applemedia: Add HEVC support to vtenc and vtdec

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2754>

subprojects/gst-plugins-bad/sys/applemedia/vtdec.c

@@ -105,6 +105,8 @@ static void gst_vtdec_session_output_callback (void
     CMTime duration);
 static gboolean compute_h264_decode_picture_buffer_length (GstVtdec * vtdec,
     GstBuffer * codec_data, int *length);
+static gboolean compute_hevc_decode_picture_buffer_length (GstVtdec * vtdec,
+    GstBuffer * codec_data, int *length);
 static gboolean gst_vtdec_compute_reorder_queue_length (GstVtdec * vtdec,
     CMVideoCodecType cm_format, GstBuffer * codec_data);
 static void gst_vtdec_set_latency (GstVtdec * vtdec);
@@ -115,6 +117,8 @@ static GstStaticPadTemplate gst_vtdec_sink_template =
     GST_PAD_SINK,
     GST_PAD_ALWAYS,
     GST_STATIC_CAPS ("video/x-h264, stream-format=avc, alignment=au,"
+        " width=(int)[1, MAX], height=(int)[1, MAX];"
+        "video/x-h265, stream-format=(string){ hev1, hvc1 }, alignment=au,"
         " width=(int)[1, MAX], height=(int)[1, MAX];"
         "video/mpeg, mpegversion=2, systemstream=false, parsed=true;"
         "image/jpeg;"
@@ -535,6 +539,8 @@ gst_vtdec_set_format (GstVideoDecoder * decoder, GstVideoCodecState * state)
   caps_name = gst_structure_get_name (structure);
   if (!strcmp (caps_name, "video/x-h264")) {
     cm_format = kCMVideoCodecType_H264;
+  } else if (!strcmp (caps_name, "video/x-h265")) {
+    cm_format = kCMVideoCodecType_HEVC;
   } else if (!strcmp (caps_name, "video/mpeg")) {
     cm_format = kCMVideoCodecType_MPEG2Video;
   } else if (!strcmp (caps_name, "image/jpeg")) {
@@ -551,7 +557,9 @@ gst_vtdec_set_format (GstVideoDecoder * decoder, GstVideoCodecState * state)
     }
   }
 
-  if (cm_format == kCMVideoCodecType_H264 && state->codec_data == NULL) {
+  if ((cm_format == kCMVideoCodecType_H264
+          || cm_format == kCMVideoCodecType_HEVC)
+      && state->codec_data == NULL) {
     GST_INFO_OBJECT (vtdec, "no codec data, wait for one");
     return TRUE;
   }
@@ -788,7 +796,12 @@ create_format_description_from_codec_data (GstVtdec * vtdec,
   gst_buffer_map (codec_data, &map, GST_MAP_READ);
   atoms = CFDictionaryCreateMutable (NULL, 0, &kCFTypeDictionaryKeyCallBacks,
       &kCFTypeDictionaryValueCallBacks);
+
+  if (cm_format == kCMVideoCodecType_HEVC)
+    gst_vtutil_dict_set_data (atoms, CFSTR ("hvcC"), map.data, map.size);
+  else
     gst_vtutil_dict_set_data (atoms, CFSTR ("avcC"), map.data, map.size);
+
   gst_vtutil_dict_set_object (extensions,
       CFSTR ("SampleDescriptionExtensionAtoms"), (CFTypeRef *) atoms);
   gst_buffer_unmap (codec_data, &map);
@@ -1148,6 +1161,11 @@ gst_vtdec_compute_reorder_queue_length (GstVtdec * vtdec,
             &vtdec->reorder_queue_length)) {
       return FALSE;
     }
+  } else if (cm_format == kCMVideoCodecType_HEVC) {
+    if (!compute_hevc_decode_picture_buffer_length (vtdec, codec_data,
+            &vtdec->reorder_queue_length)) {
+      return FALSE;
+    }
   } else {
     vtdec->reorder_queue_length = 0;
   }
@@ -1195,6 +1213,34 @@ compute_h264_decode_picture_buffer_length (GstVtdec * vtdec,
   return TRUE;
 }
 
+static gboolean
+compute_hevc_decode_picture_buffer_length (GstVtdec * vtdec,
+    GstBuffer * codec_data, int *length)
+{
+  /* This value should be level dependent (table A.8)
+   * but let's assume the maximum possible one for simplicity. */
+  const gint max_luma_ps = 35651584;
+  const gint max_dpb_pic_buf = 6;
+  gint max_dbp_size, pic_size_samples_y;
+
+  if (vtdec->video_info.width == 0 || vtdec->video_info.height == 0)
+    return FALSE;
+
+  /* A.4.2 */
+  pic_size_samples_y = vtdec->video_info.width * vtdec->video_info.height;
+  if (pic_size_samples_y <= (max_luma_ps >> 2))
+    max_dbp_size = max_dpb_pic_buf * 4;
+  else if (pic_size_samples_y <= (max_luma_ps >> 1))
+    max_dbp_size = max_dpb_pic_buf * 2;
+  else if (pic_size_samples_y <= ((3 * max_luma_ps) >> 2))
+    max_dbp_size = (max_dpb_pic_buf * 4) / 3;
+  else
+    max_dbp_size = max_dpb_pic_buf;
+
+  *length = MIN (max_dbp_size, 16);
+  return TRUE;
+}
+
 static void
 gst_vtdec_set_latency (GstVtdec * vtdec)
 {

