Technology and Trends
 USENET Archives
  
Path: gmdzi!unido!mcsun!uunet!samsung!uakari.primate.wisc.edu!doug.cae.wisc.edu!epperly
From: epp...@osnome.che.wisc.edu (Tom Epperly)
Newsgroups: gnu.gcc.help
Subject: How does gcc compare?
Message-ID: <EPPERLY.91May2121516@osnome.che.wisc.edu>
Date: 2 May 91 17:15:16 GMT
Organization: University of Wisconsin-Madison, Chemical Engineering
Lines: 3

How does gcc compare to the DECstation 3100 Ultrix 4.1 C compiler?
How does it compare to other compilers on other machines?
				Tom

Path: gmdzi!unido!mcsun!uunet!samsung!caen!uakari.primate.wisc.edu!doug.cae.wisc.edu!epperly
From: epp...@osnome.che.wisc.edu (Tom Epperly)
Newsgroups: gnu.gcc.help
Subject: Re: How does gcc compare?
Message-ID: <EPPERLY.91May2135459@osnome.che.wisc.edu>
Date: 2 May 91 18:54:59 GMT
References: <EPPERLY.91May2121516@osnome.che.wisc.edu>
Organization: University of Wisconsin-Madison, Chemical Engineering
Lines: 12
In-reply-to: epperly@osnome.che.wisc.edu's message of 2 May 91 12:15:16

Perhaps, I should ask some more specific questions.  All of the
following questions are in reference to a DECstation 3100 running
Ultrix 4.1.

	1) Should I use gcc or cc to compile X11R4?
	2) Can cc include libraries compiled with gcc(eg. -lX11 when
	   X11 is compiled with gcc)?
	3) Should I use gcc or cc to compile GNU Emacs?
	4) Should I use gcc or cc to compile TeX?

				Tom Epperly
				epp...@osnome.che.wisc.edu

Path: gmdzi!unido!mcsun!uunet!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!
magnus.acs.ohio-state.edu!csn!boulder!news!grunwald
From: grun...@foobar.colorado.edu (Dirk Grunwald)
Newsgroups: gnu.gcc.help
Subject: Re: How does gcc compare?
Message-ID: <1991May2.235022.11031@colorado.edu>
Date: 2 May 91 23:50:22 GMT
References: <EPPERLY.91May2121516@osnome.che.wisc.edu>
	<EPPERLY.91May2135459@osnome.che.wisc.edu>
Sender: ne...@colorado.edu (The Daily Planet)
Reply-To: grun...@foobar.colorado.edu
Organization: University of Colorado at Boulder
Lines: 33
In-Reply-To: epperly@osnome.che.wisc.edu's message of 2 May 91 18:54:59 GMT
Nntp-Posting-Host: foobar.colorado.edu

>>>>> On 2 May 91 18:54:59 GMT, epp...@osnome.che.wisc.edu (Tom Epperly) said:

TE> Perhaps, I should ask some more specific questions.  All of the
TE> following questions are in reference to a DECstation 3100 running
TE> Ultrix 4.1.

TE> 	1) Should I use gcc or cc to compile X11R4?
--
cc

TE> 	2) Can cc include libraries compiled with gcc(eg. -lX11 when
TE> 	   X11 is compiled with gcc)?
--
yes

TE> 	3) Should I use gcc or cc to compile GNU Emacs?
--
either. I have more luck with gcc, because the optimization isn't as broken
as MIPS CC.

TE> 	4) Should I use gcc or cc to compile TeX?
--
Again, I used gcc because optimization works.


gcc 1.39 is about 10% slower than 'cc' at corresponding optimization
levels.

Tests with alpha releases of 2.0 (with a very preliminary mips port)
show that it produces as efficient code as the MIPS compiler, unless
you use -O4 on the MIPS compiler (which isn't documented on my Ultrix
machine anyway).  I expect this will improve as people refine the mips
port.

Path: gmdzi!unido!fauern!ira.uka.de!sol.ctr.columbia.edu!lll-winken!
uunet!rayssd!fjr
From: f...@sgfb.ssd.ray.com (Fred J. Roeber)
Newsgroups: gnu.gcc.help
Subject: Re: How does gcc compare?
Message-ID: <405@sgfb.ssd.ray.com>
Date: 6 May 91 02:26:31 GMT
References: <EPPERLY.91May2121516@osnome.che.wisc.edu> 
<EPPERLY.91May2135459@osnome.che.wisc.edu> <1991May2.235022.11031@colorado.edu>
Sender: ne...@rayssd.ssd.ray.com
Organization: Raytheon Company, Portsmouth, RI
Lines: 24

In article <1991May2.2...@colorado.edu> grun...@foobar.colorado.edu
   writes:
>
>TE> 	3) Should I use gcc or cc to compile GNU Emacs?
>--
>either. I have more luck with gcc, because the optimization isn't as broken
>as MIPS CC.
>
>gcc 1.39 is about 10% slower than 'cc' at corresponding optimization
>levels.
>
>Tests with alpha releases of 2.0 (with a very preliminary mips port)
>show that it produces as efficient code as the MIPS compiler, unless
>you use -O4 on the MIPS compiler (which isn't documented on my Ultrix
>machine anyway).  I expect this will improve as people refine the mips
>port.

