1 |
I'd agree with Miguel. In general, I can't tell the difference in most |
2 |
applications. I did some tests on OpenOffice a few years ago and found the 8 |
3 |
hour compile time wasn't worth it in terms of performance. Where Gentoo |
4 |
shines is in giving you the increased ability to shrink your installed size. |
5 |
I have nothing on this machine that I don't specifically want there and I |
6 |
work overly hard to keep my compile flags exactly right. |
7 |
|
8 |
That said, I'd be interested in seeing your results. It would be nice to |
9 |
know that all of this work is beneficial for more than just keeping me |
10 |
happy. |
11 |
|
12 |
Mike |
13 |
|
14 |
On 4/30/07, Miguel Sousa Filipe <miguel.filipe@×××××.com> wrote: |
15 |
> |
16 |
> Hi there, |
17 |
> |
18 |
> On 4/29/07, Peter A. H. Peterson <pedro@×××××××××××.net> wrote: |
19 |
> > Hi Everyone, |
20 |
> > |
21 |
> > My name is Peter Peterson and I represent a group of a grad students |
22 |
> > at UCLA. We're in a computer systems performance analysis course and we |
23 |
> > were hoping to do a general performance comparison of gentoo vs. a |
24 |
> > popular binary i386-compatible distribution (probably ubuntu) in some |
25 |
> > "real-world" server tests to try and meaningfully calculate the |
26 |
> > performance gains that local compilation provides. (For example, |
27 |
> > apache2 requests processed per second on the same hardware.) |
28 |
> > |
29 |
> > I've subscribed to this list because we want the gentoo community to |
30 |
> > be involved in helping us design the tests so that we can hopefully |
31 |
> > all feel good about what and how we are testing the systems. |
32 |
> > |
33 |
> > We have no particular outcome in mind; our group represents a wide |
34 |
> > range of computer users, from Mac, Linux, and Windows enthusiasts, and |
35 |
> > we have all used a wide variety of Linux distributions. We have simply |
36 |
> > noticed that much of the discussion of gentoo's performance advantage |
37 |
> > is anecdotal and we're genuinely hoping to provide some meaningful |
38 |
> > experimental data for discussion. Also, if anyone knows of any |
39 |
> > available benchmark data or papers on this subject, we'd love to hear |
40 |
> > about them. There was apparently a paper on slashdot a couple of years |
41 |
> > ago, but the host it was on appears to now be squatted. For that |
42 |
> > matter, if this is a well understood or closed issue (for example, if |
43 |
> > the statistics that people quote are actually from good experimental |
44 |
> > data) please let us know. |
45 |
> > |
46 |
> > Is anyone here interested in discussing this project? We are |
47 |
> > specifically interested in discussing methodology, testing suits, |
48 |
> > CFLAGS and other options. Our desire is not to "trick out" gentoo or |
49 |
> > ubuntu, but rather quantify the performance benefit that gentoo has |
50 |
> > over binary distributions with "normal" compile flags (whatever normal |
51 |
> > is). |
52 |
> > |
53 |
> |
54 |
> A good CFLAGS would be something not very agressive, something like: |
55 |
> -march=<cpuType> -O3 or -O2 and at most -fomit-frame-pointer. |
56 |
> (Scientific workloads can speedup considerably with: -ffast-math) |
57 |
> |
58 |
> Having experienced and done some benchmarks with gentoo and other |
59 |
> distros on servers and on scientific workstations. |
60 |
> What I found is that sometimes gentoo lacks critical performance |
61 |
> patches in glibc that are applied to mainstream distros (redhat, |
62 |
> suse..etc) that provide boosts in memcpy, memset, etc..(I remmember a |
63 |
> discussion about that some years ago). |
64 |
> What I also found out is that the compiler flags only affect |
65 |
> workloads that are very compute intensive. not something that depends |
66 |
> almost completely on FSB load or IO load.. like most server |
67 |
> workloads... -O3 doesn't do much to a working set full of |
68 |
> unpredictable branches (like server workloads usually are) and low IPC |
69 |
> rate. |
70 |
> |
71 |
> I really do believe performance boost from gentoo to be practically |
72 |
> negligible. The difference will only be apreciable in very few corner |
73 |
> cases. Most distros also optimize critical aplications such has: |
74 |
> openssl, mplayer.. reducing the possible corner cases. |
75 |
> |
76 |
> Anyways, doing a "academic" benchmark would be a good idea. |
77 |
> |
78 |
> Something like: |
79 |
> micro-benchmarks: |
80 |
> - stream (mem bandwith benchmark) |
81 |
> - ?? |
82 |
> |
83 |
> macro-benchmarks: |
84 |
> - apache2 + gzip + php(make it cpu intensive, not IO intensive) |
85 |
> - xmlmark ? |
86 |
> - kernelbench |
87 |
> - pybench ? |
88 |
> - openssl bench |
89 |
> |
90 |
> |
91 |
> about methodology: |
92 |
> - same system, same bios version, same disks. |
93 |
> - All OSes must be installed in the same disk partitions. |
94 |
> - the will be trouble about the kernel config: |
95 |
> - for mainstream distros you should use the kernel that is provided. |
96 |
> - for gentoo, gentoo-sources configured by someone which is |
97 |
> experienced, and informed about configuration impacts on performance |
98 |
> (ideally a kernel hacker?). |
99 |
> |
100 |
> - should use the stable versions in gentoo portage? |
101 |
> - or should use the same application versions used on mainstream distro? |
102 |
> |
103 |
> |
104 |
> |
105 |
> -- |
106 |
> Miguel Sousa Filipe |
107 |
> -- |
108 |
> gentoo-performance@g.o mailing list |
109 |
> |
110 |
> |