  const int num_encoder = get_aom_encoder_count();

  fprintf(stderr, "Usage: %s <options> -o dst_filename src_filename \n",
          exec_name);

  fprintf(stderr, "\nOptions:\n");
  arg_show_usage(stderr, main_args);
  fprintf(stderr, "\nEncoder Global Options:\n");
  arg_show_usage(stderr, global_args);
  fprintf(stderr, "\nRate Control Options:\n");
  arg_show_usage(stderr, rc_args);
  fprintf(stderr, "\nTwopass Rate Control Options:\n");
  arg_show_usage(stderr, rc_twopass_args);
  fprintf(stderr, "\nKeyframe Placement Options:\n");
  arg_show_usage(stderr, kf_args);
#if CONFIG_AV1_ENCODER
  fprintf(stderr, "\nAV1 Specific Options:\n");
  arg_show_usage(stderr, av1_args);
#endif
  fprintf(stderr,
          "\nStream timebase (--timebase):\n"
          "  The desired precision of timestamps in the output, expressed\n"
          "  in fractional seconds. Default is 1/1000.\n");
  fprintf(stderr, "\nIncluded encoders:\n\n");

  for (i = 0; i < num_encoder; ++i) {
    const AvxInterface *const encoder = get_aom_encoder_by_index(i);
    const char *defstr = (i == (num_encoder - 1)) ? "(default)" : "";
    fprintf(stderr, "    %-6s - %s %s\n", encoder->name,
            aom_codec_iface_name(encoder->codec_interface()), defstr);
  }
  fprintf(stderr, "\n        ");
  fprintf(stderr, "Use --codec to switch to a non-default encoder.\n\n");

  exit(EXIT_FAILURE);
}

#if CONFIG_AV1_ENCODER
#define ARG_CTRL_CNT_MAX NELEMENTS(av1_arg_ctrl_map)
#endif

#if !CONFIG_WEBM_IO
typedef int stereo_format_t;
struct WebmOutputContext {
  int debug;
};
#endif

/* Per-stream configuration */
struct stream_config {
  struct aom_codec_enc_cfg cfg;
  const char *out_fn;
  const char *stats_fn;
#if CONFIG_FP_MB_STATS
  const char *fpmb_stats_fn;
#endif
  stereo_format_t stereo_fmt;
  int arg_ctrls[ARG_CTRL_CNT_MAX][2];
  int arg_ctrl_cnt;
  int write_webm;
#if CONFIG_HIGHBITDEPTH
  // whether to use 16bit internal buffers
  int use_16bit_internal;
#endif
};

struct stream_state {
  int index;
  struct stream_state *next;
  struct stream_config config;
  FILE *file;
  struct rate_hist *rate_hist;
  struct WebmOutputContext webm_ctx;
  uint64_t psnr_sse_total;
  uint64_t psnr_samples_total;
  double psnr_totals[4];
  int psnr_count;
  int counts[64];
  aom_codec_ctx_t encoder;
  unsigned int frames_out;
  uint64_t cx_time;
  size_t nbytes;
  stats_io_t stats;
#if CONFIG_FP_MB_STATS
  stats_io_t fpmb_stats;
#endif
  struct aom_image *img;
  aom_codec_ctx_t decoder;
  int mismatch_seen;
};

static void validate_positive_rational(const char *msg,
                                       struct aom_rational *rat) {
  if (rat->den < 0) {
    rat->num *= -1;
    rat->den *= -1;
  }

  if (rat->num < 0) die("Error: %s must be positive\n", msg);

  if (!rat->den) die("Error: %s has zero denominator\n", msg);
}

static void parse_global_config(struct AvxEncoderConfig *global, char **argv) {
  char **argi, **argj;
  struct arg arg;
  const int num_encoder = get_aom_encoder_count();

  if (num_encoder < 1) die("Error: no valid encoder available\n");

