Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
Wassup with Cray Fortran compiler!?
740 views
Skip to first unread message
FortranFan
Aug 16, 2014, 11:45:24 AM8/16/14
to
Does anyone know of a way to gain "recreational" access to the latest Cray Fortran compiler for some short period of time?
I'm very interested in many features in the latest Fortran standard (submodules, coarrays, blocks with exit, recursive components of allocatable type, etc.) as well as those in TS 29113 (for enhanced interoperability with C) that I can't get to or use reliably in the compiler versions I have. I find it very tiring and highly frustrating to sit around and wait until the compilers I use have fully functional and bug-free implementations of all these features; it's really a "mixed bag" out there right now, as everyone knows. My interest is mainly in learning about these new features by actually trying them out in some test code and gauging what is truly possible with the latest Fortran standard and how easy it will be do the things I want to do.
As posted by Ian Chivers in this thread (https://groups.google.com/forum/#!topic/comp.lang.fortran/mqJBRAJyX7I) and reported in their article at the Fortran Forum, Cray compiler appears fully conformant with Fortran 2008 as well as most of TS 29113 (how does Cray do it when other vendors struggle so much!).
Hence I wonder if it is possible to get "trial access" to Cray Fortran compiler without an account at a top-secret/"ivory tower" supercomputer facility?
Thanks,
I'm very interested in many features in the latest Fortran standard (submodules, coarrays, blocks with exit, recursive components of allocatable type, etc.) as well as those in TS 29113 (for enhanced interoperability with C) that I can't get to or use reliably in the compiler versions I have. I find it very tiring and highly frustrating to sit around and wait until the compilers I use have fully functional and bug-free implementations of all these features; it's really a "mixed bag" out there right now, as everyone knows. My interest is mainly in learning about these new features by actually trying them out in some test code and gauging what is truly possible with the latest Fortran standard and how easy it will be do the things I want to do.
As posted by Ian Chivers in this thread (https://groups.google.com/forum/#!topic/comp.lang.fortran/mqJBRAJyX7I) and reported in their article at the Fortran Forum, Cray compiler appears fully conformant with Fortran 2008 as well as most of TS 29113 (how does Cray do it when other vendors struggle so much!).
Hence I wonder if it is possible to get "trial access" to Cray Fortran compiler without an account at a top-secret/"ivory tower" supercomputer facility?
Thanks,
Paul Anton Letnes
Aug 19, 2014, 3:54:13 AM8/19/14
to
get a Cray compiler. I've had the pleasure of working on one a couple of
years ago, it was very well run and so on, but I don't think you can get
the compiler for "recreational" use. That's a shame, though. They
probably have Fortran high on their list of priorities, since otherwise,
noone would buy their systems.
I do know that Intel gives away non-commercial licenses for Linux that
you can install on a machine for private, non-commercial,
teach-yourself-ifort use. If you don't like running Linux, you can
always install it in a VirtualBox virtual machine for free. Ifort comes
with coarrays, and I think some of the other features you describe.
Paul
Steve Lionel
Aug 19, 2014, 11:09:26 AM8/19/14
to
On 8/19/2014 3:54 AM, Paul Anton Letnes wrote:
> I do know that Intel gives away non-commercial licenses for Linux that
> you can install on a machine for private, non-commercial,
> teach-yourself-ifort use. If you don't like running Linux, you can
> always install it in a VirtualBox virtual machine for free. Ifort comes
> with coarrays, and I think some of the other features you describe.
FortranFan already has the Intel compiler. As he wrote earlier, he wants
> I do know that Intel gives away non-commercial licenses for Linux that
> you can install on a machine for private, non-commercial,
> teach-yourself-ifort use. If you don't like running Linux, you can
> always install it in a VirtualBox virtual machine for free. Ifort comes
> with coarrays, and I think some of the other features you describe.
to use features ifort doesn't yet support.
