Linux-Misc Digest #515, Volume #25               Mon, 21 Aug 00 13:13:02 EDT

Contents:
  Whats the best window manager? ("Database")
  Re: Best Linux Distribution ("Peter T. Breuer")
  Re: Best Free E-Mail for Linux? (Ron J Theriault)
  Convert running Partition to RAID-1 ??? (Alfred Schott)
  Re: Reality Check - NY Times Article ("Dan Jacobson")
  Re: Other UNIX OS filesystems under linux? (Grant Edwards)
  Re: Whats the best window manager? (Grant Edwards)
  Re: Is Mandrake Really Red Hat... ("Peter T. Breuer")
  Re: Whats the best window manager? ("Peter T. Breuer")
  Re: Other UNIX OS filesystems under linux? ("Peter T. Breuer")
  Re: Best Linux Distribution (Jean-David Beyer-valinux)
  Re: Some weird xterm behaviour! ("Andrew N. McGuire ")
  Re: PC-NFS ("T. Odensson")
  Re: Is Mandrake Really Red Hat... (Kenneth Rørvik)
  Re: which libs for CAD rpogram? ([EMAIL PROTECTED])
  Re: remote host ("Guennadi V. Liakhovetski")
  Re: WWW: TuxJournal online (Matthias Arndt)
  Re: Is Mandrake Really Red Hat... (Johan Kullstam)
  Re: Installing 2nd hard disk ([EMAIL PROTECTED])
  Re: Is Mandrake Really Red Hat... (moonie;))

----------------------------------------------------------------------------

From: "Database" <[EMAIL PROTECTED]>
Subject: Whats the best window manager?
Date: Mon, 21 Aug 2000 16:11:37 GMT

Whats the best window manager?


database



------------------------------

From: "Peter T. Breuer" <[EMAIL PROTECTED]>
Subject: Re: Best Linux Distribution
Date: 21 Aug 2000 16:27:14 GMT

Luc Van Bogaert <[EMAIL PROTECTED]> wrote:
: On Sun, 20 Aug 2000 03:10:45 GMT, William W. wrote:

:>If there is a
:>good resource out there, then we should point people who ask, "Which
:>distro is the best" to that resource. If not, what's wrong with offering
:>our own opinions about why we like our favourite distributions, so long
:>as we don't start flaming each other about it?

: Finally! The first intelligent remark since the original message was
: posted... thanks.

This is about as innocent and intelligent a strategy as proposing that
catholics and protestants in northern ireland start painting on the 
walls of the town hall reasons why they prefer their religion to
others.

(hint: the intelligent people won't comment or take flame bait. That
only leaves the rest).

Peter

------------------------------

From: [EMAIL PROTECTED] (Ron J Theriault)
Subject: Re: Best Free E-Mail for Linux?
Date: 21 Aug 2000 11:12:16 -0500


In article <[EMAIL PROTECTED]>,
 [EMAIL PROTECTED] (David Efflandt) writes:
>On 18 Aug 2000 14:05:27 -0500, Ron J Theriault <[EMAIL PROTECTED]> wrote:
>>
>>What's the best free e-mail service to use for someone
>>who runs Linux?
>
>Your own.  But usually whatever connects you to the internet will provide
>mail service.

Do they all assume you're on a PC, and suck the e-mail down 
to your local disk?

>>What's the easiest free Web hosting site to use for
>>someone familiar with Unix and Apache?
>
>Virtualave.net runs CGI as you anywhere in your webspace (FreeBSD) and in
>my experience has been most reliable.  They have many users, but also many
>boxes.  They require a banner ad like they all do, but you have your
>choice of putting it on the page if you don't want the annoying popup.

They ALL don't.  eWebCity.com doesn't require ads, but all you
get is MS ASP on the server.

>David Efflandt  [EMAIL PROTECTED]  http://www.de-srv.com/
>http://www.autox.chicago.il.us/  http://www.berniesfloral.net/
>http://hammer.prohosting.com/~cgi-wiz/  http://cgi-help.virtualave.net/
>

-- 
Ron Theriault :  CS Department, Texas A&M Univ.            
[EMAIL PROTECTED]  http://www.cs.tamu.edu/staff/ron

"Victimless crime" is a euphemism for "political crime".
-- 
Ron Theriault :  CS Department, Texas A&M Univ.            
[EMAIL PROTECTED]  http://www.cs.tamu.edu/staff/ron

"Victimless crime" is a euphemism for "political crime".

