1 |
Joshua Murphy wrote: |
2 |
> On Wed, Jul 27, 2011 at 5:51 PM, Dale<rdalek1967@×××××.com> wrote: |
3 |
> |
4 |
>> Michael Mol wrote: |
5 |
>> |
6 |
>>> On Wed, Jul 27, 2011 at 5:07 PM, Dale<rdalek1967@×××××.com> wrote: |
7 |
>>> |
8 |
>>> |
9 |
>>>> Neil Bothwick wrote: |
10 |
>>>> |
11 |
>>>> |
12 |
>>>>> On Wed, 27 Jul 2011 10:58:28 -0500, Dale wrote: |
13 |
>>>>> |
14 |
>>>>> |
15 |
>>>>> |
16 |
>>>> I have 16Gbs here. It's not like I'm going to run out or anything. I |
17 |
>>>> can |
18 |
>>>> put half on tmpfs and still have 8Gbs left. That is more than enough to |
19 |
>>>> compile even OOo with no space problems. |
20 |
>>>> |
21 |
>>>> Thoughts? |
22 |
>>>> |
23 |
>>>> |
24 |
>>> This is Gentoo, where all us users are reputed to spend their days |
25 |
>>> passing around benchmarks of "emerge -e world", right? |
26 |
>>> |
27 |
>>> Try it. :) |
28 |
>>> |
29 |
>>> |
30 |
>>> |
31 |
>>> |
32 |
>> Yep, we do that. Mine is about 12 hours. I would post it but I got a |
33 |
>> Blocker instead. I'll work on that and post it later. lol |
34 |
>> |
35 |
>> Point is, I have more than enough memory to test the theory that putting |
36 |
>> portage's work directory on tmpfs is faster/slower than a hard drive. I |
37 |
>> tested it and it really didn't make much difference. Most of the time it |
38 |
>> was slower but a couple times it was faster but only by a few seconds. |
39 |
>> |
40 |
>> I guess drives are just a lot faster nowadays or at least fast enough. I |
41 |
>> dunno. |
42 |
>> |
43 |
>> Dale |
44 |
>> |
45 |
>> :-) :-) |
46 |
>> |
47 |
>> |
48 |
>> |
49 |
> I would hazard a guess here that, rather than it being a benefit from |
50 |
> improvement in drive speeds, it's moreso an improvement in the |
51 |
> kernel's file caching. As I understand it (likely incorrectly) tmpfs |
52 |
> essentially does little more than inject the given file into the |
53 |
> filesystem cache, with a 'keep this here' flag attached to it. As |
54 |
> files are accessed on the disk, they're pulled into the filesystem |
55 |
> cache anyhow, and they stay there as long as they're being used and |
56 |
> there's room to keep them. With tmpfs, every file you put into it |
57 |
> stays until explicitly removed, wheras letting the kernel handle the |
58 |
> selected list of files in the cache only keeps the ones the kernel |
59 |
> feels are worth keeping. I am curious, though, whether *not* using |
60 |
> -pipe will actually improve performance when building in tmpfs, as |
61 |
> you're already sidestepping most of the overhead of creating files |
62 |
> with tmpfs, so piping data rather than using intermediate files just |
63 |
> creates extra memory usage overhead (which, while cheaper than disk |
64 |
> i/o and filesystem metadata updates, is still overhead). |
65 |
> |
66 |
> Another likely source of performance loss in using tmpfs over physical |
67 |
> disk to hold the build is that, by design, tmpfs occupies a portion of |
68 |
> the filesystem cache. During a build, every header imported, every |
69 |
> library linked, and every process that runs to make it all come |
70 |
> together gets touched. When you touch files on disk, some or all of |
71 |
> them get pulled into the filesystem cache, so keeping the entire build |
72 |
> tree in the cache may well leave more frequently used files being |
73 |
> dropped from cache to trade back and forth between one another. |
74 |
> |
75 |
> Again, all of this comes with a "I likely have no idea what I'm |
76 |
> talking about, but it sounds convincing" disclaimer. |
77 |
> |
78 |
> :-) |
79 |
> |
80 |
> |
81 |
|
82 |
Sounds better than my "I dunno" tho. LOL |
83 |
|
84 |
I know it doesn't make much sense tho, at least not based on what |
85 |
several others thought would happen way back in the day. Back then, |
86 |
very few people had enough ram to really test this. Those who did, |
87 |
well, it may not be their machine to run the test. ;-) |
88 |
|
89 |
I may test this again one day. All you have to do is create a set and |
90 |
emerge it. Reboot to make sure the cache is cleaned 100%, mount tmpfs |
91 |
and emerge the set again. Compare the times and see what hits the fan, |
92 |
theory or reality. :/ |
93 |
|
94 |
Dale |
95 |
|
96 |
:-) :-) |