linux-origin
[Top] [All Lists]

Re: kernel compile time

To: Kanoj Sarcar <kanoj@xxxxxxxxxxxxxxxxxxx>
Subject: Re: kernel compile time
From: Rajagopal Ananthanarayanan <ananth@xxxxxxx>
Date: Thu, 01 Jun 2000 10:36:00 -0700
Cc: linux-origin@xxxxxxxxxxx, skunx@xxxxxxxxxxxx
References: <200006011716.KAA82918@xxxxxxxxxxxxxxxxxxx>
Sender: owner-linux-origin@xxxxxxxxxxx
Kanoj Sarcar wrote:
> 
> >
> >
> > How many cpus does this machine have?
> 
> How stupid of me - 8 cpus, 128Mb (on 1 node).
> 
> Kanoj
> 
> >
> > Simon.
> >
> > > Here's some more "time" results:
> > >
> > > make
> > > 742.60user 98.79system 13:28.03elapsed 104%CPU (0avgtext+0avgdata 
> > > 0maxresiden
> > > t)k0inputs+0outputs (497099major+616071minor)pagefaults 0swaps
> > >
> > > make -j4
> > > 741.31user 127.30system 4:14.02elapsed 341%CPU (0avgtext+0avgdata 
> > > 0maxresiden
> > > t)k0inputs+0outputs (497100major+616061minor)pagefaults 0swaps
> > >
> > > make -j6
> > > 751.95user 127.98system 4:00.42elapsed 365%CPU (0avgtext+0avgdata 
> > > 0maxresiden
> > > t)k0inputs+0outputs (497100major+616009minor)pagefaults 0swaps
> > >
> > > make -j8
> > > 754.64user 129.48system 4:07.33elapsed 357%CPU (0avgtext+0avgdata 
> > > 0maxresiden
> > > t)k0inputs+0outputs (497100major+615999minor)pagefaults 0swaps
> > >

Kanoj, earlier you had said the wall time looks bad.
I don't understand. For the 1-way make, it took 13:28 == 808 seconds,
of which 742sec were in user mode (> 90%) ... which is typical of
large compilation workloads.

Going to -j4, elapsed time went from 808sec -> 254sec (4:14)
which is a scaling of ~3.2 times ... if -j4 had had an
elapsed time of 202sec (3:22) then it would have been perfect
scaling. But add in the fact that -j4 vs. NO-J would introduce
additional complexity in the make itself, which now spawns
more processes, etc. ... this shows in the increased system time
for -j4; of course the extra system time also counts increased
lock contention. For this particular workload (primarily user-mode,
little-or-no-shared parallelism), -j4 did pretty good.

Finally, -j6 & -j8 may not discover any extra parallelism
in the makes ... one gross way to check this is to watch "top"
and see how many "cc" processes are running on an average.
If it stays at 4, then -j4 is all that'll work.


-- 
--------------------------------------------------------------------------
Rajagopal Ananthanarayanan ("ananth")
Member Technical Staff, SGI.
--------------------------------------------------------------------------

<Prev in Thread] Current Thread [Next in Thread>