Author |
Message |
jaschris
|
Posted: Sat Oct 06, 2007 12:12 am |
|
Joined: Sat Mar 19, 2005 1:24 pm
Posts: 10
|
I was hoping that would work for both of us, mine does _seem_ better, but i still get some pauses.
Try
Code: # top | grep rrd
or just "top" and watch for the below processes, but sometimes they are just a blip and the above is easier to watch
Many of my pauses seem to happen at the same time as these monitoring agents. Maybe causing some interrupts, you see anything similar?
output:
20899 root 29 0 10972 4516 3028 R 6.7 0.4 0:00.20 rrd_Disk.pl
20925 root 23 0 4716 2192 1516 S 0.3 0.2 0:00.01 rrd_MBtemp.pl
20929 root 23 0 4720 2196 1516 S 0.3 0.2 0:00.01 rrd_MBfan.pl
20929 root 19 0 5132 3072 2208 R 2.3 0.3 0:00.08 rrd_MBfan.pl
20940 root 22 0 4716 2196 1516 S 0.3 0.2 0:00.01 rrd_Load.pl
20964 root 24 0 5084 2788 2044 R 0.7 0.3 0:00.02 rrd_MythFS.pl
20969 root 18 0 4720 2192 1516 S 0.3 0.2 0:00.01 rrd_MBtemp.pl
20973 root 21 0 4720 2196 1516 S 0.3 0.2 0:00.01 rrd_MBfan.pl
20973 root 29 0 5132 3044 2180 R 5.3 0.3 0:00.17 rrd_MBfan.pl
|
|
Top |
|
 |
grante
|
Posted: Sat Oct 06, 2007 9:21 pm |
|
Joined: Mon Jun 27, 2005 4:42 pm
Posts: 321
Location:
Minneapolis, Minnesota, USA
|
jaschris wrote: I was hoping that would work for both of us, mine does _seem_ better, but i still get some pauses. Try Code: # top | grep rrd or just "top" and watch for the below processes, but sometimes they are just a blip and the above is easier to watch
I don't have any of those processes running at all.
_________________ Grant
|
|
Top |
|
 |
jaschris
|
Posted: Tue Oct 09, 2007 1:32 pm |
|
Joined: Sat Mar 19, 2005 1:24 pm
Posts: 10
|
I don't have them running all the time either.
But when they did kick off every 5 -10 minutes (and ran for a few seconds) is when I also saw additional pauses. Commenting them out has made the myth box as stable/smooth as previous releases of knoppmyth.
Now if only my ipod transcodings wouldn't take 2 hours per recorded hour to perform

|
|
Top |
|
 |
grante
|
Posted: Tue Oct 09, 2007 2:57 pm |
|
Joined: Mon Jun 27, 2005 4:42 pm
Posts: 321
Location:
Minneapolis, Minnesota, USA
|
jaschris wrote: I don't have them running all the time either.
But when they did kick off every 5 -10 minutes (and ran for a few seconds) is when I also saw additional pauses. Commenting them out has made the myth box as stable/smooth as previous releases of knoppmyth.
Commenting them out where?
_________________ Grant
|
|
Top |
|
 |
tkoster
|
Posted: Tue Oct 09, 2007 5:39 pm |
|
Joined: Mon Apr 04, 2005 10:50 am
Posts: 120
|
This may be a "no brainer" question, but have you tried running myth without XvMC?
Switching to either of the other options (standard, libmpeg2) improved things immensely for me and and other than your use of a pci slot, you seem to have more juice in your box than I do.
With earlier editions XvMC was a must, but that hasn't been the case lately because of other improvements in myth. So if you haven't yet, give it a try. If you have, pardon my intrusion.
|
|
Top |
|
 |
grante
|
Posted: Tue Oct 09, 2007 9:48 pm |
|
Joined: Mon Jun 27, 2005 4:42 pm
Posts: 321
Location:
Minneapolis, Minnesota, USA
|
tkoster wrote: This may be a "no brainer" question, but have you tried running myth without XvMC? Yes. I thought I mentioned it earlier: without xvmc CPU usage is 100% and stuttering is pretty bad (both video and audio). Quote: Switching to either of the other options (standard, libmpeg2) improved things immensely for me and and other than your use of a pci slot, you seem to have more juice in your box than I do. [Aargh. I just realized I've been claiming the box has a 3GHz P4. It's only a 2.4GHz P3 -- I'd been conflating two machines.] Neither is anywhere close to watchable for 720p or 1080i. Quote: With earlier editions XvMC was a must, but that hasn't been the case lately because of other improvements in myth. So if you haven't yet, give it a try. If you have, pardon my intrusion.