------------------------------

From: Alfred Schott <[EMAIL PROTECTED]>
Subject: Convert running Partition to RAID-1 ???
Date: Mon, 21 Aug 2000 18:01:20 +0200

What is the best way to add RAID-1 functionality to a running System.
What I have is a 8GB Drive with root, /boot and swap Partition.
I have a second 8GB Drive which I want to use with the existing for
RAID-1.

Any Ideas?

Alfred
-- 
============================================================
@work: [EMAIL PROTECTED]                 Tel.:0661/82459
@home: [EMAIL PROTECTED]
============================================================

------------------------------

From: "Dan Jacobson" <[EMAIL PROTECTED]>
Crossposted-To: gnu.misc.discuss
Subject: Re: Reality Check - NY Times Article
Date: Mon, 21 Aug 2000 08:38:35 +0800

> >> But a key problem remains: Wall Street is not buying the Linux pitch.

One would hope not, e.g., Wall Street is not buying the FSF or GNU pitch... because 
there is no pitch... and what if they did buy
the Linux pitch?  Would we then have operating system + banner ad?
--
www.geocities.com/jidanni  ... fix e-mail address to reply; ¿n¤¦¥§
Tel:+886-4-5854780; starting in year 2001: +886-4-25854780



------------------------------

From: [EMAIL PROTECTED] (Grant Edwards)
Subject: Re: Other UNIX OS filesystems under linux?
Date: Mon, 21 Aug 2000 16:37:09 GMT

In article <8nrhe9$4vs$[EMAIL PROTECTED]>, William Baird wrote:

>We have a requirement for 'mission essential' systems to be
>recovered in less than 4 hours.  Even with tapes, we have found
>that it often goes past the deadline.
>
>What I would like to do is set up a 'backup server' where the
>other machines dump their drives to mirror them each night.  A
>cheap linux machine would be ideal.  Just have a bank of drives
>that would be dumped to each night with a one for one
>correspondance.

What sort of things are you trying to guard against?  If you're
worried about disk drives failing, then I'd look into RAID.  If
you're worried about applications going nuts and spewing drivel
into the filesystem, then RAID won't do you much good, and
nightly dumps are probably your best bet.

-- 
Grant Edwards                   grante             Yow!  I always have fun
                                  at               because I'm out of my
                               visi.com            mind!!!

------------------------------

From: [EMAIL PROTECTED] (Grant Edwards)
Subject: Re: Whats the best window manager?
Date: Mon, 21 Aug 2000 16:39:47 GMT

In article <Zwco5.15927$[EMAIL PROTECTED]>, Database wrote:

>Whats the best window manager?

You're about as likely to get useful answers to that question
as you would get by asking:

  What's the best painting?
  What's the best song?
  What's the best movie?
  What's the best dessert?

Asking N people will generate somewhere between N and N^2
answers.

;)

Personally, I use fvwm2.

-- 
Grant Edwards                   grante             Yow!  INSIDE, I have the
                                  at               same personality disorder
                               visi.com            as LUCY RICARDO!!

------------------------------

From: "Peter T. Breuer" <[EMAIL PROTECTED]>
Crossposted-To: alt.os.linux,comp.os.linux.x
Subject: Re: Is Mandrake Really Red Hat...
Date: 21 Aug 2000 16:38:28 GMT

In comp.os.linux.misc Kenneth Rorvik <[EMAIL PROTECTED]> wrote:
: [EMAIL PROTECTED] (Johan Kullstam) wrote in
: <[EMAIL PROTECTED]>: 


:>> Oh well, but the "optimization" would still apply to the later
:>> generation Pentiums, plus K5, K6's and up.
:>
:>this assumption would be wrong.
:>
:>> Even if the performance gain is minimal. 
:>
:>it's not minimal -- it's *negative*.

: Please explain :)

Negative is negative. What's to explain? If you attempt to reorder
machine instructions to gain advantage for a particular architecture,
you will not be optimizing it for a second architecture. In fact
you will be specializing the code and thereby rendering ineffective
the strategies used by the second architecture to "optimize on the
fly", because those strategies are aimed at the general case.

The effect is particularly noticable wrt i586 and i686 architectures.
That's because the i686 contains gigantic internal logic aimed at
optimizing 386 code dynamically.  Any attempt to second guess it at
compile time (or, indeed, replace the 386 code with specialized 586
code) disadvantages that internal mechanism.  I'm tallking about the
branch heuristics here in particular.

You've also got to consider the pipeline optimizations and instruction
reorderings that can take place.  It'll be more difficult to reorder
specialized instructions that generic ones (the RISC/CISC effect).

There is also wastage due to the different memory alignment schemes,
but I don't think it will be nocicable going 586 to 686.

You'd win again on a 787 architecture, though (:-).

Peter

------------------------------

From: "Peter T. Breuer" <[EMAIL PROTECTED]>
Subject: Re: Whats the best window manager?
Date: 21 Aug 2000 16:44:35 GMT

Database <[EMAIL PROTECTED]> wrote:
: Whats the best window manager?

What's the most idiotic question asked by people who don't think what
their question means before asking?

Have you any criteria, or will "the pinkest" suffice as an answer?

Peter

------------------------------

From: "Peter T. Breuer" <[EMAIL PROTECTED]>
Subject: Re: Other UNIX OS filesystems under linux?
Date: 21 Aug 2000 16:43:20 GMT

William Baird <[EMAIL PROTECTED]> wrote:
: What I would like to do is set up a 'backup server' where the other machines
: dump their drives to mirror them each night.  A cheap linux machine would be
: ideal.  Just have a bank of drives that would be dumped to each night with a
: one for one correspondance.

: However, we are working in a heterogenous UNIX environment: Tru64, Solaris, 
: and linux.  We are trying to move to strictly linux, but that is taking time 
: due to the amount of proprietary software, so we have to work with what we 
: have.

Use rsync across the net (possibly under ssh if you care ..).

: Merely dumping and restoring isn't what I'd like to do.  I'd much rather just
: pull disk(s), set the SCSI ID and stick it into whichever system has lost its
: drive(s).  Thus be off and running in minutes even.

: The question being is it possible to have other OS fs's under linux?  

: I know you can play with DOS and MS OS partitions, but what about other
: UNIX's?

Look at the kernel configuration options and find out! What's stopping
you, since you have a "mission critical" situation? You should know
that most FS types have supprt of one kind or another in linux! Have
you never done a kernel compile and checked all those little boxes one
way or another? (yes, that was a very silly and ignorant question ...
about as bad as asking what CPU architectures are supported by linux.
The answers are in the faq and under your nose :-).

FWIW ufs has fairly good support in linux. I don't know what Tru64
uses.  But have you considered 64/32 bit issues?

Peter

------------------------------

From: Jean-David Beyer-valinux <[EMAIL PROTECTED]>
Subject: Re: Best Linux Distribution
Date: Mon, 21 Aug 2000 12:50:41 -0400

Luc Van Bogaert wrote:

> On Sat, 19 Aug 2000 20:23:45 -0500, Andrew N. McGuire  wrote:
>
> >In the end
> >it will be you who has to use it, so instead of starting a holy war
> >here, why don't you do some research for yourself and not bug the NG
> >with inflammatory questions.
>
> Well, one thing I already noticed about some people on this newsgroup,
> is that they are *real friendly* to other people, who are just asking
> for some simple advice. I didn't realize that a simple question like
> this would be enough to start a holy war, jeeeez! What's wrong with you
> people? Has the Linux hype gone to your heads or something? Don't want
> newbies here? To busy with intelligent stuff to answer stupid questions
> about Linux distributions?
>
> Anyway, I will do as you recommend and do some digging myself if I need
> any help or advice about Linux, and think twice before I turn to this
> newsgroup.

I am sorry it has come to this.

