[lfs-dev] LFS and 32-bit systems

Bruce Dubbs bruce.dubbs at gmail.com
Sun Dec 11 16:00:34 PST 2016


Douglas R. Reno wrote:
>
>
> On Sun, Dec 11, 2016 at 5:11 PM, Bruce Dubbs <bruce.dubbs at gmail.com
> <mailto:bruce.dubbs at gmail.com>> wrote:
>
>     akhiezer wrote:
>
>             From: Bruce Dubbs <bruce.dubbs at gmail.com
>             <mailto:bruce.dubbs at gmail.com>>
>             Date: Sun, 11 Dec 2016 11:29:42 -0600
>             Subject: [lfs-dev] LFS and 32-bit systems
>
>             Do any of you still have a 32-bit Intel system that you use to
>             test/run LFS?
>
>
>
>         'Yes', nine different 32-bit machines; plus three 64-bit running
>         32-bit os.
>
>         *BUT*, these are for not-by-the-books builds: instead, they're for
>         builds/tests that derive in parts from the books.
>
>
>
>             I am thinking about rewriting the LFS preface section on
>             architecture to say:
>
>             The primary target architecture of LFS is the AMD/Intel x86_64
>             (64-bit)
>             CPU. In addition instructions are included that allow a build
>             on older
>             32-bit Intel CPUs, but testing is not regularly done on that
>             architecture.
>
>
>
>         s/regularly done/done regularly/
>
>
>     Minor, but OK.
>
>             The instructions in this book are also known to work, with some
>             modifications, with the Power PC and ARM CPUs.  ...
>
>             I would then remove the second and third paragraphs and the
>             small table
>             between them.
>
>
>
>         Both paras and table are ok to leave in, until as/when 32-bit is
>         dropped altogether.
>
>
>     When I originally did that, the stats were from a Core2Duo as
>     documented.  Even that is a bit long-in-the-tooth.  I don't know how
>     valid that comparison is today.
>
>         It may be useful/'nice' to include stats from recent by-the-book
>         32-/64-
>         bit (properly-comparable) builds.
>
>
>     I could do that if I had the stats.  I don't have a reasonable way to
>     compare today.  I don't think giving numbers for a 3.3 GHz i7-5820K
>     really is a good platform for the stats.
>
>     My latest build took 64 minutes (vice Core2 at 190.6 minutes), but
>     that was at -j8 and omitted some, but not all tests in chapter 6.

> I can remotely wake up my Core2 if you want those stats (same system, not
> much has changed). I have a Xeon-based system here already with a 2.6 GHz
> CPU with 4 cores if you want those instead.

You have too much going on right now.  I'd rather you concentrate on BLFS.
This is not a priority.

What we need is a full jhalfs build with a 32-bit kernel and another full 
jhalfs build with a 64-bit kernel.  The best cpu would probably be an 
older i3 or i5 at 3 GHz or so.

   -- Bruce




More information about the lfs-dev mailing list