[mythtv-users] 0.19, XvMC, and Prebuffer pause

[Brian Wood]

On Mar 1, 2006, at 1:42 PM, Dylan R. Semler wrote:

> <http://www.mythtv.org/wiki/index.php/PCI_Latency>
>
>>> I found my IDE controller latency was set to zero, so I reset it to
>>> 32 (20hex).  I'm not at home to try playback right now, but I was
>>> wondering what others who are having this problem find for their
>>> IDE/SATA controller latency.  Maybe this could fix us up?!?!  I'll
>>> report back later tonight what I find out.
>>>
>>
>> So far, with a very limited number of samples, the people who have
>> found their disk controllers set to zero have either been unable to
>> reset the latency value, or have not seen any positive results from
>> doing so.
>>
>> Only the machines which have drive controller latency set to 32 or 64
>> initially have had much luck in adjusting them.
>>
>> Again, this is from a *very* limited number of samples (6 to be  
>> exact)
>> so please report your results, positive or negative, as well as the
>> chipset your mobo is using (perhaps on the WiKi page ??)
>
> I just tried changing the latency on my three machines.  I have two  
> Dell
> 8200 desktops from about 2002 and a Sager laptop from December  
> 2005.  In
> all three machines, the initial IDE latency was 0 and changing it with
>
> setpci -v -s 00:1f.1 latency_timer=b0 (or any other value)
>
> did not seem to do anything.  At least lspci -v continued to display
> their latency as 0 (I have not been able to test this with Myth yet).
> Furthermore, I found that if tried this command on ANY controller  
> whose
> intial latency was 0, the command did nothing.  If the initial latency
> was not zero, the command would change the latency to the desired
> value.  I Also tested that if a controller ititially had a non-zero
> latency, and I changed it to zero, I was still able to change it  
> back to
> other values.  I have no idea why this is or what can be done, but I
> thought I'd add some more datapoints to the set.  If anyone wants the
> maker of these mobos, I can probably find that out somehow.
>

That matches out what I have recorded. The one incident of somebody  
setting a zero latency to something else could not be repeated and I  
take it as bad data.

I think by spec latency does not have to be settable unless the  
device is capable of bus mastering and bursting two or more shots  
over the bus, but I can't imagine a drive controller that wouldn't  
want to be able to do that. The latency might just be fixed at some  
unknown value and not readable by lspci for some reason.

I also don't know if the "zero" is "real" or an artifact of reading  
it. setpci does not report an error so it seems to me that either  
lspci or setpci is not behaving well. I just can't see an HDD  
controler wanting to "really" have a zero latency value, perhaps the  
designers figure that mouse movement is more important data to move  
on the bus ??

("Windows has detected that the mouse has moved, please re-boot  
immediately")

It wouldn't be the first time that manufacturers (especially Intel)  
have ignored specs. Maybe they figure they have a right to do that  
since they were in on the initial PCI design ??

I was digging through my junk drawer yesterday and came upon an old  
"VESA" bus video card, anybody want a really good deal :-) 256K RAM  
(that's "K", not "M").

I'm still trying to learn more and accumulate more data, so much  
thanks for the info.