[VCWG-G144/G133] bitstream conformance for bitstreams contains SWICTH frames

apps/avmenc.c

@@ -220,6 +220,8 @@ const arg_def_t *kf_args[] = { &g_av2_codec_arg_defs.fwd_kf_enabled,
                                &g_av2_codec_arg_defs.sframe_dist,
                                &g_av2_codec_arg_defs.sframe_mode,
                                &g_av2_codec_arg_defs.sframe_type,
+                               &g_av2_codec_arg_defs.sframe_refresh_all,
+                               &g_av2_codec_arg_defs.sframe_conformance_option,
                                NULL };
 
 // TODO(bohanli): Currently all options are supported by the key & value API.
@@ -1185,6 +1187,14 @@ static int parse_stream_params(struct AvxEncoderConfig *global,
       config->cfg.sframe_mode = arg_parse_uint(&arg);
     } else if (arg_match(&arg, &g_av2_codec_arg_defs.sframe_type, argi)) {
       config->cfg.sframe_type = arg_parse_uint(&arg);
+    } else if (arg_match(&arg, &g_av2_codec_arg_defs.sframe_refresh_all,
+                         argi)) {
+      config->cfg.sframe_refresh_all = arg_parse_uint(&arg);
+    } else if (arg_match(&arg, &g_av2_codec_arg_defs.sframe_replace_kf, argi)) {
+      config->cfg.sframe_replace_kf = arg_parse_uint(&arg);
+    } else if (arg_match(&arg, &g_av2_codec_arg_defs.sframe_conformance_option,
+                         argi)) {
+      config->cfg.sframe_conformance_option = arg_parse_uint(&arg);
     } else if (arg_match(&arg, &g_av2_codec_arg_defs.enable_lcr, argi)) {
       config->cfg.enable_lcr = arg_parse_uint(&arg);
     } else if (arg_match(&arg, &g_av2_codec_arg_defs.enable_ops, argi)) {

av2/arg_defs.c

@@ -290,6 +290,18 @@ const av2_codec_arg_definitions_t g_av2_codec_arg_defs = {
       ARG_DEF(NULL, "sframe-mode", 1, "S-Frame insertion mode (0..2)"),
   .sframe_type = ARG_DEF(NULL, "sframe-type", 1,
                          "(0: Regular S-Frame (default), 1: RAS frame)"),
+  .sframe_refresh_all =
+      ARG_DEF(NULL, "sframe-refresh-all", 1,
+              "S-Frame refreshes all ref slots (0: off (default), 1: on)"),
+  .sframe_replace_kf =
+      ARG_DEF(NULL, "sframe-replace-kf", 1,
+              "Replace N consecutive keyframes with S-frames (0: off)"),
+  .sframe_conformance_option = ARG_DEF(
+      NULL, "sframe-conformance-option", 1,
+      "How to satisfy the BAWP/IBC/OPFL/RefineMV bitstream-conformance "
+      "requirement around switch frames (0: disable refmvbank (default), "
+      "1: keep refmvbank, disable BAWP/IntraBC after switch, "
+      "2: disable OPFL refinement and RefineMV)"),
   .enable_lcr =
       ARG_DEF(NULL, "enable-lcr", 1,
               "Enable layer config record (LCR) OBU (0: off (default), 1: on)"),

av2/arg_defs.h

@@ -109,6 +109,9 @@ typedef struct av2_codec_arg_definitions {
   arg_def_t sframe_dist;
   arg_def_t sframe_mode;
   arg_def_t sframe_type;
+  arg_def_t sframe_refresh_all;
+  arg_def_t sframe_replace_kf;
+  arg_def_t sframe_conformance_option;
   arg_def_t enable_lcr;
   arg_def_t enable_ops;
   arg_def_t num_ops;

av2/av2_cx_iface.c

@@ -1411,6 +1411,13 @@ static avm_codec_err_t set_encoder_config(AV2EncoderConfig *oxcf,
 