  /* Initialize default parameters */
  memset(global, 0, sizeof(*global));
  global->codec = get_aom_encoder_by_index(num_encoder - 1);
  global->passes = 0;
  global->color_type = I420;
  /* Assign default deadline to good quality */
  global->deadline = AOM_DL_GOOD_QUALITY;

  for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
    arg.argv_step = 1;

    if (arg_match(&arg, &codecarg, argi)) {
      global->codec = get_aom_encoder_by_name(arg.val);
      if (!global->codec)
        die("Error: Unrecognized argument (%s) to --codec\n", arg.val);
    } else if (arg_match(&arg, &passes, argi)) {
      global->passes = arg_parse_uint(&arg);

      if (global->passes < 1 || global->passes > 2)
        die("Error: Invalid number of passes (%d)\n", global->passes);
    } else if (arg_match(&arg, &pass_arg, argi)) {
      global->pass = arg_parse_uint(&arg);

      if (global->pass < 1 || global->pass > 2)
        die("Error: Invalid pass selected (%d)\n", global->pass);
    } else if (arg_match(&arg, &usage, argi))
      global->usage = arg_parse_uint(&arg);
    else if (arg_match(&arg, &deadline, argi))
      global->deadline = arg_parse_uint(&arg);
    else if (arg_match(&arg, &good_dl, argi))
      global->deadline = AOM_DL_GOOD_QUALITY;
    else if (arg_match(&arg, &use_yv12, argi))
      global->color_type = YV12;
    else if (arg_match(&arg, &use_i420, argi))
      global->color_type = I420;
    else if (arg_match(&arg, &use_i422, argi))
      global->color_type = I422;
    else if (arg_match(&arg, &use_i444, argi))
      global->color_type = I444;
    else if (arg_match(&arg, &use_i440, argi))
      global->color_type = I440;
    else if (arg_match(&arg, &quietarg, argi))
      global->quiet = 1;
    else if (arg_match(&arg, &verbosearg, argi))
      global->verbose = 1;
    else if (arg_match(&arg, &limit, argi))
      global->limit = arg_parse_uint(&arg);
    else if (arg_match(&arg, &skip, argi))
      global->skip_frames = arg_parse_uint(&arg);
    else if (arg_match(&arg, &psnrarg, argi))
      global->show_psnr = 1;
    else if (arg_match(&arg, &recontest, argi))
      global->test_decode = arg_parse_enum_or_int(&arg);
    else if (arg_match(&arg, &framerate, argi)) {
      global->framerate = arg_parse_rational(&arg);
      validate_positive_rational(arg.name, &global->framerate);
      global->have_framerate = 1;
    } else if (arg_match(&arg, &out_part, argi))
      global->out_part = 1;
    else if (arg_match(&arg, &debugmode, argi))
      global->debug = 1;
    else if (arg_match(&arg, &q_hist_n, argi))
      global->show_q_hist_buckets = arg_parse_uint(&arg);
    else if (arg_match(&arg, &rate_hist_n, argi))
      global->show_rate_hist_buckets = arg_parse_uint(&arg);
    else if (arg_match(&arg, &disable_warnings, argi))
      global->disable_warnings = 1;
    else if (arg_match(&arg, &disable_warning_prompt, argi))
      global->disable_warning_prompt = 1;
    else
      argj++;
  }

  if (global->pass) {
    /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
    if (global->pass > global->passes) {
      warn("Assuming --pass=%d implies --passes=%d\n", global->pass,
           global->pass);
      global->passes = global->pass;
    }
  }
  /* Validate global config */
  if (global->passes == 0) {
#if CONFIG_AV1_ENCODER
    // Make default AV1 passes = 2 until there is a better quality 1-pass
    // encoder
    if (global->codec != NULL && global->codec->name != NULL)
      global->passes = (strcmp(global->codec->name, "av1") == 0) ? 2 : 1;
#else
    global->passes = 1;
#endif
  }
}

static void open_input_file(struct AvxInputContext *input) {
  /* Parse certain options from the input file, if possible */
  input->file = strcmp(input->filename, "-") ? fopen(input->filename, "rb")
                                             : set_binary_mode(stdin);

  if (!input->file) fatal("Failed to open input file");

  if (!fseeko(input->file, 0, SEEK_END)) {
    /* Input file is seekable. Figure out how long it is, so we can get
     * progress info.
     */
    input->length = ftello(input->file);
    rewind(input->file);
  }

