libav: Integrate FFmpeg's DSD support with GstDsd caps

Code is partially based on the DSD of Robert Tiemann <rtie@gmx.de>:
https://gitlab.freedesktop.org/rtiemann/gstreamer/-/tree/dsd

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3901>

subprojects/gst-libav/docs/gst_plugins_cache.json

@@ -6659,7 +6659,7 @@
                 "long-name": "libav DSD (Direct Stream Digital), least significant bit first decoder",
                 "pad-templates": {
                     "sink": {
-                        "caps": "audio/x-dsd:\n           lsbf: true\n         planar: false\n",
+                        "caps": "audio/x-dsd:\n         format: DSDU8\n reversed-bytes: true\n         layout: interleaved\n",
                         "direction": "sink",
                         "presence": "always"
                     },
@@ -6687,7 +6687,7 @@
                 "long-name": "libav DSD (Direct Stream Digital), least significant bit first, planar decoder",
                 "pad-templates": {
                     "sink": {
-                        "caps": "audio/x-dsd:\n           lsbf: true\n         planar: true\n",
+                        "caps": "audio/x-dsd:\n         format: DSDU8\n reversed-bytes: true\n         layout: non-interleaved\n",
                         "direction": "sink",
                         "presence": "always"
                     },
@@ -6715,7 +6715,7 @@
                 "long-name": "libav DSD (Direct Stream Digital), most significant bit first decoder",
                 "pad-templates": {
                     "sink": {
-                        "caps": "audio/x-dsd:\n           lsbf: false\n         planar: false\n",
+                        "caps": "audio/x-dsd:\n         format: DSDU8\n reversed-bytes: false\n         layout: interleaved\n",
                         "direction": "sink",
                         "presence": "always"
                     },
@@ -6743,7 +6743,7 @@
                 "long-name": "libav DSD (Direct Stream Digital), most significant bit first, planar decoder",
                 "pad-templates": {
                     "sink": {
-                        "caps": "audio/x-dsd:\n           lsbf: false\n         planar: true\n",
+                        "caps": "audio/x-dsd:\n         format: DSDU8\n reversed-bytes: false\n         layout: non-interleaved\n",
                         "direction": "sink",
                         "presence": "always"
                     },

subprojects/gst-libav/ext/libav/gstavcodecmap.c

@@ -34,6 +34,7 @@
 
 #include <gst/video/video.h>
 #include <gst/audio/audio.h>
+#include <gst/audio/gstdsd.h>
 #include <gst/pbutils/codec-utils.h>
 
