Marcin Slusarz
2011-Sep-10 22:54 UTC
[Nouveau] [Mesa-dev] [PATCH v2] nouveau: Add support for XvMC acceleration
On Sat, Sep 10, 2011 at 06:59:39PM +0200, Maarten Lankhorst wrote:> Hey ?, > > On 09/10/2011 06:28 PM, ? Emeric wrote: > > Hi Maarten, > > I tried the v2 patch with my good old geforce 8600M GS, and it worked well. > > The CPU usage seems a little higher than usual, I'd said 60% instead of 40% > > for a 720p video, but I'll try with a higher bitrate video and without > > building with debug parameters to see if I can found a performance gain. > > The only problem is that I cannot seek through the video without an segfault > > on that error "X11 error: BadAccess (attempt to access private resource > > denied)", I attached a gdb backtrace. It appears to be related to the > > drawing of the mplayer OSD. > > > That's either a bug in mplayer or in the state tracker. What's oy and ox > set to at the time of the crash? and where does xvimage->data point to :) >This is a bug in xf86-video-nouveau actually, and it leads to memory corruption in mplayer. Patch below fixes it, but OSD still doesn't work. ---
Marcin Slusarz
2011-Sep-13 13:52 UTC
[Nouveau] [Mesa-dev] [PATCH v2] nouveau: Add support for XvMC acceleration
On Sun, Sep 11, 2011 at 12:54:20AM +0200, Marcin Slusarz wrote:> On Sat, Sep 10, 2011 at 06:59:39PM +0200, Maarten Lankhorst wrote: > > Hey ?, > > > > On 09/10/2011 06:28 PM, ? Emeric wrote: > > > Hi Maarten, > > > I tried the v2 patch with my good old geforce 8600M GS, and it worked well. > > > The CPU usage seems a little higher than usual, I'd said 60% instead of 40% > > > for a 720p video, but I'll try with a higher bitrate video and without > > > building with debug parameters to see if I can found a performance gain. > > > The only problem is that I cannot seek through the video without an segfault > > > on that error "X11 error: BadAccess (attempt to access private resource > > > denied)", I attached a gdb backtrace. It appears to be related to the > > > drawing of the mplayer OSD. > > > > > That's either a bug in mplayer or in the state tracker. What's oy and ox > > set to at the time of the crash? and where does xvimage->data point to :) > > > This is a bug in xf86-video-nouveau actually, and it leads to memory corruption > in mplayer. > > Patch below fixes it, but OSD still doesn't work. >With commit f626d99edf17a1d3203b257f6fe6993229ea6866 in mesa OSD in mplayer works (Yeah! ;). I've submitted patch for mplayer to disable OSD instead of crashing. Ben, can you review and push patch in the previous mail? (xv: support AI44/IA44) Thanks. Marcin