Is there a counterpart to the handy "installer" version of gfortran
http://users.humboldt.edu/finneyb/gfortran-windows-20111011.exe
but for 64-bit Windows?  The TDM version appears to need
several pieces.  The equation.com version is listed as "use at
your own risk" on the wiki.

On Nov 16, 10:58=A0am, "tho...@antispam.ham" <tho...@ifa.hawaii.edu>
wrote:
> Is there a counterpart to the handy "installer" version of gfortranhttp:/=
/users.humboldt.edu/finneyb/gfortran-windows-20111011.exe
> but for 64-bit Windows? =A0The TDM version appears to need
> several pieces. =A0The equation.com version is listed as "use at
> your own risk" on the wiki.

I have used the equation.com versions for a couple of years now and
haven't encountered any "risk".

Al Greynolds
www.ruda.com

Am 16.11.2011 19:20, schrieb Al Greynolds:
> On Nov 16, 10:58 am, "tho...@antispam.ham"<tho...@ifa.hawaii.edu>
> wrote:
>> Is there a counterpart to the handy "installer" version of gfortranhttp://users.humboldt.edu/finneyb/gfortran-windows-20111011.exe
>> but for 64-bit Windows?  The TDM version appears to need
>> several pieces.  The equation.com version is listed as "use at
>> your own risk" on the wiki.
>
> I have used the equation.com versions for a couple of years now and
> haven't encountered any "risk".

Well, at least the version I (and Polyhedron) tried gave an internal 
compiler error when compiling the Polyhedron benchmark 
(http://www.polyhedron.com/compare0html) while the MinGW64 builds at 
sourceforge.net work. As the latter are done by the MinGW64 developers, 
I would favour them. I also heard from someone having link errors with 
the equation.com builds.

However, if they work for you, there is no reason to switch.

Tobias

On 11/16/2011 2:16 PM, Tobias Burnus wrote:
> Am 16.11.2011 19:20, schrieb Al Greynolds:
>> On Nov 16, 10:58 am, "tho...@antispam.ham"<tho...@ifa.hawaii.edu>
>> wrote:
>>> Is there a counterpart to the handy "installer" version of
>>> gfortranhttp://users.humboldt.edu/finneyb/gfortran-windows-20111011.exe
>>> but for 64-bit Windows? The TDM version appears to need
>>> several pieces. The equation.com version is listed as "use at
>>> your own risk" on the wiki.
>>
>> I have used the equation.com versions for a couple of years now and
>> haven't encountered any "risk".
>
> Well, at least the version I (and Polyhedron) tried gave an internal
> compiler error when compiling the Polyhedron benchmark
> (http://www.polyhedron.com/compare0html) while the MinGW64 builds at
> sourceforge.net work. As the latter are done by the MinGW64 developers,
> I would favour them. I also heard from someone having link errors with
> the equation.com builds.
>
> However, if they work for you, there is no reason to switch.
>

The mingw64 cross compilers from the cygwin installation menu have 
worked well for me (with the exception of no gain for OpenMP).  I 
suppose they derive from sourceforge.


-- 
Tim Prince

"Tobias Burnus" <burnus@net-b.de> wrote in message 
news:4EC40BFF.1010305@net-b.de...

> Well, at least the version I (and Polyhedron) tried gave an internal 
> compiler error when compiling the Polyhedron benchmark 
> (http://www.polyhedron.com/compare0html) while the MinGW64 builds at 
> sourceforge.net work. As the latter are done by the MinGW64 developers, I 
> would favour them. I also heard from someone having link errors with the 
> equation.com builds.

There is a problem if you have more than one version of gfortran on
your computer in that the environmental variables as set via
Start->Control Panel->System->Advanced->Environmental Variables can
only be consistent with one version at a time.  Set up a Command
Prompt shortcut under the root directory of each version you have
installed which links to a *.bat file that modifies the PATH and
perhaps the EQ_LIBRARY_PATH or LIBRARY_PATH appropriately.  Use
only the shortcut appropriate to the compiler you want to run.
This way you can have several command prompts open simultaneously,
each set up for a different version of gfortran.

ISTR that one of the versions, I think it's the one from the
mingw 64 page, comes with mangled names for everything in the bin
directory.  I wrote a Fortran program that fixes these names.
Obviously you don't need to mangle names in Windows because you
can sort out different versions by having different environmental
variables.

