1 |
On 15/03/13 08:31, Mateusz Kowalczyk wrote: |
2 |
> On 14/03/13 23:52, Dale wrote: |
3 |
>> Mateusz Kowalczyk wrote: |
4 |
>>> On 14/03/13 22:41, Dale wrote: |
5 |
>>>> Grant Edwards wrote: |
6 |
>>>>> On 2013-03-14, Dale <rdalek1967@×××××.com> wrote: |
7 |
>>>>> |
8 |
>>>>> ... |
9 |
|
10 |
>>>>> RedHat maintainers aren't stupid (you can probably tell I've never |
11 |
>>>>> used RH) – they will release packages optimized for architectures |
12 |
>>>>> they will run on. Overall you might get very slight performance |
13 |
>>>>> boost because of some CFLAG you enable but you might as well have |
14 |
>>>>> worse performance because you don't know as much about |
15 |
>>>>> optimizations as the RH maintainers and developers. Bah, you can |
16 |
>>>>> even find examples on Gentoo wiki where compiling certain packages |
17 |
>>>>> with certain flags actually makes them slower and not faster where |
18 |
>>>>> usually the opposite is the case. |
19 |
Further, when we did the tests I mentioned before (exactly what Dale was |
20 |
asking about in fact) - we had 3 identical machines for testing in |
21 |
parallel ... Celerons at the time. While setting up, it became clear |
22 |
that while gentoo was working well on my P4 laptop, cloning it onto the |
23 |
celeron gave performance worse than a default i386 debian. So after a |
24 |
bit of swatting on compiler flags I tuned it closer to the architecture, |
25 |
did an overnight rebuild and we went from there ... and it could only |
26 |
"shade" i386 default debian about 10% ... mostly. |
27 |
|
28 |
1. The upshot is that I consider its actually easier to shoot yourself |
29 |
in the foot performance wise if you get it wrong than it is to get it right. |
30 |
|
31 |
2. Tuning for a particular load/job *WILL* make the machine more |
32 |
unsuitable to other load/job profiles. |
33 |
|
34 |
3. On the same hardware, any distro can/should be made to perform |
35 |
identically if tuned by someone in the know (or made worse) |
36 |
|
37 |
4. Gentoo is easier to tune (make better ... or worse :) |
38 |
|
39 |
BillK |