I think some of what happens is that there are those who administer Unix
and Linux systems for a living and after they have heard the same question
over and over again, and then get it one more time after a hard day, they
can lose patience.

I am reminded of the following joke:

There is a recruiting sargeant at the local draft board. Twice a day, a
large group of draftees enters the room. He gives each one a clipboard
with a long questionnaire on it. Twice a day, he instructs them to fill
out the forms and to take them to the guy at the desk behind the Left
door. Go through the Left door, not the Right door he says twice a day.

It finally happened that someone tried to go through the Right door and
the sargeant lost his temper: How Many TImes do I have to tell you to go
through the Left Door!!!!!!

It can seem like this to long-time posters on this (or any other) board.
Try to forgive them. It is not really stupidity or a haughty attitude, but
just human nature. It is easy to forget that posters here are people, too.

--
Jean-David Beyer               .~.
Shrewsbury, New Jersey         /V\
Registered Linux User 85642.  /( )\
Registered Machine    73926.  ^^-^^




------------------------------

From: "Andrew N. McGuire " <[EMAIL PROTECTED]>
Subject: Re: Some weird xterm behaviour!
Date: Mon, 21 Aug 2000 11:50:31 -0500

On 21 Aug 2000, Thomas Dickey quoth:

~~ Date: 21 Aug 2000 14:58:59 GMT
~~ From: Thomas Dickey <[EMAIL PROTECTED]>
~~ Newsgroups: comp.os.linux.misc
~~ Subject: Re: Some weird xterm behaviour!
~~ 
~~ Andrew N. McGuire  <[EMAIL PROTECTED]> wrote:
~~ 
~~ > No kidding, still, catting a binary on a remote machine should not
~~ > make your localhost print!  Or do you disagree?  You don't think it
~~ > would be a little bit evil, albeit funny, if while you left yourself
~~ > logged into a remote host, the sysadmin of that machine (or anyone
~~ > with access to your tty) started wasting trees in your office? To
~~ > me that is a bug, not a feature.
~~ 
~~ you are missing the point: the holes that you're "seeing" are only
~~ there for people who have permissions that aren't present on a normally
~~ configured system.

No, not at all.   I log into a remote machine (on say /dev/pts/3),
root on that machine does this:

cat somepdf.gz > /dev/pts/3

My printer at home starts printing.

Thats wrong!
 
~~ (rxvt has some support for printers, but it doesn't work properly - not
~~ the only bug therein).

I suppose in the end that does not matter, as xterm still should not
do that.  I just tried it on Solaris, and the /usr/openwin/bin/xterm
on Solaris 7 does not display this behaviour.  So what I am "seeing"
is a bug.

anm
-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
~ Andrew N. McGuire                                                      ~
~ [EMAIL PROTECTED]                                              ~
~ "Plan to throw one away; you will, anyhow." - Frederick P. Brooks, Jr. ~
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


------------------------------

From: "T. Odensson" <[EMAIL PROTECTED]>
Subject: Re: PC-NFS
Date: Mon, 21 Aug 2000 11:52:20 -0500

Thanks for the start but I need to get pcnfsd installed/configured.  On
Solaris it is rpc.pcnfsd, this is the daemon that I need to install or some
type of similar daemon/service.  I have looked at LDP and the only thing it
says about PC-NFS is:

>You don't want to run PC-NFS. You want to run samba.
>samba is far better than PC-NFS and it works with Windows 3 for Workgroups
and later >versions of Windows. It's faster and more secure too. Use it.
Really.

Well... this is of no help, really.


I guess to be a little more specific, I am using WebNFS and it needs user
authentication, pcnfsd will provide that authentication.  Otherwise it will
use the user/group nobody and you will not have any rights.  I do NOT want
to assign those file to nobody.  I need to be able to "login" into WebNFS.
Anyone have any ideas?