I did want to take a moment to let folks know that the option to get a
non-commercial Fortran Linux license from us is going away very soon, so
if you are thinking of getting that you should do it now.
https://registrationcenter.intel.com/RegCenter/NComForm.aspx?ProductID=1523&pass=yes
--
Steve Lionel
Developer Products Division
Intel Corporation
Merrimack, NH
For email address, replace "invalid" with "com"
User communities for Intel Software Development Products
http://software.intel.com/en-us/forums/
Intel Software Development Products Support
http://software.intel.com/sites/support/
My Fortran blog
http://www.intel.com/software/drfortran
Refer to http://software.intel.com/en-us/articles/optimization-notice
for more information regarding performance and optimization choices in
Intel software products.
FortranFan
Aug 19, 2014, 2:58:23 PM8/19/14
to
I was looking at the websites of ORNL and NESRC government-run supercomputer site which have the latest Cray compiling environment and again, it is a shame they can't even create a "sandbox" (apart from their "top secret" research) where ordinary citizens can make use of the supercomputer facilities. Hell, I can't open an account even if I'm willing to pay a reasonable fee to use the facilites - WTF, so much for all the taxes I pay!? I know, I know all the standard excuses about funding, security, blah, blah,..
Anyways if Cray is always going to be "first off the block" and if anyone from their Fortran compiler team is listening, they might want to think about making something available that Fortran enthusiasts who can "play around with" the latest updates (particularly the standards related enhancements) and even get a lot of testing done that way (a la Intel and their IFORT compiler). It'll be doing the Fortran community a great service, if they so cared.
Another thought: perhaps the software teams, at least the Fortran compiler teams, at Cray and Intel can partner up in terms of feature implementations; both Cray and Intel advertise a supposedly significant partnership in hardware anyway.
As mentioned by Steve Lionel from Intel in a subsequent post, I do have access to their latest Fortran compiler. Per my work process, IFORT is the compiler for all "production" code on Windows and Linux; just wrapping up the paperwork for a PO to acquire a few additional IFORT licenses. In a project I'm currently working on, some of Intel's implementations of TS 29113, esp. removal of restrictions wrt TYPE(C_PTR) and inoperable types (per BIND(C)) is a godsend - it allows me to make certain Fortran "classes" available to a much broader (i.e., non-Fortranner) audience in ways they want to and are accustomed to consuming.
So I want to see what else I can do with Fortran 2008 and TS 29113 and also what Fortran 201X will have in store. In the world I live in, even coders in science and engineering (mostly C++/Blitz++ but also C#, MATLAB, Python, etc.) get access to newer language features and fancy tools extensions at a pace they can't keep up with. Where as I would want to "evangelize" Fortran but is usually just waiting and feeling I'm getting too ahead of myself with my management who can't think any less of Fortran.
Stansfield Temmelmeier
Aug 21, 2014, 5:29:02 AM8/21/14
to
On 2014-08-19, Steve Lionel <steve....@intel.invalid> wrote:
> I did want to take a moment to let folks know that the option to get a
> non-commercial Fortran Linux license from us is going away very soon, so
> if you are thinking of getting that you should do it now.
> https://registrationcenter.intel.com/RegCenter/NComForm.aspx?ProductID=1523&pass=yes
Disastrous news. Does it help to register now and you become grandfather for
> I did want to take a moment to let folks know that the option to get a
> non-commercial Fortran Linux license from us is going away very soon, so
> if you are thinking of getting that you should do it now.
> https://registrationcenter.intel.com/RegCenter/NComForm.aspx?ProductID=1523&pass=yes
future releases or we can only get a license for the last version soon and
after it there will be no new version of non commercial users?
Thanks,
Stan
FX
Aug 21, 2014, 2:38:57 PM8/21/14
to
>> I did want to take a moment to let folks know that the option to get a
>> non-commercial Fortran Linux license from us is going away very soon
>
> Disastrous news.
>> non-commercial Fortran Linux license from us is going away very soon
>
The non-commercial license (whose terms are very restrictive) is widely
abused, at least in the academic world. I'm sad there will be one fewer
compiler available for Fortran enthusiasts, but the existing situation
was weird.