subprojects/gst-plugins-bad/sys/applemedia/vtenc.c

@@ -71,7 +71,6 @@
 #include "vtutil.h"
 #include <gst/pbutils/codec-utils.h>
 
-#define VTENC_DEFAULT_USAGE       6     /* Profile: Baseline  Level: 2.1 */
 #define VTENC_DEFAULT_BITRATE     0
 #define VTENC_DEFAULT_FRAME_REORDERING TRUE
 #define VTENC_DEFAULT_REALTIME FALSE
@@ -239,7 +238,6 @@ gst_vtenc_base_init (GstVTEncClass * klass)
         gst_pad_template_new ("sink", GST_PAD_SINK, GST_PAD_ALWAYS, caps));
   }
 
-
   src_caps = gst_caps_new_simple (codec_details->mimetype,
       "width", GST_TYPE_INT_RANGE, min_width, max_width,
       "height", GST_TYPE_INT_RANGE, min_height, max_height,
@@ -268,6 +266,11 @@ gst_vtenc_base_init (GstVTEncClass * klass)
           "stream-format", G_TYPE_STRING, "avc",
           "alignment", G_TYPE_STRING, "au", NULL);
       break;
+    case kCMVideoCodecType_HEVC:
+      gst_structure_set (gst_caps_get_structure (src_caps, 0),
+          "stream-format", G_TYPE_STRING, "hvc1",
+          "alignment", G_TYPE_STRING, "au", NULL);
+      break;
     case GST_kCMVideoCodecType_Some_AppleProRes:
       if (g_strcmp0 (codec_details->mimetype, "video/x-prores") == 0) {
         G_GNUC_BEGIN_IGNORE_DEPRECATIONS;
@@ -720,7 +723,7 @@ gst_vtenc_stop (GstVideoEncoder * enc)
 }
 
 static CFStringRef
-gst_vtenc_profile_level_key (GstVTEnc * self, const gchar * profile,
+gst_vtenc_h264_profile_level_key (GstVTEnc * self, const gchar * profile,
     const gchar * level_arg)
 {
   char level[64];
@@ -763,6 +766,37 @@ gst_vtenc_profile_level_key (GstVTEnc * self, const gchar * profile,
   return ret;
 }
 
+static CFStringRef
+gst_vtenc_hevc_profile_level_key (GstVTEnc * self, const gchar * profile,
+    const gchar * level_arg)
+{
+  gchar *key = NULL;
+  CFStringRef ret = NULL;
+
+  if (profile == NULL || !strcmp (profile, "main"))
+    profile = "Main";
+  else if (!strcmp (profile, "main-10"))
+    profile = "Main10";
+  else if (!strcmp (profile, "main-422-10"))
+    /* TODO: this should probably be guarded with a version check (macOS 12.3+ / iOS 15.4+)
+     * https://developer.apple.com/documentation/videotoolbox/kvtprofilelevel_hevc_main10_autolevel */
+    profile = "Main42210";
+  else {
+    GST_ERROR_OBJECT (self, "invalid profile: %s", profile);
+    return ret;
+  }
+
+  /* VT does not support specific levels for HEVC */
+  key = g_strdup_printf ("HEVC_%s_AutoLevel", profile);
+  ret = CFStringCreateWithBytes (NULL, (const guint8 *) key, strlen (key),
+      kCFStringEncodingASCII, 0);
+
+  GST_INFO_OBJECT (self, "negotiated profile and level %s", key);
+
+  g_free (key);
+  return ret;
+}
+
 static gboolean
 gst_vtenc_negotiate_profile_and_level (GstVTEnc * self, GstStructure * s)
 {
@@ -771,9 +805,16 @@ gst_vtenc_negotiate_profile_and_level (GstVTEnc * self, GstStructure * s)
 
