From mboxrd@z Thu Jan 1 00:00:00 1970 From: Roger Subject: Re: "Decode videos" Date: Mon, 6 Dec 2010 20:11:16 -0900 Message-ID: <20101207051116.GD4066@localhost2.local> References: <20101206200707.4793a30a@amd64linuxmsi.rentzhouse.lan> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Content-Disposition: inline In-Reply-To: <20101206200707.4793a30a-bEIzoA7YQWm4miubQJdzGGS2/mPcPMSI/gJWpUujD0o@public.gmane.org> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: nouveau-bounces+gcfxn-nouveau=m.gmane.org-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org Errors-To: nouveau-bounces+gcfxn-nouveau=m.gmane.org-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org To: nouveau-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org List-Id: nouveau.vger.kernel.org On Mon, Dec 06, 2010 at 08:07:07PM -0500, Jimmy Rentz wrote: >Initially I would say XVMC. VDPAU expects full hw bitstream >decode. XVMC only does decode starting at idct/mb. I am guessing it >is possible to support VDPAU by doing the earlier stages in software >then handling the reset to hw. Well, unless someone figures out some >way to handle the full pipeline in hw via shaders which is something >I have no time or knowledge. > >The latest version of the video decode/vpe/nv40 patches are here: >http://people.freedesktop.org/~jb17bsome/vpe/ddx/ddx_nouveau_vpe_v1.patch >http://people.freedesktop.org/~jb17bsome/vpe/drm/drm_nouveau_vpe_v5.patch >http://people.freedesktop.org/~jb17bsome/vpe/kernel/kernel_nouveau_vpe_v6.patch >http://people.freedesktop.org/~jb17bsome/vpe/mesa/mesa_pipe_video_nouveau_vpe_v3.patch >For mesa you need to pull the pipe-video branch. All the others apply >fine against the latest branches for the ddx, drm, nouveau kernel. > >The current patches integrate vpe better with gem/ttm. This means >that relocs/fences are emitted and it doesn't just pin all memory like >before. Though, it does runs much slower compared to my last non-gem >version located under the old directories in (ddx/drm/kernel/mesa). I >am fairly certain from oprofile dumps that it is because of too many >relocs emitted or something, but I haven't followed up. > >Anyway hopes this helps. > >Jimmy Just a quick ping on this topic. I have 2 video cards with XvMC capabilities on 450P3 and 2x750P3 CPU's, of which, XvMC is required for playback of high resolution video... unless I use the rough MPlayer slow CPU scaling hacks. I hopefully plan to experiment with the XvMC patches in the near future. -- Roger http://rogerx.freeshell.org/