[pacman-dev] [PATCH 0/2] Remove REQUIREDBY usage

[Dan McGee]

On Nov 13, 2007 7:05 AM, Nagy Gabor <ngaba at bibl.u-szeged.hu> wrote:
> Hmm. Great job.
> I'm a bit doubter about alpm_splitdep: did you empty disk cache before retest?
> If you did so, this is really surprising to me: this is obviously a speed-up,
> but that is _much_ greater than I expected.
> Hmm.
> http://projects.archlinux.org/git/?p=pacman.git;a=commit;h=7653bb93997f52848b54ab80868cd6da52808a75
> we can spare the first strdup, if you backup *ptr then reset before return.
>
> Some other suggestions:
> 1. compute_requiredby now can return with an alpm_list of pmpkg_t* -s, and then
> we needn't scan pkgache to find 'pkgname' in deptest
> 2. -Qt needn't compute all the requiredby packages, it can stop if he find the
> first requiredby (no orphan).
> ...
> Summary: the results are very promising for me ;-)

pacman is the last pacman-git release (before any requiredby changes),
src/pacman/.libs/pacman is with the compute_requiredby only.

[root at dublin pacman]# sync; echo 3 > /proc/sys/vm/drop_caches; time
pacman -Qt > /dev/null
real    0m10.698s
user    0m0.060s
sys     0m0.160s
[root at dublin pacman]# sync; echo 3 > /proc/sys/vm/drop_caches; time
./src/pacman/.libs/lt-pacman -Qt > /dev/null
real    0m11.327s
user    0m0.607s
sys     0m0.127s

The above results tell me at least one thing. We don't have to worry
about optimizing pacman -Qt by stopping early- it isn't worth the
hack.

[root at dublin pacman]# sync; echo 3 > /proc/sys/vm/drop_caches; time
pacman -Qi pacman-git > /dev/null
real    0m0.686s
user    0m0.020s
sys     0m0.027s
[root at dublin pacman]# sync; echo 3 > /proc/sys/vm/drop_caches; time
./src/pacman/.libs/pacman -Qi pacman-git > /dev/null
real    0m10.712s
user    0m0.047s
sys     0m0.147s
[root at dublin pacman]# time ./src/pacman/.libs/pacman -Qi pacman-git > /dev/null
real    0m0.060s
user    0m0.033s
sys     0m0.023s

OK, so -Qi performance takes a big hit when nothing is cached, because
we actually have to load the whole DB to show the one requiredby line.
However, note that its performance when the kernel FS cache is
available is quite fast, so we have nothing to worry about there.

[root at dublin pacman]# sync; echo 3 > /proc/sys/vm/drop_caches; time
pacman -Su > /dev/null
real    0m4.573s
user    0m0.490s
sys     0m0.533s
[root at dublin pacman]# sync; echo 3 > /proc/sys/vm/drop_caches; time
./src/pacman/.libs/lt-pacman -Su > /dev/null
real    0m4.848s
user    0m0.487s
sys     0m0.550s

No difference here (note that this was a sysupgrade with no upgrades
available, so I don't think requiredby is even needed here).

[root at dublin pacman]# sync; echo 3 > /proc/sys/vm/drop_caches; time
yes "no" | pacman -Su > /dev/null
Proceed with installation? [Y/n]
real    0m13.979s
user    0m0.533s
sys     0m0.723s
[root at dublin pacman]# sync; echo 3 > /proc/sys/vm/drop_caches; time
yes "no" | ./src/pacman/.libs/lt-pacman -Su > /dev/null
Proceed with installation? [Y/n]
real    0m14.467s
user    0m0.507s
sys     0m0.623s

No huge difference here (sysupgrade with one upgrade available- qt3).

If anyone else can think of places where this switch to
compute_requiredby code would really hurt us, I'd be glad to hear it.

-Dan