Re: Interested about number crunching in Ada

[Gary Scott <garylscott@sbcglobal.net>]

anon wrote:

> Even though Ada does have a few packages that are interesting in 
> numeric code.  The idea that any language such as Ada is better than 
> FORTRAN will not go over very well. Except for college project or 
> class assignments, that might give you something to do.  But in the 
> real world, it will not fly. To fully understand this try looking at the 
> history of SISAL (see below for definition).
> 
> There are a number of other High Performance Computing (HPC) 
> languages that were design back in the 60s, 70s, 80s, and 90s.  But 
> most have since died off because the lack of funding and previous 
> work done in FORTRAN and C. LISP has stay around but it role 
> was altered to mostly non-HPC status.
> 
> As for Ada:
> 
> First, is the acceptance of Ada.  In the world of mathematic FORTRAN 
> was design to and rules that universe.  Non-GNU FORTRAN compilers 
> are optimized for mathematic while Ada is not. And until that changes 
> most high performance numeric programmers will not accept Ada in that 
> world. Outside of FORTRAN they use pure "C" and maybe Lisp.  They 
> also say no to "c++" or any other languages.
> 
> Second, most programmer do not want to spend the 100s to 1000s of 
> hours to translate or convert the libraries that are written in FORTRAN 
> to any other language including Ada. That is around 50+ years worth of 
> library source code. And that does not include the time and expense of 
> getting the copyright and or Software Patents rights to do the 
> translation. Plus, translating any code from one language to another is 
> simply boring for most programmers. 
> 
> Plus, it is a lot easier to write and understand code that is wriiten in 
> the same language.  So, for libraries that are coded in FORTRAN means 
> the project languages needs to be in FORTRAN. And that's the way HPC 
> committees like it.
> 
> Now, for High Performance Computing projects, well it hard to find open 
> source projects that deal with mathematic.  In todays world, normally 
> you must be hired and move up to the position in HPC. Jobs in the 
> fields of Aerospace, Weather, Oceanic Research, and Medical are some 
> of the primary fields. All of these require knowledge in other fields that 
> are not commonly known to most programmers. Nornally, it people in 
> these specialize fields that become programmers and they use what is 
> commonly use for programming aka FORTRAN or C. 
> 
> Some of the not so commonly known job fields are Web Servers, and 
> Neural Networks but these are not driving by higher performance 
> mathematic, but by Database and File Accessing. 
> 
> But for the best information on High Performance Mathematic 
> Computing, check with your local college or university computer 
> department in a couple of weeks after the fall semester starts. Give 
> a few days for the school to calm down into the semester routine 
> before asking.
> 
> With the newer computers having dual processors you would think that 
> most would want to see HPC coding on these desk top. But business 
> owner are hard to adopt new program or computing paradigms. And 
> scientists want 16 to 512 processors aka a supercomputer to play with 
> so to them a dual processors is only building blocks toward that design. 
> And with a price tag to match they are hard to the average programmer to 
> buy or build.
> 
> 
> As for SISAL (definition): 
> 
> SISAL is programming language that automatically parallelizes code for 
> parallel computers, but still works on single processors. it is a 
> functional language that is hightly efficient for numerical computation. 
> The Sisal project was based until the early 2000's, at the Lawrence 
> Livermore National Laboratory, but it has been canceled there. You can 
> still find the source code for SISAL on the internet. 
> 
> 
> 
> 
> In <1187235764.909133.180650@19g2000hsx.googlegroups.com>,  holst <henrikholst80@gmail.com> writes:
> 
>>Hi!
>>
>>I have stumbled upon Ada95 and I have found that a recent addition was
>>made to the language standard [1]. An addition I, a student of
>>scientific computing, are highly interested in.
>>
>>What is the best online resource to get into the core of the new high
>>performance vector and matrix features? Does there exist some book
>>(yet) which covers this area? Or any other field which might be
>>related to me (concurrency, Fortran bindings etc.)? I know C and
>>Pascal good and I have a good start into Fortran 90/95.
>>
>>I applicate your time and help. I hope that, with a push in the right
>>direction I will be a productive "Ada numerics hacker" in a near
>>future. :-)
>>
>>[1] http://www.ada-auth.org/cgi-bin/cvsweb.cgi/AIs/AI-00296.TXT
>>
>>--
>>Henrik Holst, Sweden
>>http://www.nada.kth.se/~holst/contact.shtml
>>Number of productive hours in C++: <zero> of <too many>
>>
> 
You forgot to mention that Fortran has not stood still.  It is a 
moving/advancing target.  Yes it has baggage.  But it is evolving quite 
well, within those constraints.

-- 

Gary Scott
mailto:garylscott@sbcglobal dot net

Fortran Library:  http://www.fortranlib.com

Support the Original G95 Project:  http://www.g95.org
-OR-
Support the GNU GFortran Project:  http://gcc.gnu.org/fortran/index.html

If you want to do the impossible, don't hire an expert because he knows 
it can't be done.

-- Henry Ford