   if (self->profile_level)
     CFRelease (self->profile_level);
-  self->profile_level = gst_vtenc_profile_level_key (self, profile, level);
+
+  if (self->specific_format_id == kCMVideoCodecType_HEVC)
+    self->profile_level =
+        gst_vtenc_hevc_profile_level_key (self, profile, level);
+  else
+    self->profile_level =
+        gst_vtenc_h264_profile_level_key (self, profile, level);
+
   if (self->profile_level == NULL) {
-    GST_ERROR_OBJECT (self, "unsupported h264 profile '%s' or level '%s'",
+    GST_ERROR_OBJECT (self, "unsupported profile '%s' or level '%s'",
         profile, level);
     return FALSE;
   }
@@ -822,6 +863,11 @@ gst_vtenc_negotiate_specific_format_details (GstVideoEncoder * enc)
         if (!gst_vtenc_negotiate_profile_and_level (self, s))
           goto fail;
         break;
+      case kCMVideoCodecType_HEVC:
+        self->specific_format_id = kCMVideoCodecType_HEVC;
+        if (!gst_vtenc_negotiate_profile_and_level (self, s))
+          goto fail;
+        break;
       case GST_kCMVideoCodecType_Some_AppleProRes:
         if (g_strcmp0 (self->details->mimetype, "video/x-prores") != 0) {
           GST_ERROR_OBJECT (self, "format_id == %i mimetype must be Apple "
@@ -931,35 +977,48 @@ gst_vtenc_negotiate_downstream (GstVTEnc * self, CMSampleBufferRef sbuf)
 
   switch (self->details->format_id) {
     case kCMVideoCodecType_H264:
+    case kCMVideoCodecType_HEVC:
     {
       CMFormatDescriptionRef fmt;
       CFDictionaryRef atoms;
-      CFStringRef avccKey;
+      CFStringRef boxKey;
-      CFDataRef avcc;
+      CFDataRef box;
       guint8 *codec_data;
       gsize codec_data_size;
       GstBuffer *codec_data_buf;
-      guint8 sps[3];
+      guint8 sps[12];
 
       fmt = CMSampleBufferGetFormatDescription (sbuf);
       atoms = CMFormatDescriptionGetExtension (fmt,
           kCMFormatDescriptionExtension_SampleDescriptionExtensionAtoms);
-      avccKey = CFStringCreateWithCString (NULL, "avcC", kCFStringEncodingUTF8);
+
-      avcc = CFDictionaryGetValue (atoms, avccKey);
+      if (self->details->format_id == kCMVideoCodecType_HEVC)
-      CFRelease (avccKey);
+        boxKey =
-      codec_data_size = CFDataGetLength (avcc);
+            CFStringCreateWithCString (NULL, "hvcC", kCFStringEncodingUTF8);
+      else
+        boxKey =
+            CFStringCreateWithCString (NULL, "avcC", kCFStringEncodingUTF8);
+
+      box = CFDictionaryGetValue (atoms, boxKey);
+      CFRelease (boxKey);
+      codec_data_size = CFDataGetLength (box);
       codec_data = g_malloc (codec_data_size);
-      CFDataGetBytes (avcc, CFRangeMake (0, codec_data_size), codec_data);
+      CFDataGetBytes (box, CFRangeMake (0, codec_data_size), codec_data);
       codec_data_buf = gst_buffer_new_wrapped (codec_data, codec_data_size);
 
       gst_structure_set (s, "codec_data", GST_TYPE_BUFFER, codec_data_buf,
           NULL);
 
+      if (self->details->format_id == kCMVideoCodecType_HEVC) {
+        sps[0] = codec_data[1];
+        sps[11] = codec_data[12];
+        gst_codec_utils_h265_caps_set_level_tier_and_profile (caps, sps, 12);
+      } else {
         sps[0] = codec_data[1];
         sps[1] = codec_data[2] & ~0xDF;
         sps[2] = codec_data[3];
-
         gst_codec_utils_h264_caps_set_level_and_profile (caps, sps, 3);
+      }
 
       gst_buffer_unref (codec_data_buf);
     }
@@ -1189,7 +1248,7 @@ gst_vtenc_create_session (GstVTEnc * self)
         (gdouble) self->negotiated_fps_n / (gdouble) self->negotiated_fps_d);
 
     /*
-     * https://developer.apple.com/documentation/videotoolbox/vtcompressionsession/compression_properties/profile_and_level_constants
+     * https://developer.apple.com/documentation/videotoolbox/kvtcompressionpropertykey_profilelevel
      */
     status = VTSessionSetProperty (session,
         kVTCompressionPropertyKey_ProfileLevel, self->profile_level);
@@ -1885,8 +1944,11 @@ gst_vtenc_register (GstPlugin * plugin,
 
 static const GstVTEncoderDetails gst_vtenc_codecs[] = {
   {"H.264", "h264", "video/x-h264", kCMVideoCodecType_H264, FALSE},
+  {"H.265/HEVC", "h265", "video/x-h265", kCMVideoCodecType_HEVC, FALSE},
 #ifndef HAVE_IOS
   {"H.264 (HW only)", "h264_hw", "video/x-h264", kCMVideoCodecType_H264, TRUE},
+  {"H.265/HEVC (HW only)", "h265_hw", "video/x-h265", kCMVideoCodecType_HEVC,
+      TRUE},
 #endif
   {"Apple ProRes", "prores", "video/x-prores",
       GST_kCMVideoCodecType_Some_AppleProRes, FALSE},