Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > sci.physics > #532887 > unrolled thread

LLVM to get Fortran compiler that targets parallel GPUs in clusters

Back to article view | Back to sci.physics

Contents

  LLVM to get Fortran compiler that targets parallel GPUs in clusters Sam Wormley <swormley1@gmail.com> - 2015-11-13 17:28 -0600
    Re: LLVM to get Fortran compiler that targets parallel GPUs in clusters Double-A <double-a3@hush.com> - 2015-11-15 12:38 -0800
      Re: LLVM to get Fortran compiler that targets parallel GPUs in clusters jimp@specsol.spam.sux.com - 2015-11-15 20:53 +0000
        Re: LLVM to get Fortran compiler that targets parallel GPUs in clusters Double-A <double-a3@hush.com> - 2015-11-16 13:29 -0800
          Re: LLVM to get Fortran compiler that targets parallel GPUs in clusters Puppet_Sock <puppet_sock@hotmail.com> - 2015-11-16 13:39 -0800
      Re: The US Department of Energy needs to port their code to C. GreyCloud <cumulus@mist.com> - 2015-11-16 11:53 -0700

#532887 — LLVM to get Fortran compiler that targets parallel GPUs in clusters

LLVM to get Fortran compiler that targets parallel GPUs in clusters
> http://arstechnica.com/information-technology/2015/11/llvm-to-get-fortran-compiler-that-targets-parallel-gpus-in-clusters/


> 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.
>
> 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.
>
> They'll be working closely with a division of Nvidia that makes the
> PGI Accelerator toolkit. That software is designed to allow clusters
> to use the computation abilities of GPUs on the cluster to perform
> distributed computations. PGI Accelerator's syntax is intended to be
> similar to that of OpenMP, a parallel computing API that's also
> heavily used in scientific computing.
>
> According to National Labs' announcement, LLVM was already attracting
> interest from the high performance computing community, but they were
> mostly using its CLANG C/C++ compiler. But many of the large software
> packages used in scientific computing have a mix of C++ and Fortran
> modules, providing the incentive to bring it all under one toolchain.
> Los Alamos' Patrick McCormick said “The goal of this effort is to
> fill a key gap in the LLVM infrastructure that will help it become a
> more complete solution for our overall capabilities in support of the
> advancement of science and engineering.”



-- 

sci.physics is an unmoderated newsgroup dedicated
to the discussion of physics, news from the physics
community, and physics-related social issues.

[toc] | [next] | [standalone]

#533342

On Friday, November 13, 2015 at 3:28:48 PM UTC-8, Sam Wormley wrote:
> LLVM to get Fortran compiler that targets parallel GPUs in clusters
> > http://arstechnica.com/information-technology/2015/11/llvm-to-get-fortran-compiler-that-targets-parallel-gpus-in-clusters/
> 
> 
> > 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.
> >
> > 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.
> >
> > They'll be working closely with a division of Nvidia that makes the
> > PGI Accelerator toolkit. That software is designed to allow clusters
> > to use the computation abilities of GPUs on the cluster to perform
> > distributed computations. PGI Accelerator's syntax is intended to be
> > similar to that of OpenMP, a parallel computing API that's also
> > heavily used in scientific computing.
> >
> > According to National Labs' announcement, LLVM was already attracting
> > interest from the high performance computing community, but they were
> > mostly using its CLANG C/C++ compiler. But many of the large software
> > packages used in scientific computing have a mix of C++ and Fortran
> > modules, providing the incentive to bring it all under one toolchain.
> > Los Alamos' Patrick McCormick said "The goal of this effort is to
> > fill a key gap in the LLVM infrastructure that will help it become a
> > more complete solution for our overall capabilities in support of the
> > advancement of science and engineering."


When I was studying Fortran at the university about 45 years ago, I was told my a math professor that Fortran be obsolete by the time we graduated.   My.  Some things just never seem to fade away!

Double-A

[toc] | [prev] | [next] | [standalone]

#533354

Double-A <double-a3@hush.com> wrote:
> On Friday, November 13, 2015 at 3:28:48 PM UTC-8, Sam Wormley wrote:
>> LLVM to get Fortran compiler that targets parallel GPUs in clusters
>> > http://arstechnica.com/information-technology/2015/11/llvm-to-get-fortran-compiler-that-targets-parallel-gpus-in-clusters/
>> 
>> 
>> > 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.
>> >
>> > 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.
>> >
>> > They'll be working closely with a division of Nvidia that makes the
>> > PGI Accelerator toolkit. That software is designed to allow clusters
>> > to use the computation abilities of GPUs on the cluster to perform
>> > distributed computations. PGI Accelerator's syntax is intended to be
>> > similar to that of OpenMP, a parallel computing API that's also
>> > heavily used in scientific computing.
>> >
>> > According to National Labs' announcement, LLVM was already attracting
>> > interest from the high performance computing community, but they were
>> > mostly using its CLANG C/C++ compiler. But many of the large software
>> > packages used in scientific computing have a mix of C++ and Fortran
>> > modules, providing the incentive to bring it all under one toolchain.
>> > Los Alamos' Patrick McCormick said "The goal of this effort is to
>> > fill a key gap in the LLVM infrastructure that will help it become a
>> > more complete solution for our overall capabilities in support of the
>> > advancement of science and engineering."
> 
> 
> When I was studying Fortran at the university about 45 years ago, I was told my a math professor that Fortran be obsolete by the time we graduated.   My.  Some things just never seem to fade away!
> 
> Double-A
 
Fortran has been obsolete for new applications for decades.

