Linked by Thom Holwerda on Fri 13th Nov 2015 20:28 UTC
General Development

Today, the US Department of Energy announced that it had established a partnership with NVIDIA that would be enhancing the LLVM compiler collection. The goal will be to port an existing FORTRAN compiler that targets massively parallel GPUs. The results are expected to be released as open source in late 2016.

Cutting-edge research still universally involves Fortran; a trio of challengers wants in. While FORTRAN isn't a mainstream language, it's still heavily used in scientific computing, and there's lots of legacy code that relies on it. A lot of that code is maintained by people at the US National Labs, and the new project is being organized by staff at Lawrence Livermore, Sandia, and Los Alamos.

Order by: Score:
So ... GCC is dead but doesn't know it yet
by pfgbsd on Fri 13th Nov 2015 21:18 UTC
pfgbsd
Member since:
2011-03-12

No, they will not call it GNU/LLVM Fortran.

RMS will have a heart attack.

Reply Score: 1

Comment by Luminair
by Luminair on Fri 13th Nov 2015 23:08 UTC
Luminair
Member since:
2007-03-30

this is for the military supercomputers they're building. nuke sim / spying / hacking / tracking / ai / whatever

I assume the government's ultimate spying goal is to enter the world's data and spit out the answer to any question. kind of like google, but with everything you've ever said or done, on tap

Reply Score: 1

RE: Comment by Luminair
by The123king on Sat 14th Nov 2015 02:57 UTC in reply to "Comment by Luminair"
The123king Member since:
2009-05-28

Or maybe it's because the government's (pick (a western) one) missile control systems run on 40 year old hardware.

After all, the missiles themselves are nearly that old

Edited 2015-11-14 02:59 UTC

Reply Score: 1

RE[2]: Comment by Luminair
by Luminair on Sat 14th Nov 2015 10:09 UTC in reply to "RE: Comment by Luminair"
Luminair Member since:
2007-03-30

supercomputers for missile control, you're a character in wargames

Reply Score: 2

RE: Comment by Luminair
by Kochise on Sat 14th Nov 2015 08:21 UTC in reply to "Comment by Luminair"
Kochise Member since:
2006-03-03

"to enter the world's data and spit out the answer to any question"

42

Reply Score: 5

RE: Comment by Luminair
by Bill Shooter of Bul on Sat 14th Nov 2015 15:14 UTC in reply to "Comment by Luminair"
Bill Shooter of Bul Member since:
2006-07-14

Nuke sim ? Obviously yes.
Spying, hacking, tracking, ai?
Probably not.

If its physical related, probably FORTRAN. But, not likely anything else.

Not because the mighty FORTRAN, isn't the best choice, but because the kids don't understand the wisdom of FORTRAN.

Reply Score: 3

RE[2]: Comment by Luminair
by Kochise on Sat 14th Nov 2015 16:43 UTC in reply to "RE: Comment by Luminair"
Kochise Member since:
2006-03-03

And LISP

Reply Score: 2

RE: Comment by Luminair
by Doc Pain on Sun 15th Nov 2015 05:43 UTC in reply to "Comment by Luminair"
Doc Pain Member since:
2006-10-08

I assume the government's ultimate spying goal is to enter the world's data and spit out the answer to any question. kind of like google, but with everything you've ever said or done, on tap


Don't ask questions to a supercomputer when you're not prepared to accept the answer. :-)

Anecdote from times when there was a strong belief in "electron brains" that could answer all important questions:

The Politburo ordered the manufacturer robotron to build the GDR's most advanced and most powerful computer. It should be equipped with all resources necessary to improve the economical situation of the country and provide the government with the required steps. This action should confirm the leading role of the Politburo. The machine was built, and all the numbers, figures, facts, statistics and data was properly fed into the machine. After several days of crunching numbers, a card popped out of the card puncher *ding!* and it contained the only appropriate answer: "ABOLISH POLITBURO".

I'm sure this can be done again in Fortran. :-)

Reply Score: 2

Not mainstream?
by JLF65 on Sat 14th Nov 2015 05:36 UTC
JLF65
Member since:
2005-07-06

It certainly is for engineers. Every branch of engineering at most major universities includes three semesters of Fortran. Even many software groups work in Fortran - all MPEG codecs associated with each new spec have been done as Fortran, then converted with a Fortran-to-C converter after the specs are released.

Reply Score: 4

This is great news
by FortranMan on Sat 14th Nov 2015 13:01 UTC
FortranMan
Member since:
2011-12-21

I use Fortran all the time in my research, and look forward to another open source compiler being available, especially if it can target GPUs. They are still a real pain to access, but Fortran 2008 has built-in features that will make programming for them much easier.

Reply Score: 2

RE: This is great news
by Temcat on Sun 15th Nov 2015 14:32 UTC in reply to "This is great news"
Temcat Member since:
2005-10-18

What language would you use for your studies if not for all the existing Fortran-based software?

Reply Score: 2

RE[2]: This is great news
by FortranMan on Sun 15th Nov 2015 16:13 UTC in reply to "RE: This is great news"
FortranMan Member since:
2011-12-21

Its not actually the existing software that brings me to Fortran; rather, its the ease of writing performant code for numerical calculations.

Additionally, most mechanical engineering curricula need students to code, but C/C++ is too hard for students who otherwise are uninterested software development (pointers are hard...) and alternatives also have problems. Matlab is not a good language, and is also a commercial product with performance problems. Java, python, C#, Javascript, and many others are not fast enough.

In the end, Fortran is easy to learn, easy to use, and very fast without lots of tricks. These make it a good choice for numerical calculations, which is all that I am generally interested in doing or teaching others to do. The volume of legacy code in Fortran is really just a perk for me.

Reply Score: 2

RE[3]: This is great news
by ameasures on Sun 15th Nov 2015 21:44 UTC in reply to "RE[2]: This is great news"
ameasures Member since:
2006-01-09

The volume of legacy code in Fortran is really just a perk for me.

WHOA the legacy stuff I have seen was messy and not in a good way. It seemed a hard language to engineer anything in. For building short demos - great. For building real tools - less so.

Fortran has also suffered from major redefinitions between versions ... but then most languages do ... which is another rant altogether.

Reply Score: 1

RE[4]: This is great news
by FortranMan on Mon 16th Nov 2015 02:56 UTC in reply to "RE[3]: This is great news"
FortranMan Member since:
2011-12-21

I certainly agree that the legacy stuff if ugly and often hackish. I don't like modifying old codes, but I do access standard libraries like BLAS, LAPACK, etc. The transition between versions is also jarring, but since my projects rarely get really large the effort to update as needed is not very large for me.

Reply Score: 2

RE[4]: This is great news
by acobar on Mon 16th Nov 2015 19:06 UTC in reply to "RE[3]: This is great news"
acobar Member since:
2005-11-15

I certainly agree with you. Have seen even Fortran 66 code and .. it was beyond ugly ! And Fortran 77 code was still butt ugly !

Fact is, most really useful code I had to deal with were modernized at least to Fortran 90, 95 or 2003.

Fortran was the first language I had to "hack" (well, actually, I had to create a workaround for Microsoft Fortran limitation over invocation of subroutines and functions passed as parameters, that created the need to learn x86 assembly and .. dang, created a cascade of "must learn how to do .." that never stopped and made time fly, and even now I feel like I'm always behind schedule. Well, I guess life is this way for most of us, curious beings).

Reply Score: 2