Re: [PATCH spice-gtk v2 3/3] channel-display: Instrument code to get frame and queue statistics

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



> On Wed, Jan 23, 2019 at 08:57:36AM -0500, Frediano Ziglio wrote:
> > > On Wed, Jan 23, 2019 at 11:28:57AM +0000, Frediano Ziglio wrote:
> > > > This patch is based on some work from Victor Toso (statistics) and
> > > > Snir Sheriber (GStreamer probing).
> > > > All GstBuffers are queued into decoding_queue and a probe
> > > > is attached to the sink in order to understand when the
> > > > buffers are decoded.

Maybe I could add here:

"Previously we didn't add frames to decoding_queue in case of direct
rendering as there was nothing to dequeue them, now we need it to
compute the timing."

> > > > When a buffer is decoded the time spent and queue length
> > > > are computed.
> > > > 
> > > > Signed-off-by: Frediano Ziglio <fziglio@xxxxxxxxxx>
> > > > ---
> > > >  src/channel-display-gst.c  | 45 +++++++++++++++++++++++++++++++++-----
> > > >  src/channel-display-priv.h |  3 +++
> > > >  src/channel-display.c      |  1 +
> > > >  3 files changed, 43 insertions(+), 6 deletions(-)
> > > > 
> > > > Changes since v1:
> > > > - use G_GUINT64_FORMAT instead of PRIu64, SPICE_DEBUG is using GLib
> > > >   function which expects GNU formatting, it gives a warning compiling
> > > >   for Windows
> > > > 
> > > > diff --git a/src/channel-display-gst.c b/src/channel-display-gst.c
> > > > index 08c9f8fb..c15f94f2 100644
> > > > --- a/src/channel-display-gst.c
> > > > +++ b/src/channel-display-gst.c
> > > > @@ -248,6 +248,12 @@ static SpiceGstFrame
> > > > *get_decoded_frame(SpiceGstDecoder *decoder, GstBuffer *buf
> > > >      if (num_frames_dropped != 0) {
> > > >          SPICE_DEBUG("the GStreamer pipeline dropped %u frames",
> > > >          num_frames_dropped);
> > > >      }
> > > > +    if (gstframe) {
> > > > +        const SpiceFrame *frame = gstframe->encoded_frame;
> > > > +        uint64_t duration = g_get_monotonic_time() -
> > > > frame->creation_time;
> > > > +        SPICE_DEBUG("frame mm_time %u size %u decoded time %"
> > > > G_GUINT64_FORMAT " queue %u",
> > > > +                    frame->mm_time, frame->size, duration,
> > > > decoder->decoding_queue->length);
> > 
> > [1]
> > 
> > > > +    }
> > > >      return gstframe;
> > > >  }
> > > >  
> > > > @@ -391,6 +397,34 @@ static void app_source_setup(GstElement *pipeline
> > > > G_GNUC_UNUSED,
> > > >      gst_caps_unref(caps);
> > > >      decoder->appsrc = GST_APP_SRC(gst_object_ref(source));
> > > >  }
> > > > +
> > > > +static GstPadProbeReturn
> > > > +sink_event_probe(GstPad *pad, GstPadProbeInfo *info, gpointer data)
> > > > +{
> > > > +    SpiceGstDecoder *decoder = (SpiceGstDecoder*)data;
> > > > +
> > > > +    if (info->type & GST_PAD_PROBE_TYPE_BUFFER) { // Buffer arrived
> > > 
> > > C-style comments instead?
> > > 
> > 
> > C99 added "//" style comment, so unless you want C89 compatibility
> > the comment above IS C-style.
> > We use the new style in a lot of places for short comments.
> > 
> > > > +        GstBuffer *buffer = GST_PAD_PROBE_INFO_BUFFER(info);
> > > > +        g_mutex_lock(&decoder->queues_mutex);
> > > > +        SpiceGstFrame *gstframe = get_decoded_frame(decoder, buffer);
> > > > +        if (gstframe) {
> > > > +            free_gst_frame(gstframe);
> > > > +        }
> > 
> > [2]
> > 
> > > > +        g_mutex_unlock(&decoder->queues_mutex);
> > > > +    }
> > > > +    return GST_PAD_PROBE_OK;
> > > > +}
> > > > +
> > > > +/* This function is called to used to set a probe on the sink */
> > > > +static void
> > > > +add_elem_cb(GstBin * pipeline, GstBin * bin, GstElement * element,
> > > > SpiceGstDecoder *decoder)
> > > > +{
> > > > +    if (GST_IS_BASE_SINK(element) && decoder->appsink == NULL) {
> > > > +        GstPad *pad = gst_element_get_static_pad(element, "sink");
> > > > +        gst_pad_add_probe(pad, GST_PAD_PROBE_TYPE_BUFFER,
> > > > sink_event_probe, decoder, NULL);
> > > > +        gst_object_unref(pad);
> > > > +    }
> > > > +}
> > > >  #endif
> > > >  
> > > >  static gboolean create_pipeline(SpiceGstDecoder *decoder)
> > > > @@ -455,6 +489,7 @@ static gboolean create_pipeline(SpiceGstDecoder
> > > > *decoder)
> > > >  #endif
> > > >      }
> > > >  
> > > > +    g_signal_connect(playbin, "deep-element-added",
> > > > G_CALLBACK(add_elem_cb), decoder);
> > > >      g_signal_connect(playbin, "source-setup",
> > > >      G_CALLBACK(app_source_setup), decoder);
> > > >  
> > > >      g_object_set(playbin,
> > > > @@ -660,12 +695,10 @@ static gboolean
> > > > spice_gst_decoder_queue_frame(VideoDecoder *video_decoder,
> > > >      GST_BUFFER_DTS(buffer) = GST_CLOCK_TIME_NONE;
> > > >      GST_BUFFER_PTS(buffer) = gst_clock_get_time(decoder->clock) -
> > > >      gst_element_get_base_time(decoder->pipeline) + ((uint64_t)MAX(0,
> > > >      latency)) * 1000 * 1000;
> > > >  
> > > > -    if (decoder->appsink != NULL) {
> > > 
> > > From the commit log, I don't follow why this check can be removed
> > > 
> > 
> > Oh, we use the decoding_queue to match the input with the output and
> > compute
> > the decoded time, see [1] and [2].
> 
> Yes, that I can follow from commit log. What I cannot follow was
> what I asked, the check if (decoder->appsink != NULL) was added
> almost a year ago with:
> 
>     commit 424397546463698c2912b15449ef5e94f2c39c12
>     Author: Snir Sheriber <ssheribe@xxxxxxxxxx>
>     Date:   Thu Mar 8 15:57:31 2018 +0200
>     Gstreamer: Use GstVideoOverlay if possible
> 
> note:
> 
>     -    SpiceGstFrame *gst_frame = create_gst_frame(buffer, frame);
>     -    g_mutex_lock(&decoder->queues_mutex);
>     -    g_queue_push_tail(decoder->decoding_queue, gst_frame);
>     -    g_mutex_unlock(&decoder->queues_mutex);
>     +    if (decoder->appsink != NULL) {
>     +        SpiceGstFrame *gst_frame = create_gst_frame(buffer, frame);
>     +        g_mutex_lock(&decoder->queues_mutex);
>     +        g_queue_push_tail(decoder->decoding_queue, gst_frame);
>     +        g_mutex_unlock(&decoder->queues_mutex);
>     +    } else {
>     +        frame->free(frame);
>     +        frame = NULL;
>     +    }
>               
> 
> So, what I mean is that we might not need to check for
> decoder->appsink anymore but I don't follow why and why this
> patch is removing it.
> 