Here's another datapoint: Watching an SD recording uses 15-20%
CPU while it's _paused_. IT seems that the OSD uses up large
amounts of CPU time even when both the OSD and the video are
completely static.
I think I'm going to order an Asus M2NPV-VM tomorrow. I'm
hoping I'm going to be able to use a 120W pico-power supply:
http://www.mini-box.com/picoPSU-120-pow ... ategory=13
http://www.silentpcreview.com/article601-page1.html
I've also got to figure out what to do for an IR receiver. I
bought a usb-uirt, but its sensitive is so bad it's not really
usable. I had to take the dark red lens out to get it to work
at all when the remote was more than a few inches away. Without
the lens, it's only good out to about 8ft and then only when
you stay withing about 10% of the device's centerline. The
receive that came with the (removed) PVR350 was _way_ more
senstive.
_________________ Grant
|
|
Top |
|
 |
pilesofspam
|
Posted: Wed Oct 10, 2007 8:19 am |
|
Joined: Tue May 09, 2006 11:50 am
Posts: 102
|
I can personally recommend the homebrew receiver. I built one forever ago, connect it to a com port and it works well out of the box.
I'm going to try going from my current 2200+ to a 2800+ that's on its way. Its so close to working right it hurts.
|
|
Top |
|
 |
grante
|
Posted: Wed Oct 10, 2007 2:40 pm |
|
Joined: Mon Jun 27, 2005 4:42 pm
Posts: 321
Location:
Minneapolis, Minnesota, USA
|
pilesofspam wrote: I can personally recommend the homebrew receiver. I built one forever ago, connect it to a com port and it works well out of the box. Since serial ports are getting pretty rare these days, I thought I'd go with a USB connected device. I like to support garage operations like USB-UIRT and it seemed to get positive comments from others. Maybe I just got a lemon. I've posted my experience to the USB-UIRT troubleshooting forum -- perhaps somebody knows how to fix it. Quote: I'm going to try going from my current 2200+ to a 2800+ that's on its way. Its so close to working right it hurts.
I know what you mean. For me, the actual playback works fine,
it's the OSD that brings the system to it's knees. I'm
convinced there's a bug in myth's OSD code, but I don't
realistically have a snoball's chance in hell of finding/fixing
it. My brain and C++ aren't compatible.
_________________ Grant
|
|
Top |
|
 |
tjc
|
Posted: Wed Oct 10, 2007 6:31 pm |
|
Joined: Thu Mar 25, 2004 11:00 am
Posts: 9551
Location:
Arlington, MA
|
grante wrote: Here's another datapoint: Watching an SD recording uses 15-20% CPU while it's _paused_. IT seems that the OSD uses up large amounts of CPU time even when both the OSD and the video are completely static.
Presumably this is without XvMC right? This thread got me thinking and I did some experiments along these lines last night.
"Standard" MPEG2 decoder watching SD shows:
~16-17% CPU for mythfrontend and ~2.7-3% CPU for Xorg
Paused SD shows:
~1.3-2% CPU for mythfrontend and ~2-2.7% CPU for Xorg
"Standard XvMC" watching SD shows:
~12% CPU for mythfrontend and <2% CPU for Xorg
Paused SD shows:
~0% CPU for mythfrontend and ~0% CPU for Xorg
The XvMC number don't change noticeably whether I'm using the ChromaKeyOSD or not. (On screen artifacts are noticeably different.)
So why are my results so different? What is your output resolution? What version of the nVidia drivers are you using? Any other special GUI settings?
|
|
Top |
|
 |
grante
|
Posted: Wed Oct 10, 2007 7:22 pm |
|
Joined: Mon Jun 27, 2005 4:42 pm
Posts: 321
Location:
Minneapolis, Minnesota, USA
|
tjc wrote: grante wrote: Here's another datapoint: Watching an SD recording uses 15-20% CPU while it's _paused_. IT seems that the OSD uses up large amounts of CPU time even when both the OSD and the video are completely static. Presumably this is without XvMC right? No, that's with XvMC. Quote: This thread got me thinking and I did some experiments along these lines last night.
