[Bug 781537] NVDEC - Nvidia Decoder plugin

[GStreamer (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=781537

--- Comment #35 from atopilski at gmail.com <atopilski at gmail.com> ---
(In reply to Snir from comment #26)
> Hi
> 
> I was able to build and use the plugin on k620 and it works pretty well!!
> i had some small issues but maybe i did something wrong while
> running\building it.. 
> 
> -Sometimes it shows only the first frame and freezes, so i have to start it
> again (i'm using gst-launch-1.0 filesrc location=/tmp/file.h264 ! nvh264dec
> ! glimagesink (or ! videoconvert ! xvimagesink))
> 
> -I tried to play two movies one had 1365*1024 resolution and the other one
> 2560*1450, both were encoded same way using Nvidia encoder, but the first
> one gives:
> gstnvh264dec.c:763:gst_nvh264dec_send_decoded_frame:<nvh264dec0> Cuda buffer
> size: 2359296, not equal gstreamer buffer size: 2101248
> 
> -And one thing that i had to change is 
> --- a/sys/nvdec/gstnvh264plugin.c
> +++ b/sys/nvdec/gstnvh264plugin.c
> @@ -35,7 +35,7 @@ plugin_init (GstPlugin * plugin)
>  
>  GST_PLUGIN_DEFINE (GST_VERSION_MAJOR,
>      GST_VERSION_MINOR,
> -    nvh264dec,
> +    nvdec,
>      "Nvidia cuda decoder plugin",
> otherwise the .so file is blacklisted
> 
> Did you encounter these issues?
> 
> Which Nvidia tools are needed in order to build it? currently my driver is
> 381.22 and i'm using cuda packages from non-official repo, but i would
> prefer to use only Nvidia proprietary tools (maybe that's why i had this
> issues? i'm not sure :/)
> 
> (btw I'm building on fedora, the cuda pkgs were called differently so i had
> to change this too:
> -    PKG_CHECK_MODULES([CUDA], [cuda-8.0 cudart-8.0],, [
> +    PKG_CHECK_MODULES([CUDA], [cuda cudart],, [
> )
> 
> 
> Thank you!

Hi, I removed check of different video buffer, this happens in ths lines
pnvh264dec->width = dec_param.ulTargetWidth = PAD_ALIGN(width, 0x3F) * 4 / 3;
pnvh264dec->height = dec_param.ulTargetHeight = PAD_ALIGN(height, 0x0F) * 4 /
3;
 you have non standard video size and round algorithm not working properly.

Also in new version doing only 1 copying and after that spend less 2X memory.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.