  /* Default to 1:1 pixel aspect ratio. */
  input->pixel_aspect_ratio.numerator = 1;
  input->pixel_aspect_ratio.denominator = 1;

  /* For RAW input sources, these bytes will applied on the first frame
   *  in read_frame().
   */
  input->detect.buf_read = fread(input->detect.buf, 1, 4, input->file);
  input->detect.position = 0;

  if (input->detect.buf_read == 4 && file_is_y4m(input->detect.buf)) {
    if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
                       input->only_i420) >= 0) {
      input->file_type = FILE_TYPE_Y4M;
      input->width = input->y4m.pic_w;
      input->height = input->y4m.pic_h;
      input->pixel_aspect_ratio.numerator = input->y4m.par_n;
      input->pixel_aspect_ratio.denominator = input->y4m.par_d;
      input->framerate.numerator = input->y4m.fps_n;
      input->framerate.denominator = input->y4m.fps_d;
      input->fmt = input->y4m.aom_fmt;
      input->bit_depth = input->y4m.bit_depth;
    } else
      fatal("Unsupported Y4M stream.");
  } else if (input->detect.buf_read == 4 && fourcc_is_ivf(input->detect.buf)) {
    fatal("IVF is not supported as input.");
  } else {
    input->file_type = FILE_TYPE_RAW;
  }
}

static void close_input_file(struct AvxInputContext *input) {
  fclose(input->file);
  if (input->file_type == FILE_TYPE_Y4M) y4m_input_close(&input->y4m);
}

static struct stream_state *new_stream(struct AvxEncoderConfig *global,
                                       struct stream_state *prev) {
  struct stream_state *stream;

  stream = calloc(1, sizeof(*stream));
  if (stream == NULL) {
    fatal("Failed to allocate new stream.");
  }

  if (prev) {
    memcpy(stream, prev, sizeof(*stream));
    stream->index++;
    prev->next = stream;
  } else {
    aom_codec_err_t res;

    /* Populate encoder configuration */
    res = aom_codec_enc_config_default(global->codec->codec_interface(),
                                       &stream->config.cfg, global->usage);
    if (res) fatal("Failed to get config: %s\n", aom_codec_err_to_string(res));

    /* Change the default timebase to a high enough value so that the
     * encoder will always create strictly increasing timestamps.
     */
    stream->config.cfg.g_timebase.den = 1000;

    /* Never use the library's default resolution, require it be parsed
     * from the file or set on the command line.
     */
    stream->config.cfg.g_w = 0;
    stream->config.cfg.g_h = 0;

    /* Initialize remaining stream parameters */
    stream->config.write_webm = 1;
#if CONFIG_WEBM_IO
    stream->config.stereo_fmt = STEREO_FORMAT_MONO;
    stream->webm_ctx.last_pts_ns = -1;
    stream->webm_ctx.writer = NULL;
    stream->webm_ctx.segment = NULL;
#endif

    /* Allows removal of the application version from the EBML tags */
    stream->webm_ctx.debug = global->debug;
  }

  /* Output files must be specified for each stream */
  stream->config.out_fn = NULL;

  stream->next = NULL;
  return stream;
}

static int parse_stream_params(struct AvxEncoderConfig *global,
                               struct stream_state *stream, char **argv) {
  char **argi, **argj;
  struct arg arg;
  static const arg_def_t **ctrl_args = no_args;
  static const int *ctrl_args_map = NULL;
  struct stream_config *config = &stream->config;
  int eos_mark_found = 0;
#if CONFIG_HIGHBITDEPTH
  int test_16bit_internal = 0;
#endif

