[MPlayer-dev-eng] Question regarding MPlayer's OpenGL output

[Mostapha El-Shafey]

> That would indicate that you ended up with a GPU/driver combination that
can't use DMA for upload of 3-component textures.
> I think that is quite rare though, unless you did not align the line
starts...

I tested this on 3 machines, each has different graphics hardware (2 Intel,
1 NVIDIA), OpenGL version (but all greater than 2.1) and CPU.
One thing I forgot to mention is that I was using glTexImage2D in my paintGL
function. I now use glTexSubImage2D to upload slices that are 16 pixels in
height. I'm going to do more testing and if I find something interesting
I'll post it.

On Thu, Oct 20, 2011 at 2:21 PM, Reimar Döffinger
<Reimar.Doeffinger at gmx.de>wrote:

> On 20 Oct 2011, at 18:28, Mostapha El-Shafey <mostapha at viionsystems.com>
> wrote:
> >> A RGB texture is 3 times as much data, so it is 3 times slower
> >> by that alone. Probably GL_BGR_EXT isn't optimized since nobody
> >> (except you) uses it and is another multiple times slower.
> >
> > Thanks. That pointed me in the right direction. I'm now using little
> (10%)
> > CPU to display multiple video feeds, which is equal to MPlayer
> performance.
> > The problem was using RGB and BGR (3 bytes per pixel). I now use GL_BGRA
> and
> > have my incoming frames decoded into a 4 bytes per pixel frame instead of
> a
> > 3 bytes per pixel frame and that had a huge effect on performance. CPU
> usage
> > went down from 50% (one CPU core) to less than 10%.
>
> That would indicate that you ended up with a GPU/driver combination that
> can't use DMA for upload of 3-component textures.
> I think that is quite rare though, unless you did not align the line
> starts...
> _______________________________________________
> MPlayer-dev-eng mailing list
> MPlayer-dev-eng at mplayerhq.hu
> https://lists.mplayerhq.hu/mailman/listinfo/mplayer-dev-eng
>