   tool_cfg->enable_joint_mvd = extra_cfg->enable_joint_mvd;
   tool_cfg->enable_refinemv = extra_cfg->enable_refinemv;
+  if (cfg->enable_sframe && cfg->sframe_conformance_option == 2) {
+    // Option 3: disable RefineMV sequence-wide, applied at the point
+    // tool_cfg->enable_refinemv is derived (matching the enable_opfl_refine
+    // override below).
+    tool_cfg->enable_refinemv = 0;
+    extra_cfg->enable_refinemv = 0;
+  }
   tool_cfg->enable_mvd_sign_derive = extra_cfg->enable_mvd_sign_derive;
   // Turn off BRU if LA, AI or resize mode
   tool_cfg->enable_bru = extra_cfg->enable_bru;
@@ -1463,6 +1470,24 @@ static avm_codec_err_t set_encoder_config(AV2EncoderConfig *oxcf,
   tool_cfg->max_drl_refmvs = extra_cfg->max_drl_refmvs;
   tool_cfg->max_drl_refbvs = extra_cfg->max_drl_refbvs;
   tool_cfg->enable_refmvbank = extra_cfg->enable_refmvbank;
+  if (cfg->enable_sframe) {
+    switch (cfg->sframe_conformance_option) {
+      case 1:
+        // Option 2: keep the reference-MV bank on sequence-wide.
+        tool_cfg->enable_refmvbank = 1;
+        extra_cfg->enable_refmvbank = 1;
+        break;
+      case 2:
+        // Option 3 doesn't touch enable_refmvbank.
+        break;
+      case 0:
+      default:
+        // Option 1 (default): disable the reference-MV bank sequence-wide.
+        tool_cfg->enable_refmvbank = 0;
+        extra_cfg->enable_refmvbank = 0;
+        break;
+    }
+  }
   tool_cfg->enable_cropping_window = extra_cfg->enable_cropping_window;
   tool_cfg->crop_win_left_offset = extra_cfg->crop_win_left_offset;
   tool_cfg->crop_win_right_offset = extra_cfg->crop_win_right_offset;
@@ -1505,6 +1530,13 @@ static avm_codec_err_t set_encoder_config(AV2EncoderConfig *oxcf,
   tool_cfg->enable_high_motion = extra_cfg->enable_high_motion;
 
   tool_cfg->enable_opfl_refine = extra_cfg->enable_opfl_refine;
+  if (cfg->enable_sframe && cfg->sframe_conformance_option == 2) {
+    // Option 3: disable OPFL refinement sequence-wide, applied at the point
+    // tool_cfg->enable_opfl_refine is derived (matching the enable_refinemv
+    // override above).
+    tool_cfg->enable_opfl_refine = 0;
+    extra_cfg->enable_opfl_refine = 0;
+  }
   if (tool_cfg->enable_opfl_refine) {
     if (cfg->g_lag_in_frames == 0) {
       tool_cfg->enable_opfl_refine = 0;
@@ -1594,6 +1626,9 @@ static avm_codec_err_t set_encoder_config(AV2EncoderConfig *oxcf,
   kf_cfg->sframe_dist = cfg->sframe_dist;
   kf_cfg->sframe_mode = cfg->sframe_mode;
   kf_cfg->sframe_type = cfg->sframe_type;
+  kf_cfg->sframe_refresh_all = cfg->sframe_refresh_all;
+  kf_cfg->sframe_replace_kf = cfg->sframe_replace_kf;
+  kf_cfg->sframe_conformance_option = cfg->sframe_conformance_option;
   oxcf->unit_test_cfg.insert_sframe = extra_cfg->enable_sframe;
 
   kf_cfg->enable_keyframe_filtering = extra_cfg->enable_keyframe_filtering;
@@ -2528,6 +2563,18 @@ static avm_codec_err_t ctrl_set_enable_sframe(avm_codec_alg_priv_t *ctx,
   return update_extra_cfg(ctx, &extra_cfg);
 }
 