-- 
write(*,*) transfer((/17.392111325966148d0,6.5794487871554595D-85, &
6.0134700243160014d-154/),(/'x'/)); end

On Nov 16, 1:20=A0pm, Al Greynolds <awgreyno...@earthlink.net> wrote:
> On Nov 16, 10:58=A0am, "tho...@antispam.ham" <tho...@ifa.hawaii.edu>
> wrote:
>
> > Is there a counterpart to the handy "installer" version of gfortranhttp=
://users.humboldt.edu/finneyb/gfortran-windows-20111011.exe
> > but for 64-bit Windows? =A0The TDM version appears to need
> > several pieces. =A0The equation.com version is listed as "use at
> > your own risk" on the wiki.
>
> I have used the equation.com versions for a couple of years now and
> haven't encountered any "risk".
>
> Al Greynoldswww.ruda.com

Yes, the equation.com builds have always worked well for me. For some
reason, they seem to get press from gcc/gfortran maintainers.

> Yes, the equation.com builds have always worked well for me. For some
> reason, they seem to get press from gcc/gfortran maintainers.

.... negative press ...

On 11/16/2011 3:17 PM, CaptainKirk1966 wrote:
>
>> Yes, the equation.com builds have always worked well for me. For some
>> reason, they seem to get press from gcc/gfortran maintainers.
>
> ... negative press ...
>
If it's true that they're adding on closed source libraries etc. could 
that not be "some reason?"

-- 
Tim Prince

Tim Prince <tprince@computer.org> wrote:

> On 11/16/2011 3:17 PM, CaptainKirk1966 wrote:
> >
> >> Yes, the equation.com builds have always worked well for me. For some
> >> reason, they seem to get press from gcc/gfortran maintainers.
> >
> > ... negative press ...
> >
> If it's true that they're adding on closed source libraries etc. could
> that not be "some reason?"

That's what I recall - that the equation.com releases pretty directly
violated the license conditions under which gcc/gfortran was released.
Unless one wants to spend a lot of money on lawyers, and probably
achieve very little in return for the expenditure, negative press is
about all one can do and is what I'd expect as a reaction to something
like that. I haven't paid much attention, so that might have changed, or
then it might not have.

-- 
Richard Maine                    | Good judgment comes from experience;
email: last name at domain . net | experience comes from bad judgment.
domain: summertriangle           |  -- Mark Twain

On Nov 16, 3:45=A0pm, Tim Prince <tpri...@computer.org> wrote:
> On 11/16/2011 3:17 PM, CaptainKirk1966 wrote:
>
> >> Yes, the equation.com builds have always worked well for me. For some
> >> reason, they seem to get press from gcc/gfortran maintainers.
>
> > ... negative press ...
>
> If it's true that they're adding on closed source libraries etc. could
> that not be "some reason?"
>
> --
> Tim Prince


yep, that could be some reason ...

"The major change is the library to support parallel processing has
been replaced. "

On Wed, 16 Nov 2011 12:17:14 -0800, CaptainKirk1966 wrote:

>> Yes, the equation.com builds have always worked well for me. For some
>> reason, they seem to get press from gcc/gfortran maintainers.
> 
> ... negative press ...

Perhaps, it has to do with the fact that Equation.com at one
time explicitly stated that they changed the source code.
E.C has never responded to at least two different
gfortran developers who requested access to their 
changes to the source code.

Hmmm, let's visit their website:

http://www.equation.com/servlet/equation.cmd?fa=fortran

   This web page provides Fortran, C and C++ for Windows for
   download.  Equation Solution build the compilers from GCC.
   gFortran, gcc and g++ are high performance compilers, and
   absolutely free under General Public License.  Binaries distributed
   at this site are a variant of GCC.  Distributions of Equation
   Solution have indepentent settings.  The major change is the
   library to support parallel processing has been replaced. 

Oh my, "Binaries distributed at this site are a variant of GCC"

E.C has never explained what is meant by "are a variant".

E.C has never answered my emails about the source
code they use (and I know at least one other person).

E.C appears to violate the GPL that they so proudly 
acknowledge in the above blurb.

