From: His Holiness the Reverend Doktor Xenophon Fenderson, the Carbon(d)ated
Subject: [attempt at humor] Recent C environment startup test for the "null program"
Date:
Message-ID: <w4oog8srzrd.fsf@nemesis.irtnog.org>
I took the not.for.email guy's request for benchmark information
seriously enough that I compiled a null program (an empty main() that
returns 0) and ran it on my AlphaStation 200 4/233 (fairly old),
single-processor (EV45) at 233 MHz, 64 megaoctets of RAM, and a fairly
speedy Seagate SCSI hard disk.
The total time to load the C environment (Red Hat Linux 6.0), execute
my null program, and exit: 5 seconds! Note that this stripped-down
environment lacks all of the goodies one would expect from a well
behaved development and application environment (e.g. GUI, multiple
logins, compiler, etc). If I were to load all that into the C
environment before executing my null program, startup times are in the
1-to-2-minute range!!!
The command I used to start the C stripped-down environment may be of
interest to some, so that others can try the test out on other
(faster, dual-processor) Alphas:
>>> boot dka0 -file 4/vmlinux.gz -flags "root=/dev/sda4 init=/tmp/a.out"
HTH!!!
Tongue placed firmly in cheek,
#\Xenophon
P.S. Unfortunately, I was unable to find a stand-alone Lisp
environment that would work on my Alpha. All Lisp implementations I
found [CMUCL, Clisp, GNU EMACS] cheat by using an already running C
environment...since none of them have to deal with initializing
hardware or starting filesystem and network drivers, their startup
times are all well under 100ms.
P.P.S. Fortunately, my C implementation, Linux, like the Lisp
environments of old, only needs to be restarted once every couple of
weeks. I hear that another C environment, Windows, has to be
restarted several times a DAY, if not more often, with startup times
in the tens-of-minutes range! It is ridiculous to expect end users to
put up with such abysmal performance!!!
--
Don't trust these UNIX people. They are all demons. They kill their
parents and fork children. I don't know how they could do this with
their balls cut off but they manage. -- anonymous
From: Tim Bradshaw
Subject: Re: [attempt at humor] Recent C environment startup test for the "null program"
Date:
Message-ID: <ey3vh3017cs.fsf@cley.com>
* d wrote:
> The total time to load the C environment (Red Hat Linux 6.0), execute
> my null program, and exit: 5 seconds! Note that this stripped-down
> environment lacks all of the goodies one would expect from a well
> behaved development and application environment (e.g. GUI, multiple
> logins, compiler, etc). If I were to load all that into the C
> environment before executing my null program, startup times are in the
> 1-to-2-minute range!!!
I just timed a standalone lisp system:
from power on to pinging was 53 secs (this includes memory
checks, several stages of booting). From power on to the OS
first running is maybe half that.
from power on to a window on an X display was 83 secs (this
includes me asking for the window, I didn't hack the boot stuff
to make it open one automatically, so take maybe 5-10 secs for
that.
Power on to me logged on was just over 2 mins
reboot time to ping was 27 secs (avoids the low-level booting
and memory check).
This is on 1990-era hardware mind you, and includes more goodies than
you probably want. A modern machine should be booting 10x as fast
(100x?). I don't having less stuff in the image would make it faster,
as the stuff you don't want probably never comes off the disk.
--Tim `Meaningless Benchamrks-R-Us' Bradshaw