The SPUs seem to become VERY irrealistic

Bill maltby - LFS Related lfsbill at wlmcs.com
Fri Oct 18 14:51:32 PDT 2002


On Fri, 18 Oct 2002, Florian Hess (FLoH) wrote:
> Hi lfs-experts,

Not me! :)

> All went well - until the gcc compilation in ch5 (lfs4.0). My 
> measurement of 1 SPU = 892sec is no more realistic at all. The 
> compilation has been started 6:28p, and 10:35p (for 4h instead of 
> approx 2,5h) the process was still in the make phase. I get more and 
> more uneasy (Why I don't know. "coolness" is not really one of my 
> strengths). How long will the glibc take, then?

You have a tough wait ahead of you. Here are some timings from when I ran
on a Cyrix Pr-166 with 64MB and 256MB swap. This was *before* static and
gcc-3.X. Using gcc-3.X added over 2 hours on an IBM6x86-II (Cyrix-II) 300
with 256MB PC-133 SDRAM and 256MB swap.

wlmlx1 IBM6x86-PR166 128MB SDRAM =============================
Chapter05/Batchrun01
   creating-dirs: start       Fri May 10 21:17:05 /etc/localtime 2002
   Run copy-old-nss end       Fri May 10 19:33:29 EDT 2002
                              ----------------------------
                                                  01:43:36
Chapter06/Batchrun01
   Run configure end          Sat May 11 04:32:23 EDT 2002
   create-bash-profile: start Fri May 10 21:19:11 EDT 2002
                              ----------------------------
                                                  07:13:12
Chapter07/Batchrun01
   Run network end            Sat May 11 08:42:31 EDT 2002
   setclock: start            Sat May 11 08:42:30 EDT 2002
                              ----------------------------
                                                  00:00:01
Chapter08/Batchrun01.H1
   Run lilo end               Sat May 11 11:47:01 EDT 2002
   fstab: start               Sat May 11 10:12:12 EDT 2002
                              ----------------------------
                                                  01:34:49
Chapter08/Batchrun01
   Run lilo end               Mon May 13 19:42:51 EDT 2002
   make-modules-conf: start   Mon May 13 19:42:09 EDT 2002
                              ----------------------------
                                                  00:00:42
Chapter09/Batchrun01
   Run rebooting end          Sun May 12 08:30:08 EDT 2002
   PasswdDft: start           Sun May 12 08:30:07 EDT 2002
                              ----------------------------
                                                  00:00:01
                              ============================
                                                  10:32:21

More runs on same machine, upgraded to P-MMX 200MHz gave 09:05:44.

Watch your CPU does not overheat and start giving signal 11. Especially
during gcc/glibc (and a few other places). If it does, just wait a minute
or two and type the same make command again. Make is smart enough to pick
up where it left of (sort of). It mail fail several times this way. As
long as it fails later in the make each time, keep doing it. You will get
done.

> 
> Can that happen, and how? My system is: P150, 16mb + 300mb Swap space, 
> Debian Potato, CD1 (all development tasks installed).

If you can stumble on some memory, that will speed things a *lot*. I have
had slow and trouble when my smallest machines were 32MB. You will be
swapping a *lot*.

> 
> The dma mode has been activated at 10:00p, but he doesn't seem to 
> accelerate the compilation.

Unfortunately, most of the work is in the CPU and memory (except for your
sadly swapping a lot) so the DMA won't help much. Too little memory for it
to really pay off.

> Any hints to change something? May it be that the compiler is caught 
> in a never-ending loop?

If you are making a log file, tail -10 every once in a while. You should
see different programs being operated on at different times.

> I'm afraid to have to sleep on the sofa in the living room because the 
> pc (with the hdd having hard work) is installed in my chamber :-(.

What? You are not used to it yet?  ;)

> FLoH.

-- 
Bill Maltby
billm at wlmcs.com

-- 
Unsubscribe: send email to listar at linuxfromscratch.org
and put 'unsubscribe lfs-support' in the subject header of the message



More information about the lfs-support mailing list