Re: DRI on Savage: works great, minor issues, couple questions
> > > The other issue is more annoying: there is visual noise whenever there > > > is significant OpenGL activity in a full-screen window Transient noisy bars two or three pixel wide. What surprises me is that I've seen them both vertical and horizontal. > TwisterK has a lower memory bandwidth. Yep, it does look like memory bandwidth issues. Any hints on how I could write a testcase to trigger the issue? Juliusz --- This SF.Net email is sponsored by BEA Weblogic Workshop FREE Java Enterprise J2EE developer tools! Get your free copy of BEA WebLogic Workshop 8.1 today. http://ads.osdn.com/?ad_id=4721&alloc_id=10040&op=click -- ___ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel
Re: DRI on Savage: works great, minor issues, couple questions
On Mon, 19 Jul 2004 15:08:11 -0400 Alex Deucher <[EMAIL PROTECTED]> wrote: > On 19 Jul 2004 19:34:19 +0200, Juliusz Chroboczek <[EMAIL PROTECTED]> wrote: [snip] > > > > There are just two minor issues. I've managed to get it to lockup > > twice (X server frozen with 100% cpu, the system can be rebooted > > remotely), but haven't been able to reproduce the problem since. > > > > The other issue is more annoying: there is visual noise whenever there > > is significant OpenGL activity in a full-screen window (rendered at > > 640x400, scaled to 1024x768); this happens both with amoeba and > > MPlayer. There is no such problem with 640x400 windows. > > hmmm...what exactly does it look like? Felix may know more about that. I havn't seen such noise. But I have a different integrated chipset: ProSavageDDR. TwisterK has a lower memory bandwidth. It could be the 3D engine and the hardware scaler fighting for memory bandwidth. That's pure speculation though. > > > > > Oh, a few questions if any of you have got time. Without any > > customisation, the server reports that it found an AGP 4x chipset, but > > runs in AGP 1x. Setting the server to AGP 4x in the XF86Config > > appears to work. Could anyone explain what's going on? (Oh, and what > > is it supposed to speed up? Texture upload?) > > The server defaults to 1x like (all?) other 3d drivers as some older > chipsets had problems with higher agp speeds. Theoretically it should > speed up AGP transfers (just textures at the moment), but in the real > world rarely seems to affect performance. AFAIK it doesn't make any difference on integrated graphics chips. They access both AGP memory and (shared) system memory with the full bandwidth. > [snip] | Felix Kühling <[EMAIL PROTECTED]> http://fxk.de.vu | | PGP Fingerprint: 6A3C 9566 5B30 DDED 73C3 B152 151C 5CC1 D888 E595 | --- This SF.Net email is sponsored by BEA Weblogic Workshop FREE Java Enterprise J2EE developer tools! Get your free copy of BEA WebLogic Workshop 8.1 today. http://ads.osdn.com/?ad_idG21&alloc_id040&op=click -- ___ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel
Re: DRI on Savage: works great, minor issues, couple questions
On 19 Jul 2004 19:34:19 +0200, Juliusz Chroboczek <[EMAIL PROTECTED]> wrote: > Dear all, > > Many thanks to all of those involved in getting the DRI working on the > Savage. > > I just installed yesterday the 20040718 snapshot on a laptop with a > Twister-K running an up-to-date Debian Testing. The installation went > without a hitch once I realised there's an automated installation > script (I'll read the README next time). > > A nice side-effect of the installation is decent XVideo support (the > stock XFree86 server does nearest-neighbour scaling). There was a hw vertical interpolated scaling enable bit that the stock driver forgot to set. I found it when merging S3 and Tim's driver. The new code also uses the BCI on integrated chips like the twister for mastered image transfer and colorspace conversion. I also have a patch for some obsure colorkey bug that was sent to me, but alas, it has slipped my mind for a while and I haven't had a chance to test and commit it. > > There are just two minor issues. I've managed to get it to lockup > twice (X server frozen with 100% cpu, the system can be rebooted > remotely), but haven't been able to reproduce the problem since. > > The other issue is more annoying: there is visual noise whenever there > is significant OpenGL activity in a full-screen window (rendered at > 640x400, scaled to 1024x768); this happens both with amoeba and > MPlayer. There is no such problem with 640x400 windows. hmmm...what exactly does it look like? Felix may know more about that. > > Oh, a few questions if any of you have got time. Without any > customisation, the server reports that it found an AGP 4x chipset, but > runs in AGP 1x. Setting the server to AGP 4x in the XF86Config > appears to work. Could anyone explain what's going on? (Oh, and what > is it supposed to speed up? Texture upload?) The server defaults to 1x like (all?) other 3d drivers as some older chipsets had problems with higher agp speeds. Theoretically it should speed up AGP transfers (just textures at the moment), but in the real world rarely seems to affect performance. > > The server complains it cannot open /dev/agp_bios (I'm running devfs), > but doesn't seem to mind. What's that? I've never heard of that. > > Finally, does anyone know what scaling algorithm is used by the XVideo > implementation? It looks like nearest-neighbour when downscaling, > bilinear when upscaling by a reasonable amount, and linear in the Y > direction when upscaling a lot. I don't remember off hand. As I recall it varies somewhat depending on what factor you are scaling it by. Check the source (savage_video.c). > > Thanks, > > Juliusz Chroboczek > > P.S. I'll try to check the list archives, but please be so kind as to > CC me with any replies. > --- This SF.Net email is sponsored by BEA Weblogic Workshop FREE Java Enterprise J2EE developer tools! Get your free copy of BEA WebLogic Workshop 8.1 today. http://ads.osdn.com/?ad_id=4721&alloc_id=10040&op=click -- ___ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel
DRI on Savage: works great, minor issues, couple questions
Dear all, Many thanks to all of those involved in getting the DRI working on the Savage. I just installed yesterday the 20040718 snapshot on a laptop with a Twister-K running an up-to-date Debian Testing. The installation went without a hitch once I realised there's an automated installation script (I'll read the README next time). A nice side-effect of the installation is decent XVideo support (the stock XFree86 server does nearest-neighbour scaling). There are just two minor issues. I've managed to get it to lockup twice (X server frozen with 100% cpu, the system can be rebooted remotely), but haven't been able to reproduce the problem since. The other issue is more annoying: there is visual noise whenever there is significant OpenGL activity in a full-screen window (rendered at 640x400, scaled to 1024x768); this happens both with amoeba and MPlayer. There is no such problem with 640x400 windows. Oh, a few questions if any of you have got time. Without any customisation, the server reports that it found an AGP 4x chipset, but runs in AGP 1x. Setting the server to AGP 4x in the XF86Config appears to work. Could anyone explain what's going on? (Oh, and what is it supposed to speed up? Texture upload?) The server complains it cannot open /dev/agp_bios (I'm running devfs), but doesn't seem to mind. What's that? Finally, does anyone know what scaling algorithm is used by the XVideo implementation? It looks like nearest-neighbour when downscaling, bilinear when upscaling by a reasonable amount, and linear in the Y direction when upscaling a lot. Thanks, Juliusz Chroboczek P.S. I'll try to check the list archives, but please be so kind as to CC me with any replies. --- This SF.Net email is sponsored by BEA Weblogic Workshop FREE Java Enterprise J2EE developer tools! Get your free copy of BEA WebLogic Workshop 8.1 today. http://ads.osdn.com/?ad_id=4721&alloc_id=10040&op=click -- ___ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel
