1 |
Beso <givemesugarr@×××××.com> posted |
2 |
d257c3560802060625o3261f5cfw4eec6d5718e3bad9@××××××××××.com, excerpted |
3 |
below, on Wed, 06 Feb 2008 14:25:24 +0000: |
4 |
|
5 |
> 2008/2/6, Duncan <1i5t5.duncan@×××.net>: |
6 |
>> |
7 |
>> -fomit-frame-pointer is the default for -O and above on amd64/x86_64 |
8 |
> |
9 |
> i tend to always add it since i don't want to always look at what the |
10 |
> arch would do later. |
11 |
|
12 |
OK, makes sense. I used to as well, but I've enough other stuff there |
13 |
now that if I can avoid duping defaults, I do! =8^) |
14 |
|
15 |
> i tend to think that Os should be gentoo's default, since it adds some |
16 |
> options for cache and disk occupation and since i've always been good |
17 |
> with it. |
18 |
|
19 |
Like I said, I used to use it, and agreed with you. I still do in |
20 |
principle, but with gcc-4.2, -O2 is enough improved it does what I want, |
21 |
now, without doing what I don't want, for the most part. There's still a |
22 |
couple bits I change, but from 4.2, there'd be about as many I'd change |
23 |
for -Os as for -O2, and -O2 is certainly better tested, so that's what I |
24 |
decided to run for at least this version. I'll have to see if that |
25 |
changes again when 4.3 comes out. |
26 |
|
27 |
> -mmmx is included in the default for -march=k8 |
28 |
> |
29 |
> when i've added it it wasn't included by default. |
30 |
|
31 |
Are you sure? For amd64/x86_64, AFAIK, -mmmx has always been there as |
32 |
it's part of the architecture definitions. (BTW I wonder if Via's new 64- |
33 |
bit Isaiah changes things in this regard; IIRC they omitted part of the |
34 |
pentium spec for some 32-bit CPU some years ago now altho it wasn't Via |
35 |
at the time but someone they later purchased, but anyway, it threw |
36 |
compilers for a loop for awhile, but hopefully they learned from that.) |
37 |
|
38 |
>> What about -mno-tls-direct-seg-refs? Why do you use that? |
39 |
> |
40 |
> this should help with virtualizazion apps. xen won't work well without |
41 |
> it with large system memory. without xen or other virt apps this |
42 |
> shouldn't influence much on the packages. |
43 |
|
44 |
OK, that answers /that/ question. My CPUs, being Opteron 290s, aren't of |
45 |
the virtualization instruction generation yet, and I've kind of avoided |
46 |
it for that reason. Thus, the flag above isn't something I need to worry |
47 |
about. Thanks! |
48 |
|
49 |
>> > also add: |
50 |
>> > LDFLAGS="-Wl,--as-needed,-O1 -Wl,--enable-new-dtags -Wl,--sort-common |
51 |
>> > -s" |
52 |
>> |
53 |
>> Could you point me to documentation on LDFLAGS in general, or at least |
54 |
>> --enable-new-dtags and --sort-common? I use --as-needed already ( |
55 |
>> documentation at http://www.gentoo.org/proj/en/qa/asneeded.xml ) |
56 |
|
57 |
> --sort-common is to prevent gaps between symbols due to alignment |
58 |
> constraints, presumably increasing efficiency layout. here's an |
59 |
> interesting topic on ldflags: |
60 |
> http://forums.gentoo.org/viewtopic-t-67777.html |
61 |
|
62 |
Thanks! Bookmarking for later reference (I'm about to go to sleep now |
63 |
and am too sleepy to make sense of much new stuff ATM). |
64 |
|
65 |
> --enable-new-dtags should be default for current binutils but it was |
66 |
> there from my 3 years old make.conf so it is still there. adding it |
67 |
> again shouldn't do harm. |
68 |
|
69 |
Again, thanks. Now that you mention it, I think I did see something to |
70 |
that effect elsewhere, but didn't worry about it since it was now the |
71 |
default. |
72 |
|
73 |
> yep, i'll try out [-j, unlimited jobs] someday. |
74 |
> for the moment i'm planning a disk |
75 |
> change, mine has already 2 years and its overall health is starting to |
76 |
> be heard when writing or reading data from the disk. |
77 |
|
78 |
Ouch! Unfortunately, my last couple disks didn't really get to that |
79 |
point. The last one in particular got too hot after the A/C died on a |
80 |
summer day when I was out, here in Phoenix. The room air temp could have |
81 |
gotten to well over 50 C, so who knows how hot the disk got! Fortunately |
82 |
