<div>Hi Junyan:</div><div><br></div><div>do you mean the following?</div><div>video frame maps to cpu memory address space, </div><div>then a deep copy is required to upload the data to gpu memory address space by clutter_gst_yv12_upload().</div><div><br></div><div>it looks inefficient, usually we can bind texture from dmabuf fd.</div><div><br></div><div><br></div><br><br><div id="">--------------------------------<br></div><div id=""><br></div><br><div id="origbody"><div style="background: #f2f2f2;">----- 原始邮件 -----<br>发件人：Junyan He <junyan.he@hotmail.com><br>收件人："halleyzhao@sina.com" <halleyzhao@sina.com>, Discussion of the development of and with GStreamer <gstreamer-devel@lists.freedesktop.org><br>主题：Re: how does clutter-gst/totem work with dmabuf buffers from v4l2 decoders?<br>日期：2019年09月10日 14点55分<br></div><br>vaapi buffer can be mapped by the GstVideoMeta->map function.<br>On 二, 2019-09-10 at 14:24 +0800, HalleyZhao wrote:<br>> Hi experts:<br>> <br>> I notice that totem uses clutter-gst as videosink, then I look into<br>> clutter-gst (https://github.com/GNOME/clutter-gst).<br>> however, I haven't found how it work with hardware accelerated<br>> decoder (usually generate dma_buf buffer for a video frame).<br>> anyone can give some hints? thanks.<br>> <br>> <br>> --------------------------------<br>> <br>> _______________________________________________<br>> gstreamer-devel mailing list<br>> gstreamer-devel@lists.freedesktop.org<br>> https://lists.freedesktop.org/mailman/listinfo/gstreamer-devel<br>_______________________________________________<br>gstreamer-devel mailing list<br>gstreamer-devel@lists.freedesktop.org<br>https://lists.freedesktop.org/mailman/listinfo/gstreamer-devel</div>