"Standard" MPEG2 decoder watching SD shows: ~16-17% CPU for mythfrontend and ~2.7-3% CPU for Xorg
Paused SD shows: ~1.3-2% CPU for mythfrontend and ~2-2.7% CPU for Xorg
"Standard XvMC" watching SD shows: ~12% CPU for mythfrontend and <2% CPU for Xorg
Paused SD shows: ~0% CPU for mythfrontend and ~0% CPU for Xorg
The XvMC number don't change noticeably whether I'm using the ChromaKeyOSD or not. (On screen artifacts are noticeably different.)
So why are my results so different? What is your output resolution? 640x480 Quote: What version of the nVidia drivers are you using? 1.0.9755-1 I also tried 1.0.7185 and 1.0.9631, with no noticable differences. Quote: Any other special GUI settings?
I tried various "sync" options (in both nvidia-settings,
xorg.conf, and the MythTV playback settings). None made any
difference. I also tried enabling UseEvents. That made things
worse (often got a completely black screen). Setting
XvmcUsesTextures to "false" improves OSD-related problems, but
screws up the horizontal scaling such that only the left-hand
2/3 of the picture is on-screen.
I think things are pretty much "stock" at the moment except for
XvmcUsesTextures is set to "true".
_________________ Grant
|
|
Top |
|
 |
tjc
|
Posted: Wed Oct 10, 2007 9:48 pm |
|
Joined: Thu Mar 25, 2004 11:00 am
Posts: 9551
Location:
Arlington, MA
|
Bizarre.
Are you using any of the OpenGL drawing stuff?
|
|
Top |
|
 |
grante
|
Posted: Thu Oct 11, 2007 9:49 am |
|
Joined: Mon Jun 27, 2005 4:42 pm
Posts: 321
Location:
Minneapolis, Minnesota, USA
|
tjc wrote: Are you using any of the OpenGL drawing stuff?
I tried it, and it didn't work at all. On the menu screens I'd get nothing but
a blank background.
_________________ Grant
|
|
Top |
|
 |
tjc
|
Posted: Thu Oct 11, 2007 7:42 pm |
|
Joined: Thu Mar 25, 2004 11:00 am
Posts: 9551
Location:
Arlington, MA
|
Dump your settings table and PM me the output along with your xorg.conf and /home/mythtv/.nvidia-settings-rc file.
Code: mysql mythconverg -e "select * from settings"
|
|
Top |
|
 |
grante
|
Posted: Fri Oct 12, 2007 11:17 am |
|
Joined: Mon Jun 27, 2005 4:42 pm
Posts: 321
Location:
Minneapolis, Minnesota, USA
|
tjc wrote: Dump your settings table and PM me the output along with your xorg.conf and /home/mythtv/.nvidia-settings-rc file. Code: mysql mythconverg -e "select * from settings"
Will do.
_________________ Grant
|
|
Top |
|
 |
grante
|
Posted: Sat Oct 13, 2007 8:31 pm |
|
Joined: Mon Jun 27, 2005 4:42 pm
Posts: 321
Location:
Minneapolis, Minnesota, USA
|
I just re-ran some CPU measurements. All measurements were
done playing back the same SD ATSC recording.
Code: Without XVMC With XVMC
FE Xorg FE Xorg Normal playback 25% 20% 26% 1% Playback w/ menu 26% 20% 80-95% 4-5% Paused w/ OSD 15% 3% 27% <1% Paused w/o OSD 15% <1% <1% <1%
Without XVMC, an OSD produces a repeatable 1-2% CPU load.
With XVMC, the OSD load varies from 27-95%. XVMC playback with
the menu displayed will periodically hit 100% CPU usage and
produce stuttering.
HD playback without XVMC pegs CPU at 100% and stutters.
HD playback with XVMC runs OK at 60-80% CPU usage, but
collapses completely when the OSD is up (100% CPU usage, 2-3
fps video, audio sometimes quites completely -- even after OSD
is gone and video playback is normal).
The machine is a Shuttle Zen ST62K
CPU: 2.4GHz P4
RAM: 512MB of PC3200 RAM clocked at 400MHz.
Chipset: ATI 9100 IGP (RS300 NBridge, IXP150 SBridge)
Video: NVidia FX5500 128MB in a 66MHz 32-bit PCI slot
Video output is S-Video running at 640x480 60Hz.
_________________ Grant
|
|
Top |
|
 |