--
FX
Steve Lionel
Aug 22, 2014, 11:37:27 AM8/22/14
to
On 8/21/2014 5:29 AM, Stansfield Temmelmeier wrote:
> Disastrous news. Does it help to register now and you become grandfather for
> future releases or we can only get a license for the last version soon and
> after it there will be no new version of non commercial users?
The latter. FX hits on the main reason for this change, and it's also
> Disastrous news. Does it help to register now and you become grandfather for
> future releases or we can only get a license for the last version soon and
> after it there will be no new version of non commercial users?
commercial users we see abusing the non-commercial license, but I've
been trying to convince the powers-that-be that there is a middle
ground. This is also behind my earlier request for pointers to
open-source Fortran software, as there is a misperception that there is
no open-source development in Fortran.
Stansfield Temmelmeier
Aug 24, 2014, 3:50:19 AM8/24/14
to
This is very misfortunate. The criminals always ruin everything for normal
people. I don't think you must find open source projects in Fortran proving
there is some legitimate private user. People can use Fortran in their own
project and it is not open source and it will not be found on the net. I
have many small projects like this.
I hope they will consider this issue again and sue companies that misuse it
instead of making the people who don't make money from it pay for it. It's
very evil upside down world and I hate criminals......
Stan
people. I don't think you must find open source projects in Fortran proving
there is some legitimate private user. People can use Fortran in their own
project and it is not open source and it will not be found on the net. I
have many small projects like this.
I hope they will consider this issue again and sue companies that misuse it
instead of making the people who don't make money from it pay for it. It's
very evil upside down world and I hate criminals......
Stan
Phillip Helbig---undress to reply
Aug 24, 2014, 7:29:26 AM8/24/14
to
In article <ltc5fr$spm$1...@speranza.aioe.org>, Stansfield Temmelmeier
license at a lower cost than a free one. Even if the cost is so low
that it doesn't matter, it does create a different impression.
If someone is abusing the non-commercial license, then the fact that
this is known means that it is known who it is, so why not go after them
rather than kill it for the hobbyists?
<s...@optika.de> writes:
> This is very misfortunate. The criminals always ruin everything for normal
> people. I don't think you must find open source projects in Fortran proving
> there is some legitimate private user. People can use Fortran in their own
> project and it is not open source and it will not be found on the net. I
> have many small projects like this.
It seems to me that it should be possible to have a non-commercial
> This is very misfortunate. The criminals always ruin everything for normal
> people. I don't think you must find open source projects in Fortran proving
> there is some legitimate private user. People can use Fortran in their own
> project and it is not open source and it will not be found on the net. I
> have many small projects like this.
license at a lower cost than a free one. Even if the cost is so low
that it doesn't matter, it does create a different impression.
If someone is abusing the non-commercial license, then the fact that
this is known means that it is known who it is, so why not go after them
rather than kill it for the hobbyists?
Bob Radford
Aug 24, 2014, 3:34:40 PM8/24/14
to
Here's an interesting find
http://physics.princeton.edu/pulsar/K1JT/devel.html
"Source Code
WSJT, MAP65, WSPR, WSPR-X, and WSJT-X are open source programs distributed
under terms of the GNU General Public License. Source code for SimJT is
available on request. KVASD (a soft-decision Reed Solomon decoder for the
JT65 mode) is distributed only in executable form because of licensing
restrictions.
The graphical user interfaces (GUIs) for WSJT and WSPR are written in
Python. Most of the number-crunching in these programs is done in Fortran,
and some system-level functions are implemented in C.
Version 2 of MAP65, WSPR-X, and WSJT-X represent the most recent shift in
WSJT-related programming practices. The GUIs for these programs are written
in C++ using the open-source Qt framework. As with WSJT and WSPR, most of
the number-crunching is done in Fortran."
Have not downloaded it but came across the page while searching for Fortran
stuff online.
bob
On 2014-08-22, Steve Lionel <steve....@intel.invalid> wrote:
http://physics.princeton.edu/pulsar/K1JT/devel.html
"Source Code
WSJT, MAP65, WSPR, WSPR-X, and WSJT-X are open source programs distributed
under terms of the GNU General Public License. Source code for SimJT is
available on request. KVASD (a soft-decision Reed Solomon decoder for the
JT65 mode) is distributed only in executable form because of licensing
restrictions.