+static avm_codec_err_t ctrl_set_sframe_conformance_option(
+    avm_codec_alg_priv_t *ctx, va_list args) {
+  struct av2_extracfg extra_cfg = ctx->extra_cfg;
+  // sframe_conformance_option lives in avm_codec_enc_cfg_t, not in
+  // av2_extracfg, so set it directly on ctx->cfg; set_encoder_config()
+  // reads it from there into kf_cfg. update_extra_cfg() still runs to
+  // re-validate and re-apply the full configuration.
+  ctx->cfg.sframe_conformance_option =
+      CAST(AV2E_SET_SFRAME_CONFORMANCE_OPTION, args);
+  return update_extra_cfg(ctx, &extra_cfg);
+}
+
 static avm_codec_err_t ctrl_set_frame_parallel_decoding_mode(
     avm_codec_alg_priv_t *ctx, va_list args) {
   struct av2_extracfg extra_cfg = ctx->extra_cfg;
@@ -4734,6 +4781,7 @@ static avm_codec_ctrl_fn_map_t encoder_ctrl_maps[] = {
   { AV2E_SET_FRAME_PARALLEL_DECODING, ctrl_set_frame_parallel_decoding_mode },
   { AV2E_SET_ENABLE_CDF_AVERAGING, ctrl_set_enable_cdf_averaging },
   { AV2E_SET_ENABLE_SFRAME, ctrl_set_enable_sframe },
+  { AV2E_SET_SFRAME_CONFORMANCE_OPTION, ctrl_set_sframe_conformance_option },
   { AV2E_SET_ENABLE_RECT_PARTITIONS, ctrl_set_enable_rect_partitions },
   { AV2E_SET_ENABLE_1TO4_PARTITIONS, ctrl_set_enable_uneven_4way_partitions },
   { AV2E_SET_MIN_PARTITION_SIZE, ctrl_set_min_partition_size },
@@ -4883,6 +4931,9 @@ static const avm_codec_enc_cfg_t encoder_usage_cfg[] = {
       0,                           // sframe_dist
       1,                           // sframe_mode
       0,                           // sframe_type
+      0,                           // sframe_refresh_all
+      0,                           // sframe_replace_kf
+      0,                           // sframe_conformance_option
       0,                           // monochrome
       0,                           // full_still_picture_hdr
       1,                           // enable_tcq
@@ -5019,6 +5070,9 @@ static const avm_codec_enc_cfg_t encoder_usage_cfg[] = {
       0,                           // sframe_dist
       1,                           // sframe_mode
       0,                           // sframe_type
+      0,                           // sframe_refresh_all
+      0,                           // sframe_replace_kf
+      0,                           // sframe_conformance_option
       0,                           // monochrome
       0,                           // full_still_picture_hdr
       0,                           // enable_tcq

av2/common/av2_common_int.h

@@ -3026,6 +3026,25 @@ typedef struct AV2Common {
    */
   bool restricted_prediction_switch;
 
+  /*!
+   * Set to 1 for a frame that is within the "risk window" opened by a coded
+   * frame with obu_type equal to OBU_SWITCH (the switch frame itself, and
+   * every following frame up to the next random access point). Used to
+   * enforce the bitstream conformance requirement constraining BAWP/IBC
+   * around switch frames.
+   */
+  bool in_switch_risk_window;
+  /*!
+   * The mlayer_id of the OBU_SWITCH frame that opened the current
+   * in_switch_risk_window. Only meaningful when in_switch_risk_window is 1.
+   */
+  int switch_risk_window_mlayer_id;
+  /*!
+   * The tlayer_id of the OBU_SWITCH frame that opened the current
+   * in_switch_risk_window. Only meaningful when in_switch_risk_window is 1.
+   */
+  int switch_risk_window_tlayer_id;
+
 } AV2_COMMON;
 
 /*!\cond */

av2/decoder/decodeframe.c

@@ -6778,6 +6778,31 @@ static INLINE void read_intrabc_params(AV2_COMMON *const cm,
   FeatureFlags *const features = &cm->features;
   features->allow_intrabc = avm_rb_read_bit(rb);
 