 /* IMPORTANT: Keep this sorted by the ffmpeg channel masks */
@@ -640,6 +641,36 @@ gst_ff_aud_caps_new (AVCodecContext * context, AVCodec * codec,
         rates = l_rates;
         break;
       }
+      case AV_CODEC_ID_DSD_LSBF:
+      case AV_CODEC_ID_DSD_MSBF:
+      case AV_CODEC_ID_DSD_LSBF_PLANAR:
+      case AV_CODEC_ID_DSD_MSBF_PLANAR:
+      {
+        const static gint l_rates[] = {
+          GST_DSD_MAKE_DSD_RATE_44x (64),
+          GST_DSD_MAKE_DSD_RATE_48x (64),
+          GST_DSD_MAKE_DSD_RATE_44x (128),
+          GST_DSD_MAKE_DSD_RATE_48x (128),
+          GST_DSD_MAKE_DSD_RATE_44x (256),
+          GST_DSD_MAKE_DSD_RATE_48x (256),
+          GST_DSD_MAKE_DSD_RATE_44x (512),
+          GST_DSD_MAKE_DSD_RATE_48x (512),
+          GST_DSD_MAKE_DSD_RATE_44x (1024),
+          GST_DSD_MAKE_DSD_RATE_48x (1024),
+          GST_DSD_MAKE_DSD_RATE_44x (2048),
+          GST_DSD_MAKE_DSD_RATE_48x (2048),
+        };
+        /* There is no clearly defined maximum number of channels in DSD.
+         * The DSF spec mentions a maximum of 6 channels, while the DSDIFF
+         * spec mentions up to 65535 channels. DSDIFF stores DSD in an
+         * interleaved, DSF in a planar fashion. But there is no reason
+         * why some other format couldn't have more than 6 interleaved
+         * channels for example. */
+        maxchannels = 65535;
+        n_rates = G_N_ELEMENTS (l_rates);
+        rates = l_rates;
+        break;
+      }
       default:
         break;
     }
@@ -2318,33 +2349,45 @@ gst_ffmpeg_codecid_to_caps (enum AVCodecID codec_id,
           NULL);
       break;
     case AV_CODEC_ID_DSD_LSBF:
-      caps =
+    case AV_CODEC_ID_DSD_MSBF:
-          gst_ff_aud_caps_new (context, NULL, codec_id, encode, "audio/x-dsd",
+    case AV_CODEC_ID_DSD_LSBF_PLANAR:
-          NULL);
+    case AV_CODEC_ID_DSD_MSBF_PLANAR:
-      gst_caps_set_simple (caps, "lsbf", G_TYPE_BOOLEAN,
+    {
-          TRUE, "planar", G_TYPE_BOOLEAN, FALSE, NULL);
+      gboolean reversed_bytes;
+      gboolean interleaved;
+
+      switch (codec_id) {
+        case AV_CODEC_ID_DSD_LSBF:
+          reversed_bytes = TRUE;
+          interleaved = TRUE;
           break;
         case AV_CODEC_ID_DSD_MSBF:
-      caps =
+          reversed_bytes = FALSE;
-          gst_ff_aud_caps_new (context, NULL, codec_id, encode, "audio/x-dsd",
+          interleaved = TRUE;
-          NULL);
-      gst_caps_set_simple (caps, "lsbf", G_TYPE_BOOLEAN,
-          FALSE, "planar", G_TYPE_BOOLEAN, FALSE, NULL);
           break;
         case AV_CODEC_ID_DSD_LSBF_PLANAR:
-      caps =
+          reversed_bytes = TRUE;
-          gst_ff_aud_caps_new (context, NULL, codec_id, encode, "audio/x-dsd",
+          interleaved = FALSE;
-          NULL);
-      gst_caps_set_simple (caps, "lsbf", G_TYPE_BOOLEAN,
-          TRUE, "planar", G_TYPE_BOOLEAN, TRUE, NULL);
           break;
         case AV_CODEC_ID_DSD_MSBF_PLANAR:
+          reversed_bytes = FALSE;
+          interleaved = FALSE;
+          break;
+        default:
+          reversed_bytes = FALSE;
+          interleaved = FALSE;
+          break;
+      }
+
       caps =
           gst_ff_aud_caps_new (context, NULL, codec_id, encode, "audio/x-dsd",
-          NULL);
+          "format", G_TYPE_STRING, "DSDU8",
-      gst_caps_set_simple (caps, "lsbf", G_TYPE_BOOLEAN,
+          "reversed-bytes", G_TYPE_BOOLEAN, reversed_bytes,
-          FALSE, "planar", G_TYPE_BOOLEAN, TRUE, NULL);
+          "layout", G_TYPE_STRING,
+          (interleaved) ? "interleaved" : "non-interleaved", NULL);
+
       break;
+    }
     case AV_CODEC_ID_APTX:
       caps =
           gst_ff_aud_caps_new (context, NULL, codec_id, encode, "audio/aptx",

subprojects/gst-libav/ext/libav/gstavdemux.c

@@ -30,6 +30,7 @@
 /* #include <ffmpeg/avi.h> */
 #include <gst/gst.h>
 #include <gst/base/gstflowcombiner.h>
+#include <gst/audio/gstdsd.h>
 
 #include "gstav.h"
 #include "gstavcodecmap.h"
@@ -1576,6 +1577,32 @@ gst_ffmpegdemux_loop (GstFFMpegDemux * demux)
     stream->discont = FALSE;
   }
 
+  /* If we are demuxing planar DSD data, add the necessary
+   * meta to inform downstream about the planar layout. */
+  switch (avstream->codecpar->codec_id) {
+    case AV_CODEC_ID_DSD_LSBF_PLANAR:
+    case AV_CODEC_ID_DSD_MSBF_PLANAR:
+    {
+      int channel_idx;
+      int num_channels = avstream->codecpar->channels;
+      int num_bytes_per_channel = pkt.size / num_channels;
+      GstDsdPlaneOffsetMeta *plane_ofs_meta;
+
+      plane_ofs_meta = gst_buffer_add_dsd_plane_offset_meta (outbuf,
+          avstream->codecpar->channels, num_bytes_per_channel, NULL);
+
+      for (channel_idx = 0; channel_idx < num_channels; ++channel_idx) {
+        plane_ofs_meta->offsets[channel_idx] =
+            num_bytes_per_channel * channel_idx;
+      }
+
+      break;
+    }
+
+    default:
+      break;
+  }
+
   GST_DEBUG_OBJECT (demux,
       "Sending out buffer time:%" GST_TIME_FORMAT " size:%" G_GSIZE_FORMAT,
       GST_TIME_ARGS (timestamp), gst_buffer_get_size (outbuf));