  // Handle codec specific options
  if (0) {
#if CONFIG_AV1_ENCODER
  } else if (strcmp(global->codec->name, "av1") == 0) {
    // TODO(jingning): Reuse AV1 specific encoder configuration parameters.
    // Consider to expand this set for AV1 encoder control.
    ctrl_args = av1_args;
    ctrl_args_map = av1_arg_ctrl_map;
#endif
  }

  for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step) {
    arg.argv_step = 1;

    /* Once we've found an end-of-stream marker (--) we want to continue
     * shifting arguments but not consuming them.
     */
    if (eos_mark_found) {
      argj++;
      continue;
    } else if (!strcmp(*argj, "--")) {
      eos_mark_found = 1;
      continue;
    }

    if (arg_match(&arg, &outputfile, argi)) {
      config->out_fn = arg.val;
    } else if (arg_match(&arg, &fpf_name, argi)) {
      config->stats_fn = arg.val;
#if CONFIG_FP_MB_STATS
    } else if (arg_match(&arg, &fpmbf_name, argi)) {
      config->fpmb_stats_fn = arg.val;
#endif
    } else if (arg_match(&arg, &use_webm, argi)) {
#if CONFIG_WEBM_IO
      config->write_webm = 1;
#else
      die("Error: --webm specified but webm is disabled.");
#endif
    } else if (arg_match(&arg, &use_ivf, argi)) {
      config->write_webm = 0;
    } else if (arg_match(&arg, &threads, argi)) {
      config->cfg.g_threads = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &profile, argi)) {
      config->cfg.g_profile = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &width, argi)) {
      config->cfg.g_w = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &height, argi)) {
      config->cfg.g_h = arg_parse_uint(&arg);
#if CONFIG_HIGHBITDEPTH
    } else if (arg_match(&arg, &bitdeptharg, argi)) {
      config->cfg.g_bit_depth = arg_parse_enum_or_int(&arg);
    } else if (arg_match(&arg, &inbitdeptharg, argi)) {
      config->cfg.g_input_bit_depth = arg_parse_uint(&arg);
#endif
#if CONFIG_WEBM_IO
    } else if (arg_match(&arg, &stereo_mode, argi)) {
      config->stereo_fmt = arg_parse_enum_or_int(&arg);
#endif
    } else if (arg_match(&arg, &timebase, argi)) {
      config->cfg.g_timebase = arg_parse_rational(&arg);
      validate_positive_rational(arg.name, &config->cfg.g_timebase);
    } else if (arg_match(&arg, &error_resilient, argi)) {
      config->cfg.g_error_resilient = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &lag_in_frames, argi)) {
      config->cfg.g_lag_in_frames = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &dropframe_thresh, argi)) {
      config->cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &resize_allowed, argi)) {
      config->cfg.rc_resize_allowed = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &resize_width, argi)) {
      config->cfg.rc_scaled_width = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &resize_height, argi)) {
      config->cfg.rc_scaled_height = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &resize_up_thresh, argi)) {
      config->cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &resize_down_thresh, argi)) {
      config->cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &end_usage, argi)) {
      config->cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
    } else if (arg_match(&arg, &target_bitrate, argi)) {
      config->cfg.rc_target_bitrate = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &min_quantizer, argi)) {
      config->cfg.rc_min_quantizer = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &max_quantizer, argi)) {
      config->cfg.rc_max_quantizer = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &undershoot_pct, argi)) {
      config->cfg.rc_undershoot_pct = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &overshoot_pct, argi)) {
      config->cfg.rc_overshoot_pct = arg_parse_uint(&arg);
    } else if (arg_match(&arg, &buf_sz, argi)) {
      config->cfg.rc_buf_sz = arg_parse_uint(&arg);