Path: csiph.com!v102.xanadu-bbs.net!xanadu-bbs.net!feeder.erje.net!eu.feeder.erje.net!weretis.net!feeder1.news.weretis.net!news.solani.org!.POSTED!not-for-mail
From: Melzzzzz <mel@zzzzz.invalid>
Newsgroups: comp.programming.threads
Subject: Re: About scimark2 ...
Date: Sat, 7 Jun 2014 01:58:43 +0200
Organization: None
Lines: 29
Message-ID: <lmtkjj$fgk$6@solani.org>
References: <lmt7tq$7jv$2@news.albasani.net> <lmtad8$kds$5@solani.org> <lmtcjl$h3v$1@news.albasani.net> <20140606155957.393@kylheku.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
X-Trace: solani.org 1402099123 15892 eJwFwQkBACAIBLBKvAfGAYX+EdxcwbhhcJiv75rUsHZd6SpiSNLLkIIhTw/FC/VRbT7Xdz4gehEk (6 Jun 2014 23:58:43 GMT)
X-Complaints-To: abuse@news.solani.org
NNTP-Posting-Date: Fri, 6 Jun 2014 23:58:43 +0000 (UTC)
X-User-ID: eJwNwgERADEIAzBLrNAy9AzwL+H/ErqOXoao4P5m5BXV1xqdk4WGDBhPGnuthsDtXMDe+QAMihBM
X-Newsreader: Claws Mail 3.9.3 (GTK+ 2.24.23; x86_64-pc-linux-gnu)
Cancel-Lock: sha1:3JeH5E5QMKTUbQLn0pnhbHODAe4=
X-NNTP-Posting-Host: eJwFwYkBgDAIA8CVypOI4wA1+4/gHYLGfZJgQtDx+tptTCgujI7YGm9oIvZmodQ8jiTvuz8OvRCe
Xref: csiph.com comp.programming.threads:2422

On Fri, 6 Jun 2014 23:11:59 +0000 (UTC)
Kaz Kylheku <kaz@kylheku.com> wrote:

> On 2014-06-06, aminer <aminer@toto.net> wrote:
> > I have benchmarked Delphi and FreePascal against GCC C++
> > and Microsoft C+ on scimark2, and i have found that without the GCC
> > -Ox optimization options the auto-vectorization will not be enabled
> > so Delphi and FreePascal and GCC C++ scored the same speed on
> > scimark2, but when you give -O1 or -O2,  GCC C++ uses
> > auto-vectorization and it scores almost 2X times the speed of
> > Freepascal or Delphi on scimark2 on x86.
> 
> Without -O the GNU compilers do not optimize at all. The generated
> code is quite bad.

I think that gcc performs auto vectorization only at -O3 .

> 
> If removing -O options brings the performance down to Delphi, and a
> mere -O1 can beat Delphi, that suggests that Delphi doesn't optimize,
> at least with regard to the program being tested.
> 
> Or something else is wrong, which makes the programs not comparable.

Most probably pascal benchmark uses more complex code. Eg OO features...

-- 
Click OK to continue...