On Wed, 15 Apr 2009, Ciaran McCreesh wrote:

> On Tue, 14 Apr 2009 14:15:09 -0400 (EDT)
> "Benjamin R. Haskell" <[email protected]> wrote:
> 
> > So, RMD160's take about 1m30s.  SHA1's take about 1:10.  But, I'd 
> > expect computing both in a row on the same files would have a lower 
> > total time, thanks to disk cache.  Are there other hashes that get 
> > computed?
> 
> We do whatever's in the Manifest.
> 
> You might want to try c4954637fe7bb3 and see how much of a difference it 
> makes for you. In some situations it might reduce the time required by a 
> large amount.


At this point, I have ~250 packages left to update/install. c4954637fe7bb3 
reduced the checking phase by about 25%.

c4954637fe7bb3 took 7min 40s
0.36.0 took 10min 22s

The important ones for showing the difference memoization made were the 
various QT 4 packages, 9 of which use the same tar.bz2.  Using the same 
basic commandline from last time (and the same data [from the full 328 
packages]), but removing one of the 'uniq's:

332  987.6M   59.4M  (4x) /usr/portage/distfiles/gdb-6.8.tar.bz2
333    1.0G   62.7M  (4x) /usr/portage/distfiles/glibc-2.9-20081201.tar.bz2
334    1.1G   65.2M  (2x) /usr/portage/distfiles/xulrunner-1.9.0.8.tar.bz2
335    1.2G   76.4M  (1x) /usr/portage/distfiles/jdk-6u13-dlj-linux-i586.bin
336    1.3G  148.6M  (1x) 
/usr/portage/distfiles/OOo_3.0.0_LinuxIntel_install_en-US.tar.gz
337    2.3G  972.8M  (9x) 
/usr/portage/distfiles/qt-x11-opensource-src-4.5.0.tar.bz2

So paludis-0.36.0 would've been doing about twice as much work as the 
newer version.  (old total was about ~1.1GiB).

Thanks,
Ben
_______________________________________________
paludis-user mailing list
[email protected]
http://lists.pioto.org/mailman/listinfo/paludis-user

Reply via email to