🚀 go-pugleaf

RetroBBS NetNews Server

Inspired by RockSolid Light RIP Retro Guy

Article View: comp.unix.cray
Article #326

Re: Compilation of gcc/g++ on ymp-el?

#326
From: Tomislav Goles
Date: Wed, 28 Jun 1995 00:00
43 lines
2859 bytes
In article <teaDAux0H.9ns@netcom.com> tea@netcom.com (Tom Ace) writes:

   Newsgroups: comp.unix.cray,gnu.gcc
   Path: cs.utk.edu!cssun.mathcs.emory.edu!emory!gatech!news.sprintlink.net!noc.netcom.net!netcom.com!tea
   From: tea@netcom.com (Tom Ace)
   Organization: Netcom Online Communications Services (408-241-9760 login: guest)
   References: <DALort.Bpx@cnn.nas.nasa.gov> <3selil$hpf@fly.HiWAAY.net> <DAupv2.H8n@murdoch.acc.Virginia.EDU>
   Date: Wed, 28 Jun 1995 00:21:53 GMT
   Lines: 26
   Sender: tea@netcom12.netcom.com
   Xref: cs.utk.edu comp.unix.cray:2183 gnu.gcc:279

   In article <DAupv2.H8n@murdoch.acc.Virginia.EDU>
    jwa2n@galen.med.Virginia.EDU (James W. Adams) writes:

   Depends which Sun compiler you're talking about.  I was favorably impressed
   by g++'s object code and encouraged some friends to try it out.  They 
   installed gcc and benchmarked it, but they had a newer Sun C++ compiler than 
   I do.  It trounced g++; they tell me their code runs in about 40% more time 
   when compiled with g++.  This was about two months ago--I don't know the 
   name or version number of the Sun C++ compiler they were using, but they 
   were comparing it to g++ 2.6.3.  (To be fair, g++ optimization for Sparc 
   seems to do a better job in 2.7.0 than it did in 2.6.3.)

My benchmarks had g++ heavily outperform sun's 3.0 c++ compiler. Sun's
4.0 c++ compiler (their latest as far as I know) is closer to g++ but
g++ still easily outperforms it most of the time.

ps. I ran my test on sparcstation 5 and used the following gcc flags:
 -mno-epilogue -mv8 -msupersparc -O6 -fforce-mem -fforce-addr 

I tried many combinations of sun's flags but gcc/g++ always
outperformed sun (although in some cases by a very small margin).

Tom Goles
--

+-----------------------------------------------------------------------------+
| Tomislav Goles                      |  E-mail:   goles@cs.utk.edu           |
+-----------------------------------------------------------------------------+
| Justice, n.:                                                                |
|         A decision in your favor.                                           |
+-----------------------------------------------------------------------------+
Technical Details
Message-ID: <GOLES.95Jun28121330@duncan.cs.utk.edu>
Path: arc1-ca.pugleaf.net!unknown!not-for-mail
References: <DALort.Bpx@cnn.nas.nasa.gov> <3selil$hpf@fly.HiWAAY.net>

Show Full Headers
Complete Article Headers
X-Google-Language: ENGLISH,ASCII-7-bit
X-Google-Thread: f65d0,9306ff1bcb6e8b71
X-Google-Attributes: gidf65d0,public
X-Google-Thread: 10eb55,95341354f73a82f6
X-Google-Attributes: gid10eb55,public
From: goles@duncan.cs.utk.edu (Tomislav Goles)
Subject: Re: Compilation of gcc/g++ on ymp-el?
Date: 1995/06/28
Message-ID: <GOLES.95Jun28121330@duncan.cs.utk.edu>
X-Deja-AN: 105263327
distribution: inet
references: <DALort.Bpx@cnn.nas.nasa.gov> <3selil$hpf@fly.HiWAAY.net>
organization: University of Tennessee, Knoxville
newsgroups: comp.unix.cray,gnu.gcc