I had the disk partitioned decently and unmounted backup partitions of |
83 |
most stuff, which was generally retrievable. I was even able to run the |
84 |
thing in operation for awhile after it cooled back down -- NOT using the |
85 |
partitions that the head had crashed on due to the heat, of course, but I |
86 |
replaced it as soon as I could scrap the money together. |
87 |
|
88 |
That's when I decided I needed RAID. Two disks in a row going out after |
89 |
almost exactly a year each. FWIW, I also ended up replacing the A/C, and |
90 |
I've not had a problem since, 2.5 years ago now. |
91 |
|
92 |
> i've got an |
93 |
> external bigger disk that i'll partition with tmpfs for paludis and |
94 |
> portage. the only thing that i still don't know is: |
95 |
|
96 |
> 1. can i put on more than one different tmpfs?! |
97 |
|
98 |
Yes. I'm actually running several, /dev (udev), max size 2 MB, /dev/shm |
99 |
(FHS but used only by portage for PORTAGE_TMPFS, not to be confused with |
100 |
PORTAGE_TMPDIR and PKG_TMPDIR), max 50 MB, /lib64/rcscripts/init.d (old |
101 |
entry, now inactive, old baselayout-1.x), max 512 KB, and my big one, / |
102 |
tmp, 6 GB. /var/tmp is actually a symlink to the 6 gig /tmp, and both |
103 |
PORTAGE_TMPDIR and PKG_TMPDIR point to /tmp. (Note that ccache's dir is |
104 |
inside PORTAGE_TMPDIR by default according to make.conf.example. You'll |
105 |
want to put it elsewhere, OUTSIDE the tmpfs, so it doesn't die at reboot, |
106 |
rather defeating the purpose.) |
107 |
|
108 |
2. if i put one tmpfs and symlink there what i do want to use on tmpfs is |
109 |
better?! |
110 |
|
111 |
I can't quite get that question to parse. However... remember that while |
112 |
you can have multiple tmpfs mounted, if you might be using both at once, |
113 |
you need to consider the total effect on memory. Thus, it's probably |
114 |
better to only have one big (multi-gig) one, and point other things at it |
115 |
if necessary. I've used symlinks (as with /var/tmp above), but mount -- |
116 |
bind /should/ work as well, AFAIK. |
117 |
|
118 |
While we're talking symlinks, this is for cache not tmp, but it may be |
119 |
worth mentioning that I have one filesystem containing all my system |
120 |
cache stuff on RAID-0, that is, CCACHE_DIR, PORTDIR (including layman's |
121 |
subdirs and DISTDIR but not PKGDIR), and /usr/src. Those are |
122 |
traditionally in various separate locations, but here, I have them all as |
123 |
subdirs on the same RAID-0 backed filesystem for speed and because they |
124 |
don't need redundancy. CCACHE_DIR and PORTDIR can of course be reset to |
125 |
point to the desired subdirs directly, but /usr/src is a symlink to the |
126 |
appropriate subdir on my RAID-0, which I have mounted as /str (for |
127 |
striped). So /usr/src -> /str/usrsrc as a symlink. I also symlink most |
128 |
of my frequently used dirs directly from root, too. So /h -> /home |
129 |
(actually, the reverse, I mount my home filesystem on /h, and /home -> / |
130 |
h), /usr/local -> /l, /var/log -> /log, /usr/portage -> /p -> /str/ |
131 |
portage, etc. Saves a lot of typing that way, as I can refer to /l/bin/ |
132 |
script or /log/messages or /p/profiles/package.mask, for instance. |
133 |
|
134 |
Symlinks are SO useful; I'm STILL trying to figure out how I survived a |
135 |
decade on MS without them! (Actually, by W98 time, I was using a third |
136 |
party Explorer extension that allowed one to create custom system |
137 |
folders, much like My Computer is as shipped, but of course that was |
138 |
only /sort/ of like symlinks, I had yet to discover the power of the real |
139 |
thing! =8^) Of course, since the advent of --bind and friends as options |
140 |
to mount, basically the same thing could be accomplished with it by mount |
141 |
--binding the subdirs as necessary, but I don't think MS does that |
142 |
either, and symlinks work, so that's what I've used. |
143 |
|
144 |
-- |
145 |
Duncan - List replies preferred. No HTML msgs. |
146 |
"Every nonfree program has a lord, a master -- |
147 |
and if you use the program, he is your master." Richard Stallman |
148 |
|
149 |
-- |
150 |
gentoo-amd64@l.g.o mailing list |