Previously without the decoder->appsink check the decoding_queue would only
grow as there is nothing dequeuing. Now we add the probe which needs the queue
to compute the timing and also will dequeue from it.

> > > > -        SpiceGstFrame *gst_frame = create_gst_frame(buffer, frame);
> > > > -        g_mutex_lock(&decoder->queues_mutex);
> > > > -        g_queue_push_tail(decoder->decoding_queue, gst_frame);
> > > > -        g_mutex_unlock(&decoder->queues_mutex);
> > > > -    }
> > > > +    SpiceGstFrame *gst_frame = create_gst_frame(buffer, frame);
> > > > +    g_mutex_lock(&decoder->queues_mutex);
> > > > +    g_queue_push_tail(decoder->decoding_queue, gst_frame);
> > > > +    g_mutex_unlock(&decoder->queues_mutex);
> > > 
> > > Cheers,
> > > Victor
> > > 
> > > >  
> > > >      if (gst_app_src_push_buffer(decoder->appsrc, buffer) !=
> > > >      GST_FLOW_OK) {
> > > >          SPICE_DEBUG("GStreamer error: unable to push frame");
> > > > diff --git a/src/channel-display-priv.h b/src/channel-display-priv.h
> > > > index c5c3f5c8..495df7ac 100644
> > > > --- a/src/channel-display-priv.h
> > > > +++ b/src/channel-display-priv.h
> > > > @@ -45,6 +45,9 @@ struct SpiceFrame {
> > > >      uint8_t *data;
> > > >      uint32_t size;
> > > >      gpointer data_opaque;
> > > > +
> > > > +    /* stats */
> > > > +    gint64 creation_time;
> > > >  };
> > > >  
> > > >  typedef struct VideoDecoder VideoDecoder;
> > > > diff --git a/src/channel-display.c b/src/channel-display.c
> > > > index ae2c4fbc..d46b3a37 100644
> > > > --- a/src/channel-display.c
> > > > +++ b/src/channel-display.c
> > > > @@ -1657,6 +1657,7 @@ static SpiceFrame *spice_frame_new(display_stream
> > > > *st,
> > > >      frame->size = data_size;
> > > >      frame->data_opaque = in;
> > > >      spice_msg_in_ref(in);
> > > > +    frame->creation_time = g_get_monotonic_time();
> > > >      return frame;
> > > >  }
> > > >  
> > 
> > Frediano
> 
_______________________________________________
Spice-devel mailing list
Spice-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/spice-devel




[Index of Archives]     [Linux Virtualization]     [Linux Virtualization]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Security]     [Bugtraq]     [Linux OMAP]     [Linux MIPS]     [ECOS]     [Asterisk Internet PBX]     [Linux API]     [Monitors]