The problem is the huge piles of legacy math and science code no one
can afford to port to a more modern language.

Programming hell is to be tasked with maintaining legacy Fortran or 
Cobal code.


-- 
Jim Pennino

[toc] | [prev] | [next] | [standalone]

#533618

On Sunday, November 15, 2015 at 1:01:07 PM UTC-8, ji...@specsol.spam.sux.com wrote:
> Double-A <double-a3@hush.com> wrote:
> > On Friday, November 13, 2015 at 3:28:48 PM UTC-8, Sam Wormley wrote:
> >> LLVM to get Fortran compiler that targets parallel GPUs in clusters
> >> > http://arstechnica.com/information-technology/2015/11/llvm-to-get-fortran-compiler-that-targets-parallel-gpus-in-clusters/
> >> 
> >> 
> >> > 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.
> >> >
> >> > 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.
> >> >
> >> > They'll be working closely with a division of Nvidia that makes the
> >> > PGI Accelerator toolkit. That software is designed to allow clusters
> >> > to use the computation abilities of GPUs on the cluster to perform
> >> > distributed computations. PGI Accelerator's syntax is intended to be
> >> > similar to that of OpenMP, a parallel computing API that's also
> >> > heavily used in scientific computing.
> >> >
> >> > According to National Labs' announcement, LLVM was already attracting
> >> > interest from the high performance computing community, but they were
> >> > mostly using its CLANG C/C++ compiler. But many of the large software
> >> > packages used in scientific computing have a mix of C++ and Fortran
> >> > modules, providing the incentive to bring it all under one toolchain.
> >> > Los Alamos' Patrick McCormick said "The goal of this effort is to
> >> > fill a key gap in the LLVM infrastructure that will help it become a
> >> > more complete solution for our overall capabilities in support of the
> >> > advancement of science and engineering."
> > 
> > 
> > When I was studying Fortran at the university about 45 years ago, I was told my a math professor that Fortran be obsolete by the time we graduated.   My.  Some things just never seem to fade away!
> > 
> > Double-A
>  
> Fortran has been obsolete for new applications for decades.
> 
> The problem is the huge piles of legacy math and science code no one
> can afford to port to a more modern language.
> 
> Programming hell is to be tasked with maintaining legacy Fortran or 
> Cobal code.
> 
> 
> -- 
> Jim Pennino


My last project was to translate a huge assembler program into COBOL.  I would never have advised it myself.  Even applying my knowledge of COBOL internals to maximize efficiency, the COBOL version ran much slower than the assembler version.  The usual rationale is that COBOL programmers are easier to find (and cheaper) to maintain the code.  My rational always was that I would do whatever they wanted so long as I got paid.  It put the bread on the table, but sometimes made me feel like a whore.

Double-A

[toc] | [prev] | [next] | [standalone]

#533626

On Monday, November 16, 2015 at 4:29:35 PM UTC-5, Double-A wrote:
[snips]
> My last project was to translate a huge assembler program into COBOL. 
[snips]

Glurg! COBOL? So the stone knives and bear skins were turned into bronze knives and woven read matts.

[toc] | [prev] | [next] | [standalone]

#533548 — Re: The US Department of Energy needs to port their code to C.

On 11/15/15 15:02, Jeff-Relf.Me wrote:
>
> Re: LLVM to get Fortran compiler that targets parallel GPUs in clusters.
>    http://arstechnica.com/information-technology/2015/11/llvm-to-get-fortran-compiler-that-targets-parallel-gpus-in-clusters/
>
> The US Department of Energy needs to port their code to C.

What for?  Fortran was designed to signal a numerical inaccuracy inside 
loops.  C does not.  You'd have a lot of coding to do to detect 
numerical drift from what it should be.

All they have to do with the fortran code is to recompile it, that is of 
course if LLVM can do the parallelisms for you without any extra coding.

>
> To take advantage of "parallel GPUs", you need "Threads";
> MicroSoft C, Apple's CLANG, and GNU gcc, have had them for decades.
> US National Labs is "mostly using [their] CLANG C/C++ compiler".
>
> To run "Legacy Code", written in Fortran, you'd have to modify it;
> and, right now, it can't be done.

Yes, it can be done.  But then that is LLVMs challenge.

>
> Most likely, it will _never_ be done;
> the US Department of Energy is wasting money.

They want us in our homes cold and dark.


-- 
My problem is that I don't have enough middle fingers.

[toc] | [prev] | [standalone]

Back to top | Article view | sci.physics

csiph-web