On Tue, 2008-03-18 at 12:11 +0100, Tim Lauridsen wrote: > James Antill wrote: > > So while I should have been asleep I did this, please don't look if > > you are feeling queasy, and I'm pretty sure it still works the same. > > > > Basic idea behind it is that we pre-load the entire requires DB into a > > python hash+arrays, as we are losing out when we call .getRequires() > > 10,000s of times and do an SQL call a lot of those times. > > > > For me, this takes a "yum DEV update" from ~160s to ~60s. As I implied > > above it needs more work before we can check it in (probably kills > > performance for the upgrade one package case) and generally needs to be > > tided up. But what does everyone think, is there something else we could > > be doing instead ... does this give you some other great idea? > >
> The patch looks fine to me, but what is the memory impact of the patch ? > ditto what time said. The patch looks fine but This feels like it'll be huge in memory even with some of the data being shared. -sv _______________________________________________ Yum-devel mailing list [email protected] https://lists.dulug.duke.edu/mailman/listinfo/yum-devel
