| 1 |
On Tue, Oct 13, 2009 at 1:58 PM, Stefan G. Weichinger <lists@×××××.at> wrote: |
| 2 |
> Joshua Murphy schrieb: |
| 3 |
> |
| 4 |
>> Non-ricer? Well... this sorta breaks that category. |
| 5 |
> |
| 6 |
> ;-) |
| 7 |
> |
| 8 |
>> There's a rather handy tool[1] already in the stage3, I've used it |
| 9 |
>> alongside bootchart to force-load everything needed into ram during |
| 10 |
>> boot, before it's needed, so execution gets held up by i/o just a |
| 11 |
>> little less. Actually shaved a few seconds off of my desktop's bootup |
| 12 |
>> once upon a time (3.0 GHz core 2 duo on 4GB of ram, which had |
| 13 |
>> excessive eyecandy while remaining fairly lightweight). The second use |
| 14 |
>> I put it through, and this one just a little more long-term useful, |
| 15 |
>> was preloading my wm, most of my home directory (primarily all the |
| 16 |
>> config files), aterm, firefox, a few other common tools I use, and the |
| 17 |
>> libraries they were using on my system while logging in. All of my |
| 18 |
>> applications were starting in no time at all. The catch... I took the |
| 19 |
>> brute force approach, rather than using an add-on tool to |
| 20 |
>> automagically choose what to prefetch for me. There are also setting |
| 21 |
>> in the bootscripts that, if you're not already using them, will make |
| 22 |
>> use of at least a little, using tmpfs here and there, and also just |
| 23 |
>> putting /tmp and /var/tmp onto tmpfs (outside of building things like |
| 24 |
>> open office, this tends to work well). Oh, and if you really do want |
| 25 |
>> to use up all that ram.... build openoffice in tmpfs... if it could |
| 26 |
>> use all 8GB for files only, it might actually work out, I know it |
| 27 |
>> kills off when you only have part of 4GB to feed it. |
| 28 |
>> |
| 29 |
>> [1] artifice ~ # busybox readahead |
| 30 |
>> BusyBox v1.14.2 (2009-10-13 06:37:22) multi-call binary |
| 31 |
>> |
| 32 |
>> Usage: readahead [FILE]... |
| 33 |
>> |
| 34 |
>> Preload FILE(s) in RAM cache so that subsequent reads for thosefiles |
| 35 |
>> do not block on disk I/O |
| 36 |
> |
| 37 |
> I use sys-apps/preload ... is this comparable? |
| 38 |
> |
| 39 |
> I don't need to speed up boot-times as I use suspend-to-ram on my main |
| 40 |
> workstation as well, so I don't care much about booting ... |
| 41 |
> |
| 42 |
> Thanks anyway for sharing! Stefan |
| 43 |
> |
| 44 |
> |
| 45 |
|
| 46 |
Yep! sys-apps/preload goes a long way to automate what I did with |
| 47 |
busybox's readahead tool, and I'm about 99% certain they both make the |
| 48 |
same system call to do the work anyways... it's all a matter of how & |
| 49 |
where they get their list of *what* to fetch... I, as I said, took the |
| 50 |
brute force approach and made my list by hand. ;-) |
| 51 |
|
| 52 |
-- |
| 53 |
Poison [BLX] |
| 54 |
Joshua M. Murphy |
Archives