1, Add more profile limitation. 2, fix some bugs in h2m

yilun-zhangs · yilun-zhangs · commit 067a207e167d · 2024-07-04T17:34:38.000+08:00
diff --git a/code/dep_external/lib/binaural/libiamf2bear.so b/code/dep_external/lib/binaural/libiamf2bear.so
diff --git a/code/dep_external/src/binaural/iamf2bear/iamf2bear.cpp b/code/dep_external/src/binaural/iamf2bear/iamf2bear.cpp
@@ -102,7 +102,7 @@ static int getmodulepath(char *path, int buffsize)
   int count = 0, i = 0;
 #if defined(_WIN32)
   count = GetModuleFileName(NULL, path, buffsize);
-#elifdef __APPLE__
+#elif defined(__APPLE__)
   uint32_t size = MAX_PATH;
   _NSGetExecutablePath(path, &size); 
   count = size;
@@ -231,17 +231,17 @@ extern "C" EXPORT_API int ConfigureBearDirectSpeakerChannel(void *pv_thiz,
                                         std::make_pair(-180.0, 180.0),
                                         std::make_pair(-90.0, 90.0),
                                         true},
-                                Channel{"M+110",
-                                        PolarPosition{110.0, 0.0},
-                                        PolarPosition{110.0, 0.0},
-                                        std::make_pair(100.0, 120.0),
-                                        std::make_pair(0.0, 15.0),
+                                Channel{"U+030",
+                                        PolarPosition{30.0, 30.0},
+                                        PolarPosition{30.0, 30.0},
+                                        std::make_pair(30.0, 45.0),
+                                        std::make_pair(30.0, 55.0),
                                         false},
-                                Channel{"M-110",
-                                        PolarPosition{-110.0, 0.0},
-                                        PolarPosition{-110.0, 0.0},
-                                        std::make_pair(-120.0, -100.0),
-                                        std::make_pair(0.0, 15.0),
+                                Channel{"U-030",
+                                        PolarPosition{-30.0, 30.0},
+                                        PolarPosition{-30.0, 30.0},
+                                        std::make_pair(-45.0, -30.0),
+                                        std::make_pair(30.0, 55.0),
                                         false},
                             }};
 
diff --git a/code/include/IAMF_defines.h b/code/include/IAMF_defines.h
@@ -192,9 +192,4 @@ typedef enum {
   IA_CHANNEL_LAYOUT_BINAURAL,  // binaural
   IA_CHANNEL_LAYOUT_COUNT
 } IAChannelLayoutType;
-
-#define IAMF_AMBISONICS_MAX_CHANNELS 16
-#define IAMF_MIX_PRESENTATION_MAX_SUBS 1
-#define IAMF_MIX_PRESENTATION_MAX_ELEMENTS 2
-#define IAMF_MIX_PRESENTATION_MAX_CHANNELS 18
 #endif /* IAMF_DEFINES_H */
diff --git a/code/src/iamf_dec/IAMF_OBU.c b/code/src/iamf_dec/IAMF_OBU.c
@@ -247,8 +247,8 @@ uint32_t iamf_obu_get_payload_size(IAMF_OBU *obu) {
   return obu->size - (uint32_t)(obu->payload - obu->data);
 }
 
-static int _valid_profile(uint8_t profile) {
-  return profile == IAMF_PROFILE_SIMPLE || profile == IAMF_PROFILE_BASE;
+static int _valid_profile(uint8_t primary, uint8_t addional) {
+  return primary < IAMF_PROFILE_COUNT && primary <= addional;
 }
 
 IAMF_Version *iamf_version_new(IAMF_OBU *obu) {
@@ -273,9 +273,11 @@ IAMF_Version *iamf_version_new(IAMF_OBU *obu) {
       "%u.",
       (char *)&ver->iamf_code, ver->primary_profile, ver->additional_profile);
 
-  if (!_valid_profile(ver->primary_profile)) {
-    ia_loge("ia sequence header object: Invalid profile %u",
-            ver->primary_profile);
+  if (!_valid_profile(ver->primary_profile, ver->additional_profile)) {
+    ia_loge(
+        "ia sequence header object: Invalid primary profile %u or additional "
+        "profile %u.",
+        ver->primary_profile, ver->additional_profile);
     goto version_fail;
   }
 