The graphical user interfaces (GUIs) for WSJT and WSPR are written in
Python. Most of the number-crunching in these programs is done in Fortran,
and some system-level functions are implemented in C.
Version 2 of MAP65, WSPR-X, and WSJT-X represent the most recent shift in
WSJT-related programming practices. The GUIs for these programs are written
in C++ using the open-source Qt framework. As with WSJT and WSPR, most of
the number-crunching is done in Fortran."
Have not downloaded it but came across the page while searching for Fortran
stuff online.
bob
On 2014-08-22, Steve Lionel <steve....@intel.invalid> wrote:
William Clodius
Aug 24, 2014, 11:20:35 PM8/24/14
to
Phillip Helbig---undress to reply <hel...@astro.multiCLOTHESvax.de>
wrote:
It must be known that such abuses exist, which would normally require
knowing some of the abusers. It does not implying knowing all the
abusers. Further America's legal system is very expensive, and if the
known abusers are small fry the costs of suing will rarely be worth the
resutls.
wrote:
knowing some of the abusers. It does not implying knowing all the
abusers. Further America's legal system is very expensive, and if the
known abusers are small fry the costs of suing will rarely be worth the
resutls.
glen herrmannsfeldt
Aug 25, 2014, 12:18:46 AM8/25/14
to
William Clodius <wclo...@earthlink.net> wrote:
> Phillip Helbig---undress to reply <hel...@astro.multiCLOTHESvax.de>
> wrote:
(snip)
> Phillip Helbig---undress to reply <hel...@astro.multiCLOTHESvax.de>
> wrote:
>> It seems to me that it should be possible to have a non-commercial
>> license at a lower cost than a free one. Even if the cost is so low
>> that it doesn't matter, it does create a different impression.
the non-commercial version. Personally, for posts to CLF, I sometimes
compile (usually with gfortran) small programs, but not large ones.
There is an interesting feature that Dignus has for their C compiler
where you can past a program into a web page, and it will compile
it and show the generated assembly code. Convenient for testing,
not for commercial use.
>> If someone is abusing the non-commercial license, then the fact that
>> this is known means that it is known who it is, so why not go after
>> them rather than kill it for the hobbyists?
> It must be known that such abuses exist, which would normally require
> knowing some of the abusers. It does not implying knowing all the
> abusers. Further America's legal system is very expensive, and if the
> known abusers are small fry the costs of suing will rarely be worth the
> resutls.
software costs more than one might collect suing.
I would hope those selling a profitable program would find the compiler
reasonably priced and pay for it. Those without a product, but still
not satisfying non-commercial use, are more of a problem.
There are also some programs (as far as I know, not this one) free
for academic use.
-- glen
FX
Aug 25, 2014, 4:47:42 AM8/25/14
to
> I would hope those selling a profitable program would find the compiler
> reasonably priced and pay for it. Those without a product, but still
> not satisfying non-commercial use, are more of a problem.
>
> There are also some programs (as far as I know, not this one) free for
> academic use.
There are still two free (gratis) modern Fortran compilers out there:
> reasonably priced and pay for it. Those without a product, but still
> not satisfying non-commercial use, are more of a problem.
>
> There are also some programs (as far as I know, not this one) free for
> academic use.
- gfortran on Windows, Mac, Linux and many more OS'es (open source)
- Oracle Studio on Linux (and Solaris OS)
Some have unclear status:
- g95 (unmaintained)
- PathScale (used to be open source, apparently not any more)
- Open64 (status unclear to me)
--
FX
Tim Prince
Aug 25, 2014, 7:04:52 AM8/25/14
to
On 8/25/2014 4:47 AM, FX wrote:
> There are still two free (gratis) modern Fortran compilers out there:
>
> - gfortran on Windows, Mac, Linux and many more OS'es (open source)
> - Oracle Studio on Linux (and Solaris OS)
>
> Some have unclear status:
>
> - g95 (unmaintained)
> - PathScale (used to be open source, apparently not any more)
> - Open64 (status unclear to me)
>
Was PathScale a fork of Open64? That would appear to incur obligation
> There are still two free (gratis) modern Fortran compilers out there:
>
> - gfortran on Windows, Mac, Linux and many more OS'es (open source)
> - Oracle Studio on Linux (and Solaris OS)
>
> Some have unclear status:
>
> - g95 (unmaintained)
> - PathScale (used to be open source, apparently not any more)
> - Open64 (status unclear to me)
>
to continued open updates.
