Date: Tue, 11 Feb 2014 19:21:35 -0600 From: Tim Daneliuk <tundra@tundraware.com> To: freebsd-stable@freebsd.org Subject: Re: And Here I Thought buildworld/makeworld Was IO Bound Message-ID: <52FACC9F.3000009@tundraware.com> In-Reply-To: <52FA9CB3.4060609@halplant.com> References: <52F84AF8.8050007@tundraware.com> <CAFHbX1JcWpW1UOuLFj0PkSBGBsmbKfot1FN8QreHYD-M84E=4Q@mail.gmail.com> <52F96B27.4090509@tundraware.com> <CAFHbX1%2BYsSuDS%2BUVKBtRTFR0XXmzpRzDi9b7O5wP2zwwc61mbw@mail.gmail.com> <52FA70DA.7050601@tundraware.com> <52FA9CB3.4060609@halplant.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 02/11/2014 03:57 PM, Andrew J. Caines wrote:
> On 02/11/2014 01:50 PM, Tim Daneliuk wrote:
>> does buildworld/kernel even use /tmp or otherwise access tmpfs? It
>> seems to produce temporary output somewhere deep in /usr/obj. Is
>> there a way to get it to use tmpfs or write it's output to /tmp
>> instead?
>
> I've been using a mfs for /usr/obj and WRKDIRPREFIX for ports for years.
>
> md /usr/obj mfs rw,async,noatime,noauto,-s6g 0 0
> md /var/work mfs rw,async,noatime,noauto,-s2g 0 0
>
> Makes cleanup quick and easy, too.
>
>
Final Results:
I did a number of different timings including symlinking /usr/obj
to /tmp/obj when /tmp is a tmfs as well as mounting /usr/obj on an md
as suggested above. In all cases, the time to do the buildworld/kernels
was about the same.
So, I conclude the following:
When running on an SSD, at least the one I am using, the performance
is bounded by CPU, not disk.
Many thanks to all who took the time to answer my stupid questions ...
----------------------------------------------------------------------------
Tim Daneliuk tundra@tundraware.com
PGP Key: http://www.tundraware.com/PGP/
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?52FACC9F.3000009>