185.19 + VDPAU blit path == stall - Page 4

**Stephen Warren** · 10-23-09, 02:32 PM

I grabbed kaffeine-0.8.8, compiled it against xine-vdpau r281, ran a couple 3 minute clips and ~45m of a DVD through it, validated the video_out_vdpau was running through the non-zero-timestamp path in all cases, using 8400GS (G98) to a 1600x1200 60Hz DVI output, and didn't see any issues.

crisalide · 10-25-09, 05:18 PM

Stephen,

Until now, i used a G96 9500GT. This evening, I've plugged in the G98 8400gs and indeed, no stall after more than 15 minutes of live 1080i h264.
I'll do more test this week with the 9500GT and give you all details.
Thanx.

Btw:
Option "UseEvents" "False", made no difference with the 9500GT.
Option "DamageEvents" "False", makes VDPAU unable to create the presentation queue.

**Stephen Warren** · 10-26-09, 11:07 AM

OK. I'll see if I can track down a G96 and/or a 9500GT here too. Thanks for the information; it seems like it's a good clue.

**Stephen Warren** · 10-26-09, 12:06 PM

Looks like I have a repro with a G96. Nothing much useful in the log files so far, even with a debug driver. However, at least I have something to go on now...

**Stephen Warren** · 11-02-09, 12:59 PM

Good news: We have identified an issue that explains these symptoms, and a fix that solves the issue in our testing. It will be included in a future 190.* driver release.

crisalide · 11-04-09, 02:25 PM

Ah, great !!
Waiting for next release, then

Btw, what kind of issue?

**Stephen Warren** · 11-04-09, 04:44 PM

There was a problem related to the handling of timestamps. If the timestamp for the frame was soon in the future (as opposed to further out in the future) when queued, sometimes the processing for that frame would be delayed until the requested time plus some offset, in the order of seconds.

mlampard · 11-24-09, 01:38 AM

Thanks Stephen, the issue is now resolved with 195.22 drivers on my system.

jdobry · 11-24-09, 04:38 AM

@Stephen Warren: Is it fixed in 195.22 version? (http://www.nvnews.net/vbulletin/show....php?p=2129618)

EDIT: looks like fixed

crisalide · 11-24-09, 04:00 PM

Indeed, seems fixed in 195.22
Stephen, thanx for your patience and hard work

**Stephen Warren** · 11-24-09, 04:55 PM

Excellent news! Thanks very much for the confirmation.

10-23-09, 02:32 PM	#37
Stephen Warren Moderator Join Date: Aug 2005 Posts: 1,327	Re: 185.19 + VDPAU blit path == stall I grabbed kaffeine-0.8.8, compiled it against xine-vdpau r281, ran a couple 3 minute clips and ~45m of a DVD through it, validated the video_out_vdpau was running through the non-zero-timestamp path in all cases, using 8400GS (G98) to a 1600x1200 60Hz DVI output, and didn't see any issues.

10-25-09, 05:18 PM	#38
crisalide Registered User Join Date: Dec 2008 Posts: 173	Re: 185.19 + VDPAU blit path == stall Stephen, Until now, i used a G96 9500GT. This evening, I've plugged in the G98 8400gs and indeed, no stall after more than 15 minutes of live 1080i h264. I'll do more test this week with the 9500GT and give you all details. Thanx. Btw: Option "UseEvents" "False", made no difference with the 9500GT. Option "DamageEvents" "False", makes VDPAU unable to create the presentation queue.

10-26-09, 11:07 AM	#39
Stephen Warren Moderator Join Date: Aug 2005 Posts: 1,327	Re: 185.19 + VDPAU blit path == stall OK. I'll see if I can track down a G96 and/or a 9500GT here too. Thanks for the information; it seems like it's a good clue.

10-26-09, 12:06 PM	#40
Stephen Warren Moderator Join Date: Aug 2005 Posts: 1,327	Re: 185.19 + VDPAU blit path == stall Looks like I have a repro with a G96. Nothing much useful in the log files so far, even with a debug driver. However, at least I have something to go on now...

11-02-09, 12:59 PM	#41
Stephen Warren Moderator Join Date: Aug 2005 Posts: 1,327	Re: 185.19 + VDPAU blit path == stall Good news: We have identified an issue that explains these symptoms, and a fix that solves the issue in our testing. It will be included in a future 190.* driver release.

11-04-09, 02:25 PM	#42
crisalide Registered User Join Date: Dec 2008 Posts: 173	Re: 185.19 + VDPAU blit path == stall Ah, great !! Waiting for next release, then Btw, what kind of issue?

11-04-09, 04:44 PM	#43
Stephen Warren Moderator Join Date: Aug 2005 Posts: 1,327	Re: 185.19 + VDPAU blit path == stall There was a problem related to the handling of timestamps. If the timestamp for the frame was soon in the future (as opposed to further out in the future) when queued, sometimes the processing for that frame would be delayed until the requested time plus some offset, in the order of seconds.

11-24-09, 01:38 AM	#44
mlampard Registered User Join Date: Apr 2006 Location: South Australia Posts: 30	Re: 185.19 + VDPAU blit path == stall Thanks Stephen, the issue is now resolved with 195.22 drivers on my system.

11-24-09, 04:38 AM	#45
jdobry Registered User Join Date: Jan 2007 Posts: 45	Re: 185.19 + VDPAU blit path == stall @Stephen Warren: Is it fixed in 195.22 version? (http://www.nvnews.net/vbulletin/show....php?p=2129618) EDIT: looks like fixed

11-24-09, 04:00 PM	#46
crisalide Registered User Join Date: Dec 2008 Posts: 173	Re: 185.19 + VDPAU blit path == stall Indeed, seems fixed in 195.22 Stephen, thanx for your patience and hard work

11-24-09, 04:55 PM	#47
Stephen Warren Moderator Join Date: Aug 2005 Posts: 1,327	Re: 185.19 + VDPAU blit path == stall Excellent news! Thanks very much for the confirmation.