@@ -756,11 +758,6 @@ IAMF_MixPresentation *iamf_mix_presentation_new(IAMF_OBU *obu) {
           "Mix Presentation Object: num_audio_elements should not be set to "
           "0.");
       goto mix_presentation_fail;
-    } else if (val > 2) {
-      ia_logw(
-          "Mix Presentation Object: Do not support num_audio_elements more "
-          "than 2.");
-      goto mix_presentation_fail;
     }
 
     conf_s = IAMF_MALLOCZ(ElementConf, val);
diff --git a/code/src/iamf_dec/IAMF_decoder.c b/code/src/iamf_dec/IAMF_decoder.c
@@ -94,9 +94,10 @@ static int32_t FLOAT2INT24(float x) {
 
 static int32_t FLOAT2INT32(float x) {
   x = x * 2147483648.f;
-  x = MAX(x, -2147483648.f);
-  x = MIN(x, 2147483647.f);
-  return (int32_t)lrintf(x);
+  if (x > -2147483648.f && x < 2147483647.f)
+    return (int32_t)lrintf(x);
+  else
+    return (x > 0.0f ? 2147483647 : (-2147483647 - 1));
 }
 
 static void iamf_decoder_plane2stride_out(void *dst, const float *src,
@@ -1183,6 +1184,16 @@ static int iamf_database_element_add(IAMF_DataBase *db, IAMF_Object *obj) {
   return ret;
 }
 
+static struct {
+  uint32_t max_elements;
+  uint32_t max_channels;
+} _profile_limit[IAMF_PROFILE_COUNT] = {
+    {IAMF_SIMPLE_PROFILE_MIX_PRESENTATION_MAX_ELEMENTS,
+     IAMF_SIMPLE_PROFILE_MIX_PRESENTATION_MAX_CHANNELS},
+    {IAMF_BASE_PROFILE_MIX_PRESENTATION_MAX_ELEMENTS,
+     IAMF_BASE_PROFILE_MIX_PRESENTATION_MAX_CHANNELS},
+};
+
 static int iamf_database_mix_presentation_is_valid(IAMF_DataBase *db,
                                                    IAMF_MixPresentation *mp) {
   int ret = IAMF_OK;
@@ -1193,7 +1204,7 @@ static int iamf_database_mix_presentation_is_valid(IAMF_DataBase *db,
 
   if (mp->num_sub_mixes < IAMF_MIX_PRESENTATION_MAX_SUBS) return 0;
   sub = mp->sub_mixes;
-  if (sub->nb_elements > IAMF_MIX_PRESENTATION_MAX_ELEMENTS) return 0;
+  if (sub->nb_elements > _profile_limit[db->profile].max_elements) return 0;
 
   for (int e = 0; e < sub->nb_elements; ++e) {
     econf = &sub->conf_s[e];
@@ -1222,10 +1233,10 @@ static int iamf_database_mix_presentation_is_valid(IAMF_DataBase *db,
   }
 
   if (ret != IAMF_OK) return !ret;
-  if (channels > IAMF_MIX_PRESENTATION_MAX_CHANNELS) {
+  if (channels > _profile_limit[db->profile].max_channels) {
     ia_logw("Mix Presentation %" PRId64 " has %d channels, more than %u",
             mp->mix_presentation_id, channels,
-            IAMF_MIX_PRESENTATION_MAX_CHANNELS);
+            _profile_limit[db->profile].max_channels);
     return 0;
   }
 
@@ -1261,6 +1272,7 @@ int iamf_database_init(IAMF_DataBase *db) {
   db->mixPresentation = iamf_object_set_new(iamf_object_free);
   db->eViewer.freeF = free;
   db->pViewer.freeF = iamf_parameter_item_free;
+  db->profile = IAMF_PROFILE_DEFAULT;
 
   if (!db->codecConf || !db->element || !db->mixPresentation) {
     iamf_database_reset(db);
@@ -1288,13 +1300,27 @@ static int iamf_database_add_object(IAMF_DataBase *db, IAMF_Object *obj) {
   if (!obj) return IAMF_ERR_BAD_ARG;
 
   switch (obj->type) {
-    case IAMF_OBU_SEQUENCE_HEADER:
+    case IAMF_OBU_SEQUENCE_HEADER: {
+      IAMF_Version *version = (IAMF_Version *)obj;
+
+      if (version->primary_profile > db->profile) {
+        ia_loge("Unimplemented profile %u", version->primary_profile);
+        free(obj);
+        ret = IAMF_ERR_UNIMPLEMENTED;
+        break;
+      }
+
+      if (version->additional_profile < db->profile)
+        db->profile = version->additional_profile;
+
       if (db->version) {
         ia_logw("WARNING : Receive Multiple START CODE OBUs !!!");
         free(db->version);
       }
+
       db->version = obj;
       break;
+    }
     case IAMF_OBU_CODEC_CONFIG:
       ret = iamf_object_set_add(db->codecConf, (void *)obj);
       break;
diff --git a/code/src/iamf_dec/IAMF_decoder_private.h b/code/src/iamf_dec/IAMF_decoder_private.h
@@ -136,6 +136,8 @@ typedef struct IAMF_DataBase {
 