It isn't clear from the response whether Dirk is using a MIPS machine
or a DECSTATION.  It is my understanding that DEC wrote their own C
compiler for the DECSTATION; they don't use the MIPS C compiler.  Also,
everything I have heard/seen indicates that the MIPS compiler is one
of the best ones out there.  While the GNU compilers beat most vendor
supplied compilers in terms of code efficiency I would be very surprised
if they beat MIPS.  Can anyone clarify this?

Path: gmdzi!unido!mcsun!uunet!tut.cis.ohio-state.edu!unreplyable!garbage
From: tie...@CYGNUS.COM (Michael Tiemann)
Newsgroups: gnu.gcc.help
Subject: How does gcc compare?
Message-ID: <9105060632.AA00788@cygnus.com>
Date: 6 May 91 06:32:53 GMT
References: <405@sgfb.ssd.ray.com>
Sender: dae...@tut.cis.ohio-state.edu
Reply-To: tie...@cygnus.com
Distribution: gnu
Organization: Cygnus Support, Palo Alto CA; Phone +1 415 322 3811
    It isn't clear from the response whether Dirk is using a MIPS machine
    or a DECSTATION.  It is my understanding that DEC wrote their own C
    compiler for the DECSTATION; they don't use the MIPS
Lines: 15

DEC did not write their own compiler for the MIPS chip.  They used an
early, less tuned version of the MIPS compiler.  But I'm curious why
you think MIPS should have a good compiler when you can believe that
Sun, DEC, HP, and IBM all lose.  The way I see it, the fact that MIPS
competes in the same fashion (proprietary software), using the same
tools (compilers, editors, debuggers) and resources (people) as these
other companies, it would surprise me more if their compiler were
qualitatively or quantitatively different.

Sure, MIPS spends a lot on marketing their technology, but that's
their business.  Our business is to compete with them by delivering
twice the product at a fifth the cost.  How you spend your money,
whether you prefer to feed marketing or development, is your business.

Michael

Path: gmdzi!unido!mcsun!uunet!elroy.jpl.nasa.gov!usc!samsung!
uakari.primate.wisc.edu!caen!hellgate.utah.edu!csn!boulder!news!grunwald
From: grun...@foobar.colorado.edu (Dirk Grunwald)
Newsgroups: gnu.gcc.help
Subject: Re: How does gcc compare?
Message-ID: <1991May7.151352.28069@colorado.edu>
Date: 7 May 91 15:13:52 GMT
References: <EPPERLY.91May2121516@osnome.che.wisc.edu>
	<EPPERLY.91May2135459@osnome.che.wisc.edu>
	<1991May2.235022.11031@colorado.edu> <405@sgfb.ssd.ray.com>
Sender: ne...@colorado.edu (The Daily Planet)
Reply-To: grun...@foobar.colorado.edu
Organization: University of Colorado at Boulder
Lines: 19
In-Reply-To: fjr@sgfb.ssd.ray.com's message of 6 May 91 02:26:31 GMT
Nntp-Posting-Host: foobar.colorado.edu


With Ultrix 4.0, DEC supplied Mips C V2.0. With Ultrix 3.x, DEC use
Mips V1.something. DEC is supplying their own compiler suite with (I
think) Ultrix 4.2. This is good, because although the MIPS compiler
produces good code, it's not the most bug-free compiler I've ever
used.  It's also not ANSI.

My comment about alpha GCC-2 performance is based on running a single
floating-point intensive program (VM_pRAY) under Mips CC 2.1 and Gnu C
with various compiler options. GCC2 did as well or better as Mips
CC2.1, until I enabled -O4 on the Mips CC (this does interprocedural
optimization, I think -- it's not documented).

But, the alpha GCC crashes often, and the mips output .o size is 2x
the previous version. A lot remains to be done, but it'll be very
competative with the MIPS product line when it's stable (which it
isn't, so don't ask). This is more or less what Tiemann has been
posting about the SPARC port as well -- it will out-perform Suns C
compiler.

Path: gmdzi!unido!ztivax!sof3
From: sof3@ztivax.UUCP (Walter Meyer)
Newsgroups: gnu.gcc.help
Subject: Re: How does gcc compare?
Message-ID: <5348@ztivax.UUCP>
Date: 29 May 91 13:08:38 GMT
References: <1991May7.151352.28069@colorado.edu>
Organization: Siemens AG, Munich, W-Germany
Lines: 20

On 7May91 grun...@foobar.colorado.edu (Dirk Grunwald) writes:
> 
> But, the alpha GCC crashes often, and the mips output .o size is 2x
> the previous version. A lot remains to be done, but it'll be very
> competative with the MIPS product line when it's stable (which it
> isn't, so don't ask). This is more or less what Tiemann has been
> posting about the SPARC port as well -- it will out-perform Suns C
> compiler.

OK, so I take it that gcc 2.0 is not yet stable and available.  Any
wild ass guesses as to its availability date?

I need it on SVR4.  Any chance 2.0 will work on SVR4?  1.39 sure as
hell doesn't (SVR4 does NOT use the a.out object file format, it
uses ELF which is conceptually similar but totally different).  Will
the suite of gcc & related products work with SVR4?

David Smyth
david...@ztivax.siemens.com
(Follow up to this newsgroup, mail is again hopeless for a few days)

Path: gmdzi!unido!mcsun!uunet!cs.utexas.edu!milano!uudell!bigtex!james
From: ja...@bigtex.cactus.org (James Van Artsdalen)
Newsgroups: gnu.gcc.help
Subject: Re: How does gcc compare?
Message-ID: <62012@bigtex.cactus.org>
Date: 30 May 91 03:38:44 GMT
References: <1991May7.151352.28069@colorado.edu> <5348@ztivax.UUCP>
Reply-To: ja...@bigtex.cactus.org (James Van Artsdalen)
Organization: Institute of Applied Cosmology, Austin TX
Lines: 17

In <5348@ztivax.UUCP>, sof3@ztivax.UUCP (Walter Meyer) wrote:

> OK, so I take it that gcc 2.0 is not yet stable and available.

It is not stable in the sense that changes are still being made,
and much testing remains to be done.

> Any chance 2.0 will work on SVR4?  1.39 sure as hell doesn't (SVR4
> does NOT use the a.out object file format, it uses ELF which is
> conceptually similar but totally different).

Oddly enough, I built gcc 1.39+ on SysVr4 this afternoon.  It should work,
except that you get no debugging information.  Someone was working on
DWARF output for gcc 2, but I don't know the status of that.
-- 
James R. Van Artsdalen          ja...@bigtex.cactus.org   "Live Free or Die"
Dell Computer Co    9505 Arboretum Blvd Austin TX 78759         512-338-8789

Path: gmdzi!unido!mcsun!uunet!zaphod.mps.ohio-state.edu!
sample.eng.ohio-state.edu!purdue!haven.umd.edu!mimsy!dftsrv!tsingle
From: tsi...@sunland.gsfc.nasa.gov (Tim Singletary)
Newsgroups: gnu.gcc.help
Subject: Re: How does gcc compare?
Message-ID: <TSINGLE.91May30105547@sunland.gsfc.nasa.gov>
Date: 30 May 91 14:55:47 GMT
References: <1991May7.151352.28069@colorado.edu> <5348@ztivax.UUCP>
	<62012@bigtex.cactus.org>
Sender: ne...@dftsrv.gsfc.nasa.gov
Organization: Goddard Space Flight Center, Greenbelt, MD
Lines: 13

In article <62...@bigtex.cactus.org> ja...@bigtex.cactus.org (James Van Artsdalen) 
writes:
   Oddly enough, I built gcc 1.39+ on SysVr4 this afternoon.  It should work,
   except that you get no debugging information.

Does this mean that gdb can't debug a gcc compiled program on SVR4?
(Does gdb even compile-and-run on SVR4???)



--
Tim Singletary, August Automation Inc., (301) 286-7942
--
Why does The Government care whether or not I wear a seatbelt when I drive?

Path: gmdzi!unido!mcsun!uunet!cis.ohio-state.edu!tut.cis.ohio-state.edu!
unreplyable!garbage
From: youn...@V6550C.NRL.NAVY.MIL (Eric Youngdale)
Newsgroups: gnu.gcc.help
Subject: Re: How does gcc compare?
Message-ID: <009497C9.C3EA25E0.12252@v6550c.nrl.navy.mil>
Date: 1 Jun 91 22:48:57 GMT
Sender: dae...@tut.cis.ohio-state.edu
Distribution: gnu
Organization: GNUs Not Usenet
Lines: 29

>In article <62...@bigtex.cactus.org> ja...@bigtex.cactus.org 
>(James Van Artsdalen) writes:
>   Oddly enough, I built gcc 1.39+ on SysVr4 this afternoon.  It should work,
>   except that you get no debugging information.
>
>Does this mean that gdb can't debug a gcc compiled program on SVR4?
>(Does gdb even compile-and-run on SVR4???)
>

	Would it work to use GAS with the patches to generate COFF?  The SVR4
system that we have has a cof2elf utility, and according to the man page, 
"Some of the debugging information is discarded", which implies not all of it. 
Actually,  you may not even need to use cof2elf, since the native ld accepts a
COFF file for input (issuing an annoying "notice" messages whenever it sees
one.  I do not know what happens if the COFF file has debugging information in
it).

	As far as gdb is concerned, the vendor that sold us the SVR4 system
(Dell) also included gdb and emacs, and I can definitely say that both of
these seem to work quite nicely under release 2.01.  I do not know what changes
they had to make to get them to work, but they do indeed work.  I have not seen
any sources to the patches, and I have been too busy just getting the machine
up and running to ask them about this.

------------------------------------------------------------------
Eric Youngdale          INTERNET:  YOUN...@V6550C.NRL.NAVY.MIL
Naval Research Lab     	    SPAN:  11.13 (or 11277::)
Washington, DC	       FLAME-NET:  NLA0:
	Disclaimer:  The opinions expressed herein are my own and not
necessarily those of my employer.