developer.amd.com posted nearly annual updates of their open64 up to 17
months ago.
http://en.wikipedia.org/wiki/Open64 discusses other open64 forks,
including past "official" ones.
FX
Aug 25, 2014, 8:58:09 AM8/25/14
to
> Was PathScale a fork of Open64?
No, but both are derivatives of the MIPSPro compiler. Regarding PathScale
compiler, see my post "PathScale EKOPath now closed-source again?"
(https://groups.google.com/forum/#!topic/comp.lang.fortran/In3E4bQGaxc)
back in July.
--
FX
Izaak B Beekman
Aug 25, 2014, 11:23:38 AM8/25/14
to
Cray has some of its own machines that you may be able to get time on.
I think they are usually used to let potential customers do some
benchmarking before they buy some $1.e7 hardware.
--
-Zaak
I think they are usually used to let potential customers do some
benchmarking before they buy some $1.e7 hardware.
--
-Zaak
Damian Rouson
Aug 26, 2014, 10:54:20 PM8/26/14
to
On Saturday, August 16, 2014 11:45:24 AM UTC-4, FortranFan wrote:
> I'm very interested in many features in the latest Fortran standard (submodules, coarrays, blocks with exit, recursive components of allocatable type, etc.) as well as those in TS 29113 (for enhanced interoperability with C) that I can't get to or use reliably in the compiler versions I have.
For now, I think your best choice is gfortran. At least the gcc trunk (5.0) supports coarrays and most of TS 29113.
> I'm very interested in many features in the latest Fortran standard (submodules, coarrays, blocks with exit, recursive components of allocatable type, etc.) as well as those in TS 29113 (for enhanced interoperability with C) that I can't get to or use reliably in the compiler versions I have.
Damian
Paul Anton Letnes
Aug 28, 2014, 4:33:53 PM8/28/14
to
On Friday, August 22, 2014 5:37:27 PM UTC+2, Steve Lionel wrote:
> On 8/21/2014 5:29 AM, Stansfield Temmelmeier wrote:
>
> > Disastrous news. Does it help to register now and you become grandfather for
>
> > future releases or we can only get a license for the last version soon and
>
> > after it there will be no new version of non commercial users?
>
>
>
> The latter. FX hits on the main reason for this change, and it's also
>
> commercial users we see abusing the non-commercial license, but I've
>
> been trying to convince the powers-that-be that there is a middle
>
> ground. This is also behind my earlier request for pointers to
>
> open-source Fortran software, as there is a misperception that there is
>
> no open-source development in Fortran.
>
>
>
> --
>
> Steve Lionel
>
> Developer Products Division
>
> Intel Corporation
>
> Merrimack, NH
>
>
Steve,
> On 8/21/2014 5:29 AM, Stansfield Temmelmeier wrote:
>
> > Disastrous news. Does it help to register now and you become grandfather for
>
> > future releases or we can only get a license for the last version soon and
>
> > after it there will be no new version of non commercial users?
>
>
>
> The latter. FX hits on the main reason for this change, and it's also
>
> commercial users we see abusing the non-commercial license, but I've
>
> been trying to convince the powers-that-be that there is a middle
>
> ground. This is also behind my earlier request for pointers to
>
> open-source Fortran software, as there is a misperception that there is
>
> no open-source development in Fortran.
>
>
>
> --
>
> Steve Lionel
>
> Developer Products Division
>
> Intel Corporation
>
> Merrimack, NH
>
>
not sure if one data point is worth anything, but I've kept an ifort for Linux license at home to keep up with features etc. I've found it to be a very good compiler in several respects, and having a demo available can be useful when persuading an employer to fork over $ for a compiler. Installing trial editions can be a hassle, and is generally much less worth it, imho.