   Viewer eViewer;
   Viewer pViewer;
+
+  IAMF_Profile profile;
 } IAMF_DataBase;
 
 /* <<<<<<<<<<<<<<<<<< DATABASE <<<<<<<<<<<<<<<<<< */
diff --git a/code/src/iamf_dec/IAMF_types.h b/code/src/iamf_dec/IAMF_types.h
@@ -92,7 +92,12 @@ typedef struct MixFactors {
   int w_idx_offset;
 } MixFactors;
 
-typedef enum { IAMF_PROFILE_SIMPLE, IAMF_PROFILE_BASE } IAMF_Profile;
+typedef enum {
+  IAMF_PROFILE_SIMPLE,
+  IAMF_PROFILE_BASE,
+  IAMF_PROFILE_DEFAULT = IAMF_PROFILE_BASE,
+  IAMF_PROFILE_COUNT,
+} IAMF_Profile;
 
 #define U8_MASK 0xFF
 #define U16_MASK 0xFFFF
@@ -112,4 +117,15 @@ typedef enum { IAMF_PROFILE_SIMPLE, IAMF_PROFILE_BASE } IAMF_Profile;
 #define IA_CH_CATE_WEIGHT 0X200
 #define IA_CH_CATE_TOP 0X400
 
+#define IAMF_AMBISONICS_MAX_CHANNELS 16
+#define IAMF_MIX_PRESENTATION_MAX_SUBS 1
+/// @brief should be able to reconstruct one Audio Element. in simple profile.
+#define IAMF_SIMPLE_PROFILE_MIX_PRESENTATION_MAX_ELEMENTS 1
+/// @brief should be able to handle up to 16 channels in simple profile.
+#define IAMF_SIMPLE_PROFILE_MIX_PRESENTATION_MAX_CHANNELS 16
+/// @brief should be able to mix two Audio Elements in base profile.
+#define IAMF_BASE_PROFILE_MIX_PRESENTATION_MAX_ELEMENTS 2
+/// @brief should be able to handle up to 18 channels in base profile.
+#define IAMF_BASE_PROFILE_MIX_PRESENTATION_MAX_CHANNELS 18
+
 #endif /* IAMF_TYPES_H_ */
diff --git a/code/src/iamf_dec/h2m_rdr.c b/code/src/iamf_dec/h2m_rdr.c
@@ -995,9 +995,9 @@ struct h2m_rdr_t h2m_rdr_tab[] = {
     {IAMF_ZOA, BS2051_C, 8, 3, -1, (float *)zoa_bs250, 1, 7},
     {IAMF_ZOA, BS2051_D, 10, 3, -1, (float *)zoa_bs450, 1, 9},
     {IAMF_ZOA, BS2051_E, 11, 3, -1, (float *)zoa_bs451, 1, 10},
-    {IAMF_ZOA, BS2051_F, 11, 10, 11, (float *)zoa_bs370, 1, 10},
-    {IAMF_ZOA, BS2051_G, 12, 3, -1, (float *)zoa_bs490, 1, 13},
-    {IAMF_ZOA, BS2051_H, 24, 3, -1, (float *)zoa_bs9A3, 1, 22},
+    {IAMF_ZOA, BS2051_F, 12, 10, 11, (float *)zoa_bs370, 1, 10},
+    {IAMF_ZOA, BS2051_G, 14, 3, -1, (float *)zoa_bs490, 1, 13},
+    {IAMF_ZOA, BS2051_H, 24, 3, 9, (float *)zoa_bs9A3, 1, 22},
     {IAMF_ZOA, BS2051_I, 8, 3, -1, (float *)zoa_bs070, 1, 7},
     {IAMF_ZOA, BS2051_J, 12, 3, -1, (float *)zoa_bs470, 1, 11},
     {IAMF_ZOA, IAMF_312, 6, 3, -1, (float *)zoa_iamf312, 1, 5},
@@ -1010,9 +1010,9 @@ struct h2m_rdr_t h2m_rdr_tab[] = {
     {IAMF_FOA, BS2051_C, 8, 3, -1, (float *)foa_bs250, 4, 7},
