[mythtv] CPU detection/optimization

[Tom Lichti]

Daniel Kristjansson wrote:
> On Wed, 2005-03-30 at 23:57 -0500, Tom Lichti wrote:
> 
>>Daniel Kristjansson wrote:
>>
>>> time mythtv 60_second_video_file.mpg
>>
>>Not sure what that is supposed to show, other than it takes 60 seconds 
>>to play 60 seconds? Anyway, the results of that test (using a CVS build 
>>from the 23rd, one from this afternoon, both with 'pentiumpro' as the 
>>march, and one from today with 'athlon-xp' as the march) didn't show any 
>>appreciable difference doing the above test. I just catted /dev/video1 
>>to tmp.mpg for 60 seconds, and then used that as the source for the test.
> 
> 
> time gives the elapsed time, which isn't very interesting in this case,
> but it also gives the user and system time, which are..
> 
> -- Daniel
> 
> 

Ok, I didn't think there was much difference but here they are:

Last weeks build (pentiumpro):
R 1m5.295s
U 0m10.925s
S 0m2.691s

Yesterday's build (pentiumpro):
R 1m5.809s
U 0m10.999s
S 0m2.570s

Yesterday's build (athlon-xp):
R 1m6.954s
U 0m11.297s
S 0m2.879s

I only ran each test once (I was tired of watching Keanu Reeves blab to 
some chick for 60 seconds...) but I did notice on the console a TON of 
pre-buffering pauses and 'audio out of sync' messages, for all three 
runs. Not sure if that's noteworthy or not. Anyway, I'll try installing 
the athlon-xp build tonight and see how stable it is. I used all the 
flags that were suggested, including the fspmath (or whatever it was).

Tom