Paul
Johannes
Sep 16, 2014, 8:06:45 AM9/16/14
to
>
>
>
> Steve,
>
>
>
> not sure if one data point is worth anything, but I've kept an ifort for Linux license at home to keep up with features etc. I've found it to be a very good compiler in several respects, and having a demo available can be useful when persuading an employer to fork over $ for a compiler. Installing trial editions can be a hassle, and is generally much less worth it, imho.
>
>
>
> Paul
Hi all,
>
>
> Steve,
>
>
>
> not sure if one data point is worth anything, but I've kept an ifort for Linux license at home to keep up with features etc. I've found it to be a very good compiler in several respects, and having a demo available can be useful when persuading an employer to fork over $ for a compiler. Installing trial editions can be a hassle, and is generally much less worth it, imho.
>
>
>
> Paul
it happened meanwhile. I learned with sorrow that the non-commercial Intel compilers are not available anymore for new users:
https://software.intel.com/en-us/forums/topic/330003
I wonder, according to Steve Lionel, why the C/C++ compilers will be available in undefined time in future and why the Fortran companion will not?
Does the Fortran community contributes so little. Are there so few open source projects...
Best regards,
Johannes
rusi_pathan
Sep 26, 2014, 9:13:40 PM9/26/14
to
When I was using Crays (via TeraGrid) the default compilers on them were from PGI. AFAIK most people were either using PGI or GNU compilers. The number of folks who have access to a cray machine and who actually use a Cray Fortran compiler is probably even lower (and significantly lower if you include folks that use F2003 and/or Coarrays with CAF).
While Cray has a long history with Coarrays and does claim full support for Fortran 2003/2008 I don't think the compiler has been tested that much, except for Crays own internal testing. Though I must say that their compiler worked perfectly well with my own (Fortran 95) code(s).
Just my $0.02.
While Cray has a long history with Coarrays and does claim full support for Fortran 2003/2008 I don't think the compiler has been tested that much, except for Crays own internal testing. Though I must say that their compiler worked perfectly well with my own (Fortran 95) code(s).
Just my $0.02.
Anton Shterenlikht
Sep 27, 2014, 4:50:15 AM9/27/14
to
rusi_pathan <tabr...@gmail.com> writes:
>When I was using Crays (via TeraGrid) the default compilers on them were fr=
>om PGI. AFAIK most people were either using PGI or GNU compilers. The numbe=
>r of folks who have access to a cray machine and who actually use a Cray Fo=
>rtran compiler is probably even lower (and significantly lower if you inclu=
> Fortran 2003/2008 I don't think the compiler has been tested that much, ex=
>cept for Crays own internal testing. Though I must say that their compiler =
Apart from loosing out on highly optimised libraries, you'll
be loosing on the best coarray and submodule support there
is ATM. In addition, Cray compiler is the only one I'm aware
of that not only allows but supports coarray+MPI+OpenMP
programs. I think you also get much better profiling
and debugging support (Cray ATP, CrayPAT) if you use the
Cray compiler.
I bet the Cray compiler has been tested by the
user coarray code more than any other.
For example, search for European Centre for Medium-Range
Weather Forecasts (ECMWF) codes and how they
benefit from Cray compilers.
>On Saturday, August 16, 2014 10:45:24 AM UTC-5, FortranFan wrote:
>> Does anyone know of a way to gain "recreational" access to the latest Cra=
>> As posted by Ian Chivers in this thread (https://groups.google.com/forum/=
>#!topic/comp.lang.fortran/mqJBRAJyX7I) and reported in their article at the=
> Fortran Forum, Cray compiler appears fully conformant with Fortran 2008 as=
providing much of TS 18508 (further coarray features)
for some time.
> (how does Cray do it when other vendors struggle =
>so much!).
good question. My view is that they try to lead.
It seems to me that TS 18508 is based largely on
Cray extensions, experience and user feedback.