Thanks for the help,
T. Odensson



> Hi!  I presume you mean you need to connect a PC running PC-NFS to your
> NFS server running on Linux, right?  OR do you want to mount NFS
> exported filesystems from a Windows NFS server into your Linux
> filesystem?
>
> If you add the directories you want to export to "/etc/exports" (I think
> it's called) on the Linux box and mae sure rpc.mountd (and possibly a
> couple of other daemons) is running then PC-NFS should be able to see
> the NFS daemon running on Linux and mount the shares.  Use the "man
> nfsd" to learn more.  "man mount" might even have references to other
> NFS man pages to read (on Linux of course).  Lastly, there's the Linux
> NFS HOWTO which you can find at the Linux Documentation Project
> (http://www.linuxdoc.org) web site.
>
> Too bad Samba is not an alternative for your because that would be MUCH
> cheaper than buying PC-NFS licenses.  You can always run a free NFS
> server on Windows and mount the NFS exports into your Linux filesystem.
>
> Good luck!
>
> Peace....
>
> Tom
>
>
> Sent via Deja.com http://www.deja.com/
> Before you buy.



------------------------------

Crossposted-To: alt.os.linux,comp.os.linux.x
Subject: Re: Is Mandrake Really Red Hat...
From: [EMAIL PROTECTED] (Kenneth Rørvik)
Date: Mon, 21 Aug 2000 16:54:22 GMT

[EMAIL PROTECTED] (Peter T. Breuer) wrote in
<8nrlu4$fv0$[EMAIL PROTECTED]>: 

>: Please explain :)
>
>Negative is negative. What's to explain? 

Well, you just did below, thanks :) 

>If you attempt to reorder
>machine instructions to gain advantage for a particular architecture,
>you will not be optimizing it for a second architecture. In fact
>you will be specializing the code and thereby rendering ineffective
>the strategies used by the second architecture to "optimize on the
>fly", because those strategies are aimed at the general case.
>
>The effect is particularly noticable wrt i586 and i686 architectures.
>That's because the i686 contains gigantic internal logic aimed at
>optimizing 386 code dynamically.  Any attempt to second guess it at
>compile time (or, indeed, replace the 386 code with specialized 586
>code) disadvantages that internal mechanism.  I'm tallking about the
>branch heuristics here in particular.
>
>You've also got to consider the pipeline optimizations and instruction
>reorderings that can take place.  It'll be more difficult to reorder
>specialized instructions that generic ones (the RISC/CISC effect).
>
>There is also wastage due to the different memory alignment schemes,
>but I don't think it will be nocicable going 586 to 686.
>
>You'd win again on a 787 architecture, though (:-).


-- 
Kenneth Rørvik          91841353/22718452
Steenstrupsgate 5 B     [EMAIL PROTECTED]
0554 OSLO               home.no.net/stasis

------------------------------

From: [EMAIL PROTECTED]
Subject: Re: which libs for CAD rpogram?
Date: Mon, 21 Aug 2000 16:48:24 GMT

check www.mesa3d.org and www.opencascade.org
between those two you should stay buisy for a while.

hope htat helps

        phrostie

In article <[EMAIL PROTECTED]>,
  Marcus Woletz <[EMAIL PROTECTED]> wrote:
> Hello all,
>
> I'm trying to rewrite a simple CAD program for GNU/Linux.
> I've developed the program a few years ago for OS/2.
> Now I've the problem to find a apropriate graphics API.
> The lib should at least have the following functions:
>
> 1.) Path creation
> 2.) Area creation
> 3.) rotated text display
> 4.) bounding box calculation
> 5.) Hit detection
>
> it doesn't matter if the lib is for kde, gnome or X.
>
> I've looked at dgs, libart and tcl/tk.
> dgs has all the needed features, but fails
> because of ugly/buggy font display.
> (Any other GPLed DPS-lib for Linux??)
> In libart I've found no possibility
> to display text.
> tcl/tk has no feature to display rotated
> text.
>
> Or should I look at similar apps like
> killustrator and rewrite the specific parts?
>
> I appreciate any hints.
> It's urgent because I really want to build a
> ECAD program for GNU/Linux and need your help!
>
> With regards
> Marcus
>

--
phrostie
Oh I have slipped the surly bonds of dos
and danced the skies on Linux silvered wings
http://www.nownetworks.com/~phrostie/cad-tastrafy


Sent via Deja.com http://www.deja.com/
Before you buy.

