1 |
Harry Putnam posted <ubqzt9dzh.fsf@×××××××.com>, excerpted below, on Wed, |
2 |
07 Dec 2005 08:30:10 -0600: |
3 |
|
4 |
> I'm about to install gentoo on an athlon64. Is there enough 64 bit |
5 |
> software and other good reasons to use the 64 bit version? |
6 |
> |
7 |
> Although a long time linux user, I'm not particularly skilled at dealing |
8 |
> with problems (a slow learning or just thick headed I guess). |
9 |
> |
10 |
> But have cross posted this to hear from both 32 and 64 bit advocates. |
11 |
|
12 |
Thanks for mentioning the cross-posting. Few enough have the courtesy to |
13 |
do so. |
14 |
|
15 |
For the most part, the Free, Libre, and Open Source Software (FLOSS) side, |
16 |
certainly for the commonly used stuff, has long ago been ported, and will |
17 |
present little or no issues related to 64-bit. |
18 |
|
19 |
The problems, where they exist, therefore almost all relate to closed |
20 |
source unfree software that cannot be ported by the community. Even in |
21 |
these cases, however, AMD64 is actually a dual-bitness arch, and can run |
22 |
32-bit software even on a normal 64-bit system, provided support is |
23 |
activated in the kernel, and a multilib profile (the default amd64 |
24 |
profiles) is chosen. It does get a bit more complicated, but it's not |
25 |
overly so, and there is plenty of documentation and help available when |
26 |
needed. |
27 |
|
28 |
Basically, there are four applications areas and two kernel areas one |
29 |
needs to worry about, all proprietary only, games, multimedia codecs, |
30 |
open office, and browser plugins, on the userland side, video and network |
31 |
drivers, depending of course on your hardware, on the kernel side. |
32 |
|
33 |
The easiest way to solve the userland side is to merge the 32-bit |
34 |
binary-only compatibility libs (glibc, gcc, and portage's sandbox, are |
35 |
exceptions that are normally compiled with support for both bitnesses). |
36 |
This will support most 32-bit-only binary-only games, as well as the |
37 |
32-bit binary-only compatibility ebuilds for firefox, mplayer, and open |
38 |
office. Note that 64-bit versions of mplayer and firefox can be compiled |
39 |
from source as is usual for Gentoo, but because 32-bit libraries won't |
40 |
function in 64-bit apps, and certain codecs (the windows media stuff, for |
41 |
example) and plugins (flash, for example) are only available as |
42 |
proprietary binary-only 32-bit libs, those will be unavailable to the |
43 |
64-bit versions of firefox and mplayer. Whether that's a problem for you |
44 |
depends entirely on how much you depend on unfree software. |
45 |
OpenOffice.org isn't yet fully 64-bit ported, tho that is expected for the |
46 |
2.1 version. However, due to its size, even many 32-bit Gentoo users run |
47 |
the binary version of it. |
48 |
|
49 |
The kernel side is similar. Some network drivers, particularly wireless, |
50 |
are only available in proprietary 32-bit kernel modules (or require the |
51 |
bridge, I forget it's name, allowing 32-bit MSWormOS drivers to be used |
52 |
under Linux). That can be a major issue, but it's easily solved if one is |
53 |
willing to either do the research and buy only open source supported |
54 |
hardware originally, or fork over the money to switch NICs to a supported |
55 |
one, later. |
56 |
|
57 |
Both ATI and NVidia now have 64-bit proprietary modules that can be |
58 |
inserted into the kernel. This is of course only required for 3D support, |
59 |
the most common use of which is unfree games, so if you don't use |
60 |
unfree software, it's not so much of an issue. NVidia support tends to be |
61 |
better, but it's still proprietary. Again, the problem is easily solved if |
62 |
one does the research and buys something with open source drivers, such as |
63 |
ATI Radeons thru the 9250, but not later with 3D support, tho 2D support |
64 |
works. (Xorg 7.0, just now coming out, is said to include experimental |
65 |
3D support for newer ATI cards, but it's just that, experimental, and they |
66 |
had to reverse engineer the cards to get it, because ATI quit cooperating |
67 |
with the open source community after the 9250.) |
68 |
|
69 |
Here, I choose not to run what I call slaveryware, because it's not free |
70 |
(see my sig), so I don't have to worry about closed source 32-bit-only. I |
71 |
run a Radeon 9250 with the open xorg native drivers, and did the research |
72 |
on my motherboard to know everything on it, including the network chip, |
73 |
had in-kernel Linux drivers available. I don't need Open Office, so that |
74 |
doesn't bother me either. I have been running the default multilib |
75 |
profile, but since I don't run any 32-bit apps (save for booting with |
76 |
grub, which also has a 32-bit binary compatibility package in portage, |
77 |
grub-static) and compiling the 32-bit support into gcc and glibc in |
78 |
particular takes additional merging time, I'm currently considering |
79 |
switching to the no-multilib profile. |
80 |
|
81 |
Condensing that all down to a simple summary, most stuff you will run is |
82 |
available in 64-bit, no problem. The problems, with the exception of |
83 |
OOo, are mainly confined to 32-bit-only proprietaryware, but even then, |
84 |
32-bit runs quite well on the amd64 arch. Setting up dual 32 and 64-bit |
85 |
support is a bit more complex than 32-bit only, but there's the usual |
86 |
level of good Gentoo documentation and help available when it's necessary, |
87 |
so most run it with very little more difficulty than they'd have running |
88 |
32-bit x86 Gentoo. |
89 |
|
90 |
-- |
91 |
Duncan - List replies preferred. No HTML msgs. |
92 |
"Every nonfree program has a lord, a master -- |
93 |
and if you use the program, he is your master." Richard Stallman in |
94 |
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html |
95 |
|
96 |
|
97 |
-- |
98 |
gentoo-amd64@g.o mailing list |