1 |
On Tue, 24 Feb 2009 20:28:43 +0100 |
2 |
Alexis Ballier <aballier@g.o> wrote: |
3 |
> On Tue, 24 Feb 2009 18:24:16 +0000 |
4 |
> Ciaran McCreesh <ciaran.mccreesh@××××××××××.com> wrote: |
5 |
> > On Tue, 24 Feb 2009 18:16:54 +0100 |
6 |
> > Luca Barbato <lu_zero@g.o> wrote: |
7 |
> > > > You're doubling the number of files that have to be read for an |
8 |
> > > > operation that's almost purely i/o bound. On top of that, you're |
9 |
> > > > introducing a whole bunch of disk seeks in what's otherwise a |
10 |
> > > > nice linear operation. |
11 |
> > > |
12 |
> > > I see words, not numbers. |
13 |
> > |
14 |
> > Number: double. That's a '2 times'. |
15 |
> |
16 |
> That only means you're increasing the constant factor in the |
17 |
> complexity of the thing... which may very well be completely |
18 |
> negligible unless someone provides real benchmarks. |
19 |
|
20 |
In the most common case where metadata is valid, around half of the |
21 |
time it takes Paludis to work out a resolution set is spent grabbing |
22 |
metadata for ebuilds. |
23 |
|
24 |
> I would be very surprised if that "2 times" factor happens to be true, |
25 |
> because finding a string in a file is an order of magnitude simpler |
26 |
> than sourcing said file with bash. Moreover this doesn't take into |
27 |
> account disk and os cache. |
28 |
|
29 |
No no no. *Opening* the file is the slow part, not searching. The file |
30 |
wouldn't otherwise be opened at all. |
31 |
|
32 |
-- |
33 |
Ciaran McCreesh |