Performance Issues

[Adam Jackson]

On Thursday 16 March 2006 14:37, Roland Scheidegger wrote:
> Carsten Otto wrote:
> > On Thu, Mar 16, 2006 at 07:18:58PM +0100, Roland Scheidegger wrote:
> >> You could try oprofile to see where all the time is spent, chances are
> >> there are only very few functions where all cpu cycles are spent. That
> >> said, if those functions are somewhere inside the binary nvidia driver
> >> you're using then that's probably nvidia's problem. Maybe the game does
> >> something slightly unusual not accelerated by the nvidia driver.
> >
> > carsten-otto:~# opreport
> > CPU: AMD64 processors, speed 2574.59 MHz (estimated)
> > Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a
> > unit mask of 0x00 (No unit mask) count 100000 CPU_CLK_UNHALT...|
> >   samples|      %|
> > ------------------
> >   1134192 45.2940 dnetc
> >   1004802 40.1268 Xorg
> >         CPU_CLK_UNHALT...|
> >           samples|      %|
> >         ------------------
> >            776216 77.2506 anon (tgid:8509 range:0x957000-0x96d000)
> >            192975 19.2053 anon (tgid:8509 range:0x9aa000-0x9b8000)
> >             22072  2.1967 anon (tgid:8509 range:0x9f5000-0xad3000)
> >              6045  0.6016 Xorg
> >              5115  0.5091 anon (tgid:8509 range:0x989000-0x994000)
> >              2059  0.2049 anon (tgid:8509 range:0x978000-0x988000)
> >               205  0.0204 anon (tgid:8509 range:0x833000-0x836000)
> >                38  0.0038 anon (tgid:8509 range:0x75e000-0x763000)
> >                28  0.0028 anon (tgid:8509 range:0x997000-0x99c000)
> >                18  0.0018 anon (tgid:8509 range:0x99d000-0x9a9000)
> >                17  0.0017 anon (tgid:8509 range:0x84b000-0x886000)
> >                 7 7.0e-04 anon (tgid:8509 range:0x7f0000-0x7f2000)
> >                 7 7.0e-04 anon (tgid:8509 range:0x7f5000-0x7fa000)
> >    171987  6.8683 vmlinux
> >     64661  2.5822 libSDL-1.2.so.0.7.1
> >     30792  1.2297 openttd
> >
> > Any idea how I can give those anon things names?
>
>  From http://oprofile.sourceforge.net/doc/opreport.html:
> "Many applications, typically ones involving dynamic compilation into
> machine code, have executable mappings that are not backed by an ELF
> file. opreport has basic support for showing the samples taken in these
> regions; for example:...
> Currently, there is no support for getting symbol-based summaries for
> such regions. Note that, since such mappings are dependent upon
> individual invocations of a binary, these mappings are always listed as
> a dependent image, even when using --separate=none. Equally, the results
> are not affected by the --merge option."

He's using 6.8.2, for which elfloader falls into this class.

This is almost assuredly a problem in nvidia's driver.

- ajax
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://lists.x.org/archives/xorg/attachments/20060316/c501cc7c/attachment.pgp>