| 1 |
On Sat, 08 Nov 2014 15:45:30 -0800 Zac Medico wrote: |
| 2 |
> On 11/08/2014 03:33 PM, Patrick Lauer wrote: |
| 3 |
> > We can choose for "code that works reasonably fast" - portage hasn't |
| 4 |
> > gotten any noticeable work on performance in a while, and people have |
| 5 |
> > just piled on more and more features and complexity. |
| 6 |
> |
| 7 |
> Yes, as one of only 2 people who have worked on the resolver in recent |
| 8 |
> history, I agree with your statement. There have been lots of features |
| 9 |
> added (including EAPI 5 sub-slot rebuilds), with not much thought to |
| 10 |
> performance tuning. It will be interesting to do some profiling and see |
| 11 |
> what we can improve. |
| 12 |
> |
| 13 |
> > There's no reason that it takes a minute to start up, |
| 14 |
> |
| 15 |
> If it takes a minute then it probably means that |
| 16 |
> /var/cache/edb/vdb_metadata.pickle got deleted. In that case, it has to |
| 17 |
> reload lots of metadata from /var/db/pkg/*/*/*. |
| 18 |
|
| 19 |
On old hardware it may take dozens of minutes of CPU time. I have |
| 20 |
that *.pickle files, I have sqlite metadata cache, I have 100% CPU |
| 21 |
usage. It's not an I/O problem. Just take into account that due to |
| 22 |
instruction set, Lx cache, frequency and memory speed difference |
| 23 |
old CPU-based system may be 20x times slower than recent one. |
| 24 |
|
| 25 |
Best regards, |
| 26 |
Andrew Savchenko |
Archives