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=0.4 required=5.0 tests=BAYES_00,FORGED_MUA_MOZILLA
	autolearn=no autolearn_force=no version=3.4.4
X-Google-Thread: 103376,81bb2ce65a3240c3
X-Google-NewGroupId: yes
X-Google-Attributes: gida07f3367d7,domainid0,public,usenet
X-Google-Language: ENGLISH,ASCII-7-bit
Received: by 10.68.227.230 with SMTP id sd6mr5475878pbc.8.1335431785809;
        Thu, 26 Apr 2012 02:16:25 -0700 (PDT)
Path: 
 r9ni100088pbh.0!nntp.google.com!news1.google.com!goblin3!goblin.stu.neva.ru!de-l.enfer-du-nord.net!feeder2.enfer-du-nord.net!weretis.net!feeder4.news.weretis.net!news.teledata-fn.de!newsfeed.arcor.de!newsspool1.arcor-online.net!news.arcor.de.POSTED!not-for-mail
Date: Thu, 26 Apr 2012 11:16:24 +0200
From: Georg Bauhaus <rm.dash-bauhaus@futureapps.de>
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.7;
 rv:11.0) Gecko/20120327 Thunderbird/11.0.1
MIME-Version: 1.0
Newsgroups: comp.lang.ada
Subject: Re: What would you like in Ada202X?
References: <3637793.35.1335340026327.JavaMail.geo-discussion-forums@ynfi5>
 <jn9hp2$2qv$1@dont-email.me>
 <4f98639a$0$6557$9b4e6d93@newsspool4.arcor-online.net>
 <jnb06k$v52$1@dont-email.me>
In-Reply-To: <jnb06k$v52$1@dont-email.me>
Message-ID: <4f991268$0$9522$9b4e6d93@newsspool1.arcor-online.net>
Organization: Arcor
NNTP-Posting-Date: 26 Apr 2012 11:16:24 CEST
NNTP-Posting-Host: 0e023849.newsspool1.arcor-online.net
X-Trace: 
 DXC=VN;Tad[^e:185[]]\]T081ic==]BZ:af>4Fo<]lROoR1nkgeX?EC@@0_VX:i1JEGk:PCY\c7>ejV8mH1IV7EQHS<e\:DgBcV:[;
X-Complaints-To: usenet-abuse@arcor.de
Content-Type: text/plain; charset=windows-1252; format=flowed
Content-Transfer-Encoding: 7bit
Date: 2012-04-26T11:16:24+02:00
List-Id: <comp.lang.ada>

On 26.04.12 10:15, Oliver Kleinke wrote:
>>>>> 2) parallel loops / functions
>>>> Array support in the direction of utilizing parallel
>>>> micro-micro-processing powers.
>>>
>>> Already present in GNAT, implicitly though.
>>
>> Where, please? I'm very curious, since what we have found to be working is
>> either tied to GCC's type attributes, or seems to be caused by the compiler
>> working in miraculous ways: small arrays/sequences of floats can be,
>> or may be, made operands of SIMD instructions on Intel compatible
>> hardware. Tricky, and hard to control, yet worth a factor 2.
>
> Have you discovered this yet?: http://gcc.gnu.org/onlinedocs/gnat_ugn_unw/Vectorization-of-loops.html

Thanks for the pointer; we have essentially tried everything
it says. I notice that the description is optimistic about SIMD
operators for "+", and "*", "adding and some of the multiplying".
We do in addition have a positive case for "/", but not on x86,
only on x86_64. Some other things worked on x86, but not on x86_64,
such as splitting loops in two, and using auxiliary data structures,
perhaps creating enough of independence. Nevertheless, all effects
seem unpredictable (I'm not blaming GNAT or GCC), and not yet near
to what Intel's Fortran does.