+  // Bitstream conformance: for a coded video sequence containing an
+  // OBU_SWITCH frame, at least one of the BAWP/IBC/OPFL/RefineMV
+  // constraints must hold for the switch frame and every dependent inter
+  // frame that follows it (the switch frame itself, frame_type == S_FRAME,
+  // is excluded from this particular constraint). When enable_refmvbank is
+  // 1 and OPFL refinement/RefineMV are not both off sequence-wide,
+  // allow_intrabc must be 0 here.
+  if (features->allow_intrabc && cm->in_switch_risk_window &&
+      current_frame->frame_type == INTER_FRAME &&
+      cm->seq_params.enable_refmvbank &&
+      !(cm->seq_params.enable_opfl_refine == AVM_OPFL_REFINE_NONE &&
+        cm->seq_params.enable_refinemv == 0) &&
+      is_mlayer_transitively_dependent(&cm->seq_params, cm->mlayer_id,
+                                       cm->switch_risk_window_mlayer_id) &&
+      (cm->tlayer_id == cm->switch_risk_window_tlayer_id ||
+       cm->seq_params
+           .tlayer_dependency_map[cm->mlayer_id][cm->tlayer_id]
+                                 [cm->switch_risk_window_tlayer_id])) {
+    avm_internal_error(
+        &cm->error, AVM_CODEC_UNSUP_BITSTREAM,
+        "allow_intrabc must be 0 for an inter frame depending on the layer "
+        "of a preceding OBU_SWITCH frame when enable_refmvbank is 1, unless "
+        "enable_opfl_refine is REFINE_NONE and enable_refinemv is 0");
+  }
+
   if (features->allow_intrabc) {
     if (current_frame->frame_type == KEY_FRAME ||
         current_frame->frame_type == INTRA_ONLY_FRAME) {
@@ -8074,6 +8099,22 @@ static int read_uncompressed_header(AV2Decoder *pbi, OBU_TYPE obu_type,
       current_frame->frame_type =
           avm_rb_read_bit(rb) ? INTER_FRAME : INTRA_ONLY_FRAME;
     }
+
+    // Track the "risk window" opened by an OBU_SWITCH frame, used below to
+    // enforce the bitstream conformance requirement constraining BAWP/IBC
+    // around switch frames (decoder state derived from reconstructed
+    // samples can diverge across bitstreams spliced at the switch frame
+    // even though parsing stays deterministic). A CLK/OLK or an
+    // OBU_RAS_FRAME is a random access point, which closes any risk window
+    // opened by a prior OBU_SWITCH frame.
+    if (current_frame->frame_type == KEY_FRAME || obu_type == OBU_RAS_FRAME) {
+      cm->in_switch_risk_window = false;
+    } else if (obu_type == OBU_SWITCH) {
+      cm->in_switch_risk_window = true;
+      cm->switch_risk_window_mlayer_id = cm->mlayer_id;
+      cm->switch_risk_window_tlayer_id = cm->tlayer_id;
+    }
+
     current_frame->long_term_id = -1;
     if (current_frame->frame_type == KEY_FRAME) {
       const int long_term_id_plus_1 =
@@ -9303,6 +9344,28 @@ static int read_uncompressed_header(AV2Decoder *pbi, OBU_TYPE obu_type,
   else
     features->enable_bawp = 0;
 
