On Mon, Jun 02, 2014 at 01:07:43PM -0400, James Antill wrote:
> On Sun, 2014-06-01 at 18:03 +0200, Lars Ellenberg wrote:
> > On Fri, May 30, 2014 at 05:18:19PM -0400, James Antill wrote:
> > > On Fri, 2014-05-30 at 16:46 +0200, Lars Ellenberg wrote:
> > > > Hi there.
> > > > (Cc me, I'm not subscribed)
> > > >
> > > > I've got a "yum check" taking two hours.
> > > > (really single core CPU time, actually, not much IO going on there,
> > > > afaics)
> > >
> > > That is not normal. Whenever I've seen this kind of problem it's always
> > > been a problem in the rpm indexes, rpm --rebuilddb might help.
> >
> > Did that.
> > Does not help.
>
> Ok, please try the attached patch[1]. It directly asks the rpm indexes
> for an answer, and doesn't bother doing any of the yum side stuff as we
> already know the answer.
>
>
> [1] It's measurably faster here, for normal/sane pkgs. But then it's
> doing a lot less :).
>
I added timing for individual "provides" in the inner loop
of check_provides(); second column is number of calls to str_eq
(which is 0 obviously for the "new" style),
third value is the time it took (seconds.fractions),
"speedup" is "T_new/T_old".
The "interesting" part of two runs of yum check provides,
interleaved,
both interrupted once they reached the "kernel" provides.
config(jpackage-utils) 3 0.000266790
config(jpackage-utils) 0 0.000150919 1.77 x speedup
java-1.6.0-openjdk 177 0.001138930
java-1.6.0-openjdk 0 0.000332117 3.43 x speedup
java-1.6.0-openjdk(x86-64) 178 0.001145120
java-1.6.0-openjdk(x86-64) 0 0.000360966 3.17 x speedup
java_cup 7 0.000203848
java_cup 0 0.000096083 2.12 x speedup
java_cup(x86-64) 8 0.000218153
java_cup(x86-64) 0 0.000105858 2.06 x speedup
java_cup-0.10k.jar.so()(64bit) 4 0.000185966
java_cup-0.10k.jar.so()(64bit) 0 0.000099897 1.86 x speedup
java_cup-runtime-0.10k.jar.so()(64bit) 4 0.000209808
java_cup-runtime-0.10k.jar.so()(64bit) 0 0.000111103 1.89 x speedup
jline 2 0.000190973
jline 0 0.000101089 1.89 x speedup
jpackage-utils 4 0.000254869
jpackage-utils 0 0.000148058 1.72 x speedup
kbd 3 0.000252008
kbd 0 0.000157118 1.60 x speedup
kbd(x86-64) 4 0.000255108
kbd(x86-64) 0 0.000152111 1.68 x speedup
kbd-misc 2 0.000576973
kbd-misc 0 0.000462055 1.25 x speedup
kernel 64770 1.586930000
kernel 0 0.015713000 100.99 x speedup
kernel(DSA_verify) 64806 0.392415000
kernel(DSA_verify) 0 0.014446000 27.16 x speedup
kernel(FsmChangeState) 64812 0.389691000
kernel(FsmChangeState) 0 0.014361100 27.14 x speedup
kernel(FsmDelTimer) 64818 0.385489000
kernel(FsmDelTimer) 0 0.014438200 26.70 x speedup
kernel(FsmEvent) 64824 0.394106000
kernel(FsmEvent) 0 0.014229100 27.70 x speedup
kernel(FsmFree) 64830 0.389760000
kernel(FsmFree) 0 0.014483000 26.91 x speedup
kernel-drm 64779 0.390668000
kernel-drm 0 0.014301100 27.32 x speedup
kernel-drm-nouveau 64785 0.388427000
kernel-drm-nouveau 0 0.014469100 26.85 x speedup
kernel-modeset 64791 0.387349000
kernel-modeset 0 0.014324900 27.04 x speedup
kernel-uname-r 64797 0.385177000
kernel-uname-r 0 0.014467000 26.62 x speedup
kernel-x86_64 64773 0.407443000
kernel-x86_64 0 0.014701800 27.71 x speedup
kernel-xen 32280 0.195229000
kernel-xen 0 0.014184000 13.76 x speedup
So yes, that looks like a nice improvement even for packages with
a "short" provides list, and a big improvement for packages with
long provides lists.
If it's functional equivalent: please use!
I'll time a full "yum check" later,
and then an other with cProfile enabled, just for fun.
Thanks,
Lars
_______________________________________________
Yum-devel mailing list
Yum-devel@lists.baseurl.org
http://lists.baseurl.org/mailman/listinfo/yum-devel
