audioconvert: Optimize transform_caps()

If the second and next caps structures are a subset of the already existing
transformed caps we can safely skip them because we would transform them to
the same caps again.

This makes gst_pad_get_caps() on an audiotestsrc ! audioconvert !
audioconvert ! audioconvert ! fakesink pipeline about 1.7 times faster.

diff --git a/gst/audioconvert/gstaudioconvert.c b/gst/audioconvert/gstaudioconvert.c
index b3250a7..26506d2 100644 (file)
--- a/gst/audioconvert/gstaudioconvert.c
+++ b/gst/audioconvert/gstaudioconvert.c
@@ -577,6 +577,21 @@ gst_audio_convert_transform_caps (GstBaseTransform * base,
 
   for (j = 0; j < n; j++) {
     structure = gst_caps_get_structure (caps, j);
+
+    if (j > 0) {
+      GstCaps *tmp = gst_caps_new_full (gst_structure_copy (structure), NULL);
+
+      /* If the new structure is a subset of the already existing transformed
+       * caps we can safely skip it because we would transform it to the
+       * same caps again.
+       */
+      if (gst_caps_is_subset (tmp, ret)) {
+        gst_caps_unref (tmp);
+        continue;
+      }
+      gst_caps_unref (tmp);
+    }
+
     structure_name = gst_structure_get_name (structure);
 
     isfloat = strcmp (structure_name, "audio/x-raw-float") == 0;