1 |
On Wed, Jun 24, 2015 at 01:13:40PM +0200, Alan McKinnon wrote: |
2 |
> On 23/06/2015 15:05, behrouz khosravi wrote: |
3 |
> > Hello everyone. |
4 |
> > |
5 |
> > I really like to have control over my machine as much as possible. In |
6 |
> > this way I will learn a lot, so I am trying to remove all the default |
7 |
> > use flags and control them manually. |
8 |
> |
9 |
> |
10 |
> Here's some good advice: |
11 |
> |
12 |
> Don't do that. See below. |
13 |
> |
14 |
|
15 |
Nonsense - do that. If your goal is to learn how stuff works and you're |
16 |
already reasonably familiar with C/C++ so you can debug any strange |
17 |
errors that can happen, have fun. Just don't think you'll get any real |
18 |
work done ;). i.e. it might be good to do this in a virtual machine and |
19 |
still have a stable system for work. |
20 |
|
21 |
> |
22 |
> > I just don't know which "global" use flags are absolutely necessary to |
23 |
> > the system to make it snappier or secure. |
24 |
> |
25 |
> That's a bit of a nonsensical line of thought, as what you think you |
26 |
> want doesn't really exist. |
27 |
> |
28 |
> ... |
29 |
> |
30 |
> Put "-march=native" in CFLAGS |
31 |
> |
32 |
|
33 |
Yes. Also, properly setting CPU_FLAGS_X86 is another thing that can |
34 |
speed up software *if* said software supports any special instruction |
35 |
sets. Most "normal desktop software" like web browsers, email clients, |
36 |
terminals, editors, etc. probably will not get a whole lot of benefit |
37 |
either way, since most of this software is generally not CPU-bound and |
38 |
is instead network/disk bound. |
39 |
|
40 |
In the mornings I primarily use my desktop for reading email and |
41 |
browsing news with firefox (mostly on sites with minimal JavaScript), |
42 |
and I have yet to see my load averages climb higher than maybe 0.5. |
43 |
|
44 |
Any software that does anything requiring lots of math will get a boost |
45 |
from this type of stuff, though; graphics editing, most things in sci-* |
46 |
categories, audio/video transcoding, etc. |
47 |
|
48 |
Alec |
49 |
|
50 |
P.S. Just realized I don't have "-march=native" in my CFLAGS. Time to |
51 |
rice - could be getting 1% better performance. ;) |
52 |
|
53 |
P.P.S. Also, on 1% better performance: My professor for the compilers |
54 |
class I took used to (maybe still does) work at Google. Apparently |
55 |
Google sees a <1% increase in performance as *the best thing ever*, |
56 |
because it can save them a bunch of money in infrastructure and power. |
57 |
Apparently Google are the ultimate ricers. |