     {IAMF_FOA, BS2051_D, 10, 3, -1, (float *)foa_bs450, 4, 9},
     {IAMF_FOA, BS2051_E, 11, 3, -1, (float *)foa_bs451, 4, 10},
-    {IAMF_FOA, BS2051_F, 11, 10, 11, (float *)foa_bs370, 4, 10},
-    {IAMF_FOA, BS2051_G, 12, 3, -1, (float *)foa_bs490, 4, 13},
-    {IAMF_FOA, BS2051_H, 24, 3, -1, (float *)foa_bs9A3, 4, 22},
+    {IAMF_FOA, BS2051_F, 12, 10, 11, (float *)foa_bs370, 4, 10},
+    {IAMF_FOA, BS2051_G, 14, 3, -1, (float *)foa_bs490, 4, 13},
+    {IAMF_FOA, BS2051_H, 24, 3, 9, (float *)foa_bs9A3, 4, 22},
     {IAMF_FOA, BS2051_I, 8, 3, -1, (float *)foa_bs070, 4, 7},
     {IAMF_FOA, BS2051_J, 12, 3, -1, (float *)foa_bs470, 4, 11},
     {IAMF_FOA, IAMF_312, 6, 3, -1, (float *)foa_iamf312, 4, 5},
@@ -1025,9 +1025,9 @@ struct h2m_rdr_t h2m_rdr_tab[] = {
     {IAMF_SOA, BS2051_C, 8, 3, -1, (float *)soa_bs250, 9, 7},
     {IAMF_SOA, BS2051_D, 10, 3, -1, (float *)soa_bs450, 9, 9},
     {IAMF_SOA, BS2051_E, 11, 3, -1, (float *)soa_bs451, 9, 10},
-    {IAMF_SOA, BS2051_F, 11, 10, 11, (float *)soa_bs370, 9, 10},
-    {IAMF_SOA, BS2051_G, 12, 3, -1, (float *)soa_bs490, 9, 13},
-    {IAMF_SOA, BS2051_H, 24, 3, -1, (float *)soa_bs9A3, 9, 22},
+    {IAMF_SOA, BS2051_F, 12, 10, 11, (float *)soa_bs370, 9, 10},
+    {IAMF_SOA, BS2051_G, 14, 3, -1, (float *)soa_bs490, 9, 13},
+    {IAMF_SOA, BS2051_H, 24, 3, 9, (float *)soa_bs9A3, 9, 22},
     {IAMF_SOA, BS2051_I, 8, 3, -1, (float *)soa_bs070, 9, 7},
     {IAMF_SOA, BS2051_J, 12, 3, -1, (float *)soa_bs470, 9, 11},
     {IAMF_SOA, IAMF_312, 6, 3, -1, (float *)soa_iamf312, 9, 5},
@@ -1040,9 +1040,9 @@ struct h2m_rdr_t h2m_rdr_tab[] = {
     {IAMF_TOA, BS2051_C, 8, 3, -1, (float *)toa_bs250, 16, 7},
     {IAMF_TOA, BS2051_D, 10, 3, -1, (float *)toa_bs450, 16, 9},
     {IAMF_TOA, BS2051_E, 11, 3, -1, (float *)toa_bs451, 16, 10},
-    {IAMF_TOA, BS2051_F, 11, 10, 11, (float *)toa_bs370, 16, 10},
-    {IAMF_TOA, BS2051_G, 12, 3, -1, (float *)toa_bs490, 16, 13},
-    {IAMF_TOA, BS2051_H, 24, 3, -1, (float *)toa_bs9A3, 16, 22},
+    {IAMF_TOA, BS2051_F, 12, 10, 11, (float *)toa_bs370, 16, 10},
+    {IAMF_TOA, BS2051_G, 14, 3, -1, (float *)toa_bs490, 16, 13},
+    {IAMF_TOA, BS2051_H, 24, 3, 9, (float *)toa_bs9A3, 16, 22},
     {IAMF_TOA, BS2051_I, 8, 3, -1, (float *)toa_bs070, 16, 7},
     {IAMF_TOA, BS2051_J, 12, 3, -1, (float *)toa_bs470, 16, 11},
     {IAMF_TOA, IAMF_312, 6, 3, -1, (float *)toa_iamf312, 16, 5},