+  // Bitstream conformance: for a coded video sequence containing an
+  // OBU_SWITCH frame, at least one of the BAWP/IBC/OPFL/RefineMV
+  // constraints must hold for the switch frame and every dependent inter
+  // frame that follows it (the switch frame itself, frame_type == S_FRAME,
+  // is excluded from this particular constraint). If OPFL refinement and
+  // RefineMV are not both off sequence-wide, then BAWP must be off here.
+  if (features->enable_bawp && cm->in_switch_risk_window &&
+      current_frame->frame_type == INTER_FRAME &&
+      !(seq_params->enable_opfl_refine == AVM_OPFL_REFINE_NONE &&
+        seq_params->enable_refinemv == 0) &&
+      is_mlayer_transitively_dependent(seq_params, cm->mlayer_id,
+                                       cm->switch_risk_window_mlayer_id) &&
+      (cm->tlayer_id == cm->switch_risk_window_tlayer_id ||
+       seq_params->tlayer_dependency_map[cm->mlayer_id][cm->tlayer_id]
+                                        [cm->switch_risk_window_tlayer_id])) {
+    avm_internal_error(
+        &cm->error, AVM_CODEC_UNSUP_BITSTREAM,
+        "enable_bawp must be 0 for an inter frame depending on the layer of "
+        "a preceding OBU_SWITCH frame, unless enable_opfl_refine is "
+        "REFINE_NONE and enable_refinemv is 0");
+  }
+
   features->enable_intra_bawp = seq_params->enable_bawp;
 
   features->enable_cwp = seq_params->enable_cwp;

av2/encoder/encode_strategy.c

@@ -762,7 +762,18 @@ int av2_get_refresh_frame_flags(
     return (1 << cpi->common.seq_params.ref_frames) - 1;
   }
 
-  if (frame_params->frame_type == S_FRAME ||
+  // sframe_refresh_all only gates real, restricted OBU_SWITCH frames
+  // (restricted_prediction_switch, i.e. sframe_mode == 0). RAS frames
+  // (sframe_type == RAS_FRAME) and non-restricted switch frames always
+  // refresh all RPL slots: RAS frames are random access points, and a
+  // non-restricted switch frame retains no is_restricted old references to
+  // make partial refresh meaningful. cpi->is_ras_frame is not used here
+  // because it is only valid for the frame currently being encoded, not
+  // reliably for every call site reachable from this function.
+  if ((frame_params->frame_type == S_FRAME &&
+       (cpi->oxcf.kf_cfg.sframe_refresh_all ||
+        cpi->oxcf.kf_cfg.sframe_type == RAS_FRAME ||
+        !cpi->common.restricted_prediction_switch)) ||
       frame_params->frame_type == KEY_FRAME) {
     AV2_COMMON *const cm = &cpi->common;
     int refresh_frame_flags = (1 << cpi->common.seq_params.ref_frames) - 1;
@@ -794,6 +805,7 @@ int av2_get_refresh_frame_flags(
           // been output yet and whose DOH is at least the current
           // frame's DOH. (DOH requirement)
           if (cm->ref_frame_map[i] != NULL &&
+              !cm->ref_frame_map[i]->is_restricted &&
               cm->ref_frame_map[i]->implicit_output_picture &&
               !cm->ref_frame_map[i]->frame_output_done &&
               (int)cm->ref_frame_map[i]->display_order_hint >= cur_disp_order) {
@@ -847,7 +859,7 @@ int av2_get_refresh_frame_flags(
   // at least the current frame's DOH. (DOH requirement)
   for (int i = 0; i < cpi->common.seq_params.ref_frames; i++) {
     const RefCntBuffer *const buf = cpi->common.ref_frame_map[i];
-    if (buf != NULL && buf->implicit_output_picture &&
+    if (buf != NULL && !buf->is_restricted && buf->implicit_output_picture &&
         !buf->frame_output_done &&
         (int)buf->display_order_hint >= cur_disp_order) {
       olk_flags_to_keep |= (1 << i);
@@ -1278,6 +1290,8 @@ int av2_encode_strategy(AV2_COMP *const cpi, size_t *const size,
 
   cm->restricted_prediction_switch =
       (cpi->oxcf.kf_cfg.enable_sframe && cpi->oxcf.kf_cfg.sframe_mode == 0) ||
+      (cpi->oxcf.kf_cfg.sframe_replace_kf > 0 &&
+       cpi->oxcf.kf_cfg.sframe_mode == 0) ||
       cpi->oxcf.tool_cfg.g_error_resilient_mode;
 
   av2_configure_buffer_updates(cpi, frame_update_type);