In principle this is a good model - you provide
extensions, and see how popular they are. This
allows you to put new features in the standard
that you have already tried and tested.
** I might be completely wrong here, pure speculation on my part **.
Anton
>When I was using Crays (via TeraGrid) the default compilers on them were fr=
>om PGI. AFAIK most people were either using PGI or GNU compilers. The numbe=
>r of folks who have access to a cray machine and who actually use a Cray Fo=
>rtran compiler is probably even lower (and significantly lower if you inclu=
>de folks that use F2003 and/or Coarrays with CAF).
>While Cray has a long history with Coarrays and does claim full support for=
> Fortran 2003/2008 I don't think the compiler has been tested that much, ex=
>cept for Crays own internal testing. Though I must say that their compiler =
>worked perfectly well with my own (Fortran 95) code(s).
>Just my $0.02.
You'd be crazy not to use Cray compiler on a Cray system.
>Just my $0.02.
Apart from loosing out on highly optimised libraries, you'll
be loosing on the best coarray and submodule support there
is ATM. In addition, Cray compiler is the only one I'm aware
of that not only allows but supports coarray+MPI+OpenMP
programs. I think you also get much better profiling
and debugging support (Cray ATP, CrayPAT) if you use the
Cray compiler.
I bet the Cray compiler has been tested by the
user coarray code more than any other.
For example, search for European Centre for Medium-Range
Weather Forecasts (ECMWF) codes and how they
benefit from Cray compilers.
>On Saturday, August 16, 2014 10:45:24 AM UTC-5, FortranFan wrote:
>y Fortran compiler for some short period of time?
don't know
>> As posted by Ian Chivers in this thread (https://groups.google.com/forum/=
>#!topic/comp.lang.fortran/mqJBRAJyX7I) and reported in their article at the=
> Fortran Forum, Cray compiler appears fully conformant with Fortran 2008 as=
> well as most of TS 29113
I don't know about 29113, but Cray have been
providing much of TS 18508 (further coarray features)
for some time.
> (how does Cray do it when other vendors struggle =
>so much!).
good question. My view is that they try to lead.
It seems to me that TS 18508 is based largely on
Cray extensions, experience and user feedback.
In principle this is a good model - you provide
extensions, and see how popular they are. This
allows you to put new features in the standard
that you have already tried and tested.
** I might be completely wrong here, pure speculation on my part **.
Anton
rusi_pathan
Sep 27, 2014, 9:26:02 AM9/27/14
to
On Saturday, September 27, 2014 3:50:15 AM UTC-5, Anton Shterenlikht wrote:
> You'd be crazy not to use Cray compiler on a Cray system.
>
> Apart from loosing out on highly optimised libraries, you'll
>
> be loosing on the best coarray and submodule support there
>
> is ATM. In addition, Cray compiler is the only one I'm aware
>
> of that not only allows but supports coarray+MPI+OpenMP
>
> programs. I think you also get much better profiling
>
> and debugging support (Cray ATP, CrayPAT) if you use the
>
> Cray compiler.
As I said the default compilers on Cray machines often were those from PGI and not Cray. It is still true for Titan at OLCF which is the largest/fastest Cray machine. See the user guide at
> You'd be crazy not to use Cray compiler on a Cray system.
>
> Apart from loosing out on highly optimised libraries, you'll
>
> be loosing on the best coarray and submodule support there
>
> is ATM. In addition, Cray compiler is the only one I'm aware
>
> of that not only allows but supports coarray+MPI+OpenMP
>
> programs. I think you also get much better profiling
>
> and debugging support (Cray ATP, CrayPAT) if you use the
>
> Cray compiler.
https://www.olcf.ornl.gov/support/system-user-guides/titan-user-guide/#267
Perhaps Cray's own engineers and people running OCLF are indeed crazy.
rusi_pathan
Sep 27, 2014, 9:47:14 AM9/27/14
to
The paper is from Crays users group meeting. Use of Crays own compilers on their own machines is orders of magnitude behind PGI, GNU and Intel. I am not second guessing their compiler quality (as I have used it to compile my own Fortran 95 codes), just stating facts.
0 new messages