------------------------------

From: "Guennadi V. Liakhovetski" <[EMAIL PROTECTED]>
Crossposted-To: 
comp.unix.programmer,comp.unix.questions,comp.unix.misc,comp.os.linux.networking
Subject: Re: remote host
Date: Mon, 21 Aug 2000 17:40:29 +0100

Thanks to all! I was already thinking about repeatedly remotely enquiring
the previous host until you reach the first one, something like
setenv BLOGIN `rsh B whoami`
rsh B who | grep BLOGIN
...

Now can choose from a number of solutions. Thanks!
Guennadi
___

Dr. Guennadi V. Liakhovetski
Sheffield Centre for Earth Observation Science
Department of Applied Mathematics
University of Sheffield
Hicks Building, Hounsfield Road
Sheffield S3 7RH
phone: +44-(0)114-222-3798
fax:   +44-(0)114-222-3739
email: [EMAIL PROTECTED]



------------------------------

From: Matthias Arndt <[EMAIL PROTECTED]>
Subject: Re: WWW: TuxJournal online
Date: Mon, 21 Aug 2000 11:12:41 +0200
Reply-To: [EMAIL PROTECTED]

Matthias Arndt wrote:

Hello folks,
> -----BEGIN PGP SIGNED MESSAGE-----
> 
> 
> Hello folks,
> 
> the Tux Journal is now online.
> It is a Linux ezine on the web and it features my articles about Linux.
> However, some of the articles at the Tux Journal may be available at the
> very good Linux Gazette.
> 
> http://www.geocities.com/matthiasarndt/linux

Sorry, the link above is not correct. 
The right link to the TuxJournal is:
http://www.geocities.com/matthiasarndt/tuxjournal/

Matthias Arndt
-- 

... Original mail by Matthias Arndt
    IMPORTANT: Mail any replies to: [EMAIL PROTECTED]
    my ICQ number is: 40358321

------------------------------

Crossposted-To: alt.os.linux,comp.os.linux.x
Subject: Re: Is Mandrake Really Red Hat...
From: Johan Kullstam <[EMAIL PROTECTED]>
Date: 21 Aug 2000 13:01:18 -0400

[EMAIL PROTECTED] (Kenneth Rørvik) writes:

> [EMAIL PROTECTED] (Johan Kullstam) wrote in
> <[EMAIL PROTECTED]>: 
> 
> 
> >> Oh well, but the "optimization" would still apply to the later
> >> generation Pentiums, plus K5, K6's and up.
> >
> >this assumption would be wrong.
> >
> >> Even if the performance gain is
> >> minimal. 
> >
> >it's not minimal -- it's *negative*.
> 
> Please explain :)

it may be minimal, but it's still negative.  compiling stuff with
-march=pentium can give an executable which runs slower on an i686
than if it had been compiled for pentiumpro or even i486.

the i586 is a weird arch.  iirc it has two integer arithmetic
pipelines, call them A and B.  one of which, say A, can do any integer
type operation, the other, B, can do some but not all.  in addition
they can't be hitting the same register &c.  in scheduling for the
i586, you want to keep both pipelines busy.  this can mean doing some
things in a rather roundabout way in order to exploit B because A is
busy doing stuff.

the i386 and i486 only have one integer unit.  the i586 tricks are not
useful.

the i686 has three pipelines.  moreover, it has internal on-the-fly
scheduling op-code decoding to keep the pipes happy.  i586 tricks are
not needed and can even be counterproductive.

if you do not know, it's probably best to schedule to a i486.  it'll
run everywhere and i486 and i686 will both be reasonably happy.

most of the time the performance diffs are tiny.  however, i've
noticed with my own code that mis-tuning can cost you.  i've got hit
by extra 10% on a number crunching program (turbo codec simulation) by
compiling for i586 vs i386, i486 or i686 and running on a pentiumpro.
this was with some version egcs.  last i tried it with gcc 2.95.2 they
were all the same.  ymmv.

if you are considering running this on an i686, i find the whole i586
optimization thing to be dubious at best and slightly harmful at
worst.

