1 |
Hello, Alexander. |
2 |
|
3 |
You wrote: |
4 |
|
5 |
>> Keep in mind though that multiple emerges will share resources, so each |
6 |
>> process may lose speed... |
7 |
> Something that I always wondered about - does it |
8 |
> actually *lose* speed? IOW: Is it, *IN* *TOTAL* |
9 |
> slower to do multiple emerges in parallel compared |
10 |
> to doing them sequentially? |
11 |
|
12 |
Given that you have enough RAM, it may speed up things somewhat. If |
13 |
you do some cpu load monitoring (e.g., using 'top') you should see |
14 |
that cpu load during emerge process does not stay constant. There are |
15 |
99% periods (usually when compiling something), and there are 1% |
16 |
periods (sh, grep, ldconfig, emerge itself and most notably |
17 |
configure). If you are installing something large (say, KDE) you can |
18 |
get your cpu working at 100% when one instance of emerge configures |
19 |
and other does actual compilation. |
20 |
|
21 |
Mind that gcc can hog up to 150-200 megabytes of RAM per source file |
22 |
when compiling large c++ programs (e.g., kde with kdeenablefinal USE |
23 |
flag), so trying parallel emerge on systems with less than 512 megs of |
24 |
RAM space is rather pointless. |
25 |
|
26 |
-- |
27 |
WBR, |
28 |
Alexander mailto:grebenkov@×××××.com |
29 |
|
30 |
-- |
31 |
gentoo-user@g.o mailing list |