From mboxrd@z Thu Jan  1 00:00:00 1970
X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me
X-Spam-Level: 
X-Spam-Status: No, score=-1.9 required=5.0 tests=BAYES_00 autolearn=ham
	autolearn_force=no version=3.4.4
X-Google-Language: ENGLISH,ASCII-7-bit
X-Google-Thread: 103376,546c9d71cfb0c1ee
X-Google-Attributes: gid103376,public
X-Google-ArrivalTime: 2004-03-14 10:12:41 PST
Path: 
 archiver1.google.com!news1.google.com!newsfeed.stanford.edu!bloom-beacon.mit.edu!news-out.cwix.com!newsfeed.cwix.com!border1.nntp.ash.giganews.com!border2.nntp.ash.giganews.com!nntp.giganews.com!in.100proofnews.com!in.100proofnews.com!tiscali!newsfeed1.ip.tiscali.net!news.worldonline.be!not-for-mail
From: Ludovic Brenta <ludovic.brenta@insalien.org>
Newsgroups: comp.lang.ada
Subject: Re: [Announcement] New Project: ASIS for GNAT
Date: 14 Mar 2004 19:16:09 +0100
Organization: Worldonline Belgium
Sender: lbrenta@deuteronomy
Message-ID: <87brmzs39y.fsf@insalien.org>
References: <1252369.os30E2mMWi@linux1.krischik.com>
 <3zl4c.17228$mM.123491@attbi_s02>
 <46b8b50a.0403140746.95e9c51@posting.google.com>
NNTP-Posting-Host: ppp-62-235-77-194.tiscali.be
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
X-Trace: news.worldonline.be 1079287955 17499 62.235.77.194 (14 Mar 2004
 18:12:35 GMT)
X-Complaints-To: abuse@worldonline.be
NNTP-Posting-Date: Sun, 14 Mar 2004 18:12:35 +0000 (UTC)
User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.3
Xref: archiver1.google.com comp.lang.ada:6322
Date: 2004-03-14T19:16:09+01:00
List-Id: <comp.lang.ada>

rolf.ebert_nospam_@gmx.net (Rolf Ebert) writes:

> I ran gcc 3.3.3 and 3.4 prereleases on the ACATS test suite at several
> optimization levels on Linux and Windows during the last few weeks. 
> As I cannot attach a pdf or Excel file here, I am willing to send the
> summary sheet to anybody who is interested.
> 
> In short: the 3.4 prereleases are considerably better than the 3.3.3
> release.  The number of failures dropped from 24 to 7 (Linux) or 5
> (Windows) out of more than 2300 test files.  The remaining 7 errors on
> Linux are all related to the complex calculation library and don't
> show up on Windows.  The number of failures is stable at -O0 to -O2. 
> Only further optimizations and inlining (-O3, -gnatN,
> -funroll-all-loop, -finline-functions) trigger additional failures.


Thank you for these data points.  It has been my experience with 3.15p
that -gnatVa tends to trigger ICE's quite often.  As I find the
additional checks valuable, I would suggest that running the ACATS
tests with -gnatVa (and possibly -gnatwa, too) might be worthwhile.

-- 
Ludovic Brenta.