-- 
J o h a n  K u l l s t a m
[[EMAIL PROTECTED]]
sysengr

------------------------------

From: [EMAIL PROTECTED]
Subject: Re: Installing 2nd hard disk
Crossposted-To: comp.os.linux.hardware,comp.os.linux.help
Date: Mon, 21 Aug 2000 13:10:57 +0100

[EMAIL PROTECTED] did eloquently scribble:
> Peter T. Breuer <[EMAIL PROTECTED]> did eloquently scribble:
>> : The whole point of the mount command is to allow for an extendable
>> : filesystem without the need to reboot!

>> I think you're confusing two parts of the software universe.

> I think you're right...
> :)
> My mistake.

Looks like I was wrong to admit I was wrong.
I *thought* I'd managed to add a disk and fdisk it without rebooting.
(It was so long ago, I though I was mistaken. Now I know I wasn't).

I hereby retract my "I think you're right. :) My mistake"
-- 
______________________________________________________________________________
|   [EMAIL PROTECTED]   |                                                 |
|Andrew Halliwell BSc(hons)| "The day Microsoft makes something that doesn't |
|            in            |  suck is probably the day they start making     |
|     Computer science     |  vacuum cleaners" - Ernst Jan Plugge            |
==============================================================================

------------------------------

From: moonie;) <[EMAIL PROTECTED]>
Crossposted-To: alt.os.linux,comp.os.linux.x
Subject: Re: Is Mandrake Really Red Hat...
Date: Mon, 21 Aug 2000 13:06:30 -0400

On Mon, 21 Aug 2000, Peter T. Breuer wrote:
>In comp.os.linux.misc Kenneth Rorvik <[EMAIL PROTECTED]> wrote:
>: [EMAIL PROTECTED] (Johan Kullstam) wrote in
>: <[EMAIL PROTECTED]>: 
>
>
>:>> Oh well, but the "optimization" would still apply to the later
>:>> generation Pentiums, plus K5, K6's and up.
>:>
>:>this assumption would be wrong.
>:>
>:>> Even if the performance gain is minimal. 
>:>
>:>it's not minimal -- it's *negative*.
>
>: Please explain :)
>
>Negative is negative. What's to explain? If you attempt to reorder
>machine instructions to gain advantage for a particular architecture,
>you will not be optimizing it for a second architecture. In fact
>you will be specializing the code and thereby rendering ineffective
>the strategies used by the second architecture to "optimize on the
>fly", because those strategies are aimed at the general case.
>
>The effect is particularly noticable wrt i586 and i686 architectures.
>That's because the i686 contains gigantic internal logic aimed at
>optimizing 386 code dynamically.  Any attempt to second guess it at
>compile time (or, indeed, replace the 386 code with specialized 586
>code) disadvantages that internal mechanism.  I'm tallking about the
>branch heuristics here in particular.
>
>You've also got to consider the pipeline optimizations and instruction
>reorderings that can take place.  It'll be more difficult to reorder
>specialized instructions that generic ones (the RISC/CISC effect).
>
>There is also wastage due to the different memory alignment schemes,
>but I don't think it will be nocicable going 586 to 686.
>
>You'd win again on a 787 architecture, though (:-).
>
>Peter

If this argument is correct then why did GCC go to the PGCC code for the latest
release of the GCC?  If your argument is correct then ALL of Linux just took a
big leap backwards.  
--
moonie ;)

Registered Linux User #175104

KDE2
Kernel 2.4.0-test5
XFree86 4.0 Nvidia .94 drivers
RAID 0 Stripped
Test-Pilots-R-Us ;)


------------------------------


** FOR YOUR REFERENCE **

The service address, to which questions about the list itself and requests
to be added to or deleted from it should be directed, is:

    Internet: [EMAIL PROTECTED]

You can send mail to the entire list (and comp.os.linux.misc) via:

    Internet: [EMAIL PROTECTED]

Linux may be obtained via one of these FTP sites:
    ftp.funet.fi                                pub/Linux
    tsx-11.mit.edu                              pub/linux
    sunsite.unc.edu                             pub/Linux

End of Linux-Misc Digest
******************************

Reply via email to