[gentoo-dev] Compiling in a RAM disk

[gentoo-dev] Compiling in a RAM disk

[Andy Arbon]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

I seem to remember that a while back there was talk of modifying Portage
so that it could do compilations in RAM disk rather than using the HD.

I'm curious whether this idea was accepted by the devs and if so, when
it's targeted to be finished?

I have a fair amount of RAM and a slow disk, so I'm quite keen on the idea.

Cheers,

Andy
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQE+MTryX3TTUvZURBERAlcEAJsHDVcDmExfpbV8iyTfLz2BagJbxACdHFTt
zh7Rb4UkDpUKh1HLRci6DtI=
=O27F
-----END PGP SIGNATURE-----


--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Compiling in a RAM disk

[Toby Dickenson]

On Friday 24 January 2003 1:09 pm, Andy Arbon wrote:
> Hello,
>
> I seem to remember that a while back there was talk of modifying Portage
> so that it could do compilations in RAM disk rather than using the HD.
>
> I'm curious whether this idea was accepted by the devs and if so, when
> it's targeted to be finished?
>
> I have a fair amount of RAM and a slow disk, so I'm quite keen on the idea.

Try running vmstat while emerging; do you actually see any disk activity? On 
this machine with 500M ram, my last kde emerge barely touched the disks.



--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Compiling in a RAM disk

[Christopher J. PeBenito]

If you want to compile in a ramdisk, couldn't you simply mount a tmpfs
at /var/tmp/portage?

Something like this (tmpfs max size 512MB):
mount -t tmpfs none /var/tmp/portage -o size=512m

On Fri, 2003-01-24 at 07:09, Andy Arbon wrote:
> Hello,
> 
> I seem to remember that a while back there was talk of modifying Portage
> so that it could do compilations in RAM disk rather than using the HD.
> 
> I'm curious whether this idea was accepted by the devs and if so, when
> it's targeted to be finished?
> 
> I have a fair amount of RAM and a slow disk, so I'm quite keen on the idea.
> 
> Cheers,
> 
> Andy

-- 
Chris PeBenito
<pebenito@ieee.org>


"Engineering does not require science. Science helps
a lot, but people built perfectly good brick walls 
long before they knew why cement works."-Alan Cox

--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Compiling in a RAM disk

[Christopher J. PeBenito]

Well you can omit the size option, and the ramdisk size will only be
limited by the size of your memory plus swap.  If you want to limit it,
you can set the size to whatever you want.

On Fri, 2003-01-24 at 09:41, Nicholas Henke wrote:
> On 24 Jan 2003 08:39:43 -0600
> "Christopher J. PeBenito" <pebenito@ieee.org> wrote:
> 
>     > If you want to compile in a ramdisk, couldn't you simply mount a tmpfs
>     > at /var/tmp/portage?
>     > 
>     > Something like this (tmpfs max size 512MB):
>     > mount -t tmpfs none /var/tmp/portage -o size=512m
>     > 
> Sadly enough, 512 is not enough for things such as X. I am not sure if there are other packages that require more to build.
> 
> Nic 
-- 
Chris PeBenito
<pebenito@ieee.org>
AIM: PeBenito78
ICQ#: 10434387

"Engineering does not require science. Science helps
a lot, but people built perfectly good brick walls 
long before they knew why cement works."-Alan Cox

--
gentoo-dev@gentoo.org mailing list

Re: [gentoo-dev] Compiling in a RAM disk

[Evan Powers]

On Friday 24 January 2003 08:20 am, Toby Dickenson wrote:
> Try running vmstat while emerging; do you actually see any disk activity?
> On this machine with 500M ram, my last kde emerge barely touched the disks.

Yeah, I'm completely with Toby on this one.... Because Linux aggressively 
caches disk access, I really wouldn't expect compiling on a RAM disk to do 
any better. If you have enough RAM for this to be feasible, the compilation 
is probably happening in RAM anyway (disk cache), so you'd gain nothing.

The only thing I could possibly see confusing this is when you actually start 
having to swap out to disk. If page-outs that are swap-backed are very much 
more efficient than page-outs that are filesystem-backed, maybe then a RAM 
disk would give you a gain.

I guess I wouldn't be surprised if page-outs to a swap partition are faster 
than page-outs to a filesystem, that's why Linux has swap partitions in the 
first place. But does anybody have concrete numbers proving that it actually 
matters in this particular case, that is, compilation?

Evan

--
gentoo-dev@gentoo.org mailing list