Re: [xine-devel] xine performance

[Michael R. <mr...@us...>]

Hi Robin,

> > > Using dvdnav 0.9.10/libdvdnav 0.1.1 with xine 0.9.13 fixes the
> > > performance problems. Disabling the read ahead cache in dvdnav
> > > 0.9.13/libdvdnav 0.1.3 reduces the performance problems but it
> > > isn't as good as dvdnav 0.9.10. As an aside, build_mrl_list
> > > segfaults when using either version of dvdnav with xine 0.9.13.
> > > Using dvdnav 0.9.10 with xine 0.9.10 doesn't cause a segfault on
> > > my machine but it did for some people.
> >
> > Could you try the intermediate versions, too and see, which is the
> > most recent combination of versions that still works for you?
>
> The problem seems to have been introduced in 0.9.13 as 0.9.12 works
> fine. Personally I think 0.9.10 works better than 0.9.12, but that
> may be an inaccurate perception.

Just to summarize:
xine-lib 0.9.13 + xine-dvdnav 0.9.13 + libdvdnav 0.1.3 is bad
xine-lib 0.9.12 + xine-dvdnav 0.9.12 + libdvdnav 0.1.2 is ok

I looked over the ChangeLogs of xine-lib and libdvdnav, but there is 
nothing suspicious in there that could cause a performance loss. At 
least I can't see anything. (xine-dvdnav was nearly unchanged and is 
anyway only a thin conversion layer between libdvdnav and xine)

I am a bit out of ideas here, especially since performance did not 
change much for me on IA32. (It slightly improved.)

My only suggestion would be to try:
xine-lib 0.9.13 + xine-dvdnav 0.9.12 + libdvdnav 0.1.2
xine-lib 0.9.12 + xine-dvdnav 0.9.13 + libdvdnav 0.1.3
to see if xine or libdvdnav causes the problems. But doing so you will 
have problems with plugin interface versions. If you still want to try, 
you can change the iface version test in xine-dvdnav (around line 1310) 
to match the needed number, since the interface did not change.

Michael

-- 
prom_printf("Detected PenguinPages, getting out of here.\n");
	2.0.38 /usr/src/linux/arch/sparc/mm/srmmu.c