-- 
steve

On Nov 16, 6:22=A0pm, "Steven G. Kargl"
<s...@REMOVEtroutmask.apl.washington.edu> wrote:
> On Wed, 16 Nov 2011 12:17:14 -0800, CaptainKirk1966 wrote:
> >> Yes, the equation.com builds have always worked well for me. For some
> >> reason, they seem to get press from gcc/gfortran maintainers.
>
> > ... negative press ...
>
> Perhaps, it has to do with the fact that Equation.com at one
> time explicitly stated that they changed the source code.
> E.C has never responded to at least two different
> gfortran developers who requested access to their
> changes to the source code.
>
> Hmmm, let's visit their website:
>
> http://www.equation.com/servlet/equation.cmd?fa=3Dfortran
>
> =A0 =A0This web page provides Fortran, C and C++ for Windows for
> =A0 =A0download. =A0Equation Solution build the compilers from GCC.
> =A0 =A0gFortran, gcc and g++ are high performance compilers, and
> =A0 =A0absolutely free under General Public License. =A0Binaries distribu=
ted
> =A0 =A0at this site are a variant of GCC. =A0Distributions of Equation
> =A0 =A0Solution have indepentent settings. =A0The major change is the
> =A0 =A0library to support parallel processing has been replaced.
>
> Oh my, "Binaries distributed at this site are a variant of GCC"
>
> E.C has never explained what is meant by "are a variant".
>
> E.C has never answered my emails about the source
> code they use (and I know at least one other person).
>
> E.C appears to violate the GPL that they so proudly
> acknowledge in the above blurb.
>
> --
> steve

I knew I didn't like E. C. distribution, but I didn't recall why...

Thanks, it's always good to remember some things.

Fernando.

Richard Maine wrote:
>> On 11/16/2011 3:17 PM, CaptainKirk1966 wrote:
>>>> Yes, the equation.com builds have always worked well for me. For some
>>>> reason, they seem to get press from gcc/gfortran maintainers.
>>>
>>> ... negative press ...
>>>
>> If it's true that they're adding on closed source libraries etc. could
>> that not be "some reason?"
>
> That's what I recall - that the equation.com releases pretty directly
> violated the license conditions under which gcc/gfortran was released.
> Unless one wants to spend a lot of money on lawyers, and probably
> achieve very little in return for the expenditure, negative press is
> about all one can do and is what I'd expect as a reaction to something
> like that. I haven't paid much attention, so that might have changed, or
> then it might not have.

Yes, the no-reaction to any emails and not releasing their modification 
is surely one reason. Also that a some point virus scanners warned (cf. 
[1]) - though that was seemingly a false positive. (Actually, the latter 
was the main reason that I added the warning.)

Until recently, the wording was: "builds were announced at 
comp.lang.fortran and are said to work well - but is not affiliated with 
any MinGW or GCC developer. Use at your own risk"

But I changed it to "builds were announced at comp.lang.fortran; note - 
equation.com is not affiliated with any MinGW or GCC developer. Use at 
your own risk".

The reason for the change was - as mentioned before - that the 
equation.com builds fail for Polyhedron's rnflow.f90 test case 
("internal compiler error: Segmentation fault"). As it works with the 
builds from 
http://downloads.sourceforge.net/project/mingw-w64/Toolchains%20targetting%20Win64/, 
it must be a bug in their builds. As they do not react, I cannot do much 
about it. Another failure was recently mentioned on the gfortran mailing 
list; that one was a bit odd linking problem, which seemingly went away 
when choosing a different build.

Additionally, I heard from some MinGW64 developer that equation.com has 
chosen some strange build options. As I am not using/building GCC for 
Windows, I do not know the details.


Regarding Tim's
> The mingw64 cross compilers from the cygwin installation menu have worked well for me
> (with the exception of no gain for OpenMP).  I suppose they derive from sourceforge.

A MinGW64 developer has recently rewritten the POSIX thread support 
library - before it used the one of MinGW (the 32bit project). The new 
library is still work in progress, but it is said to be faster and more 
robust. I think there are currently MinGW64 builds with the old and with 
the new library.


And regarding equation.com's "64-bit quadmath buggy": That issue has 
been fixed 2011-11-07 thus their 2011-11-11 should be OK. The reason was 
that gcc 4.7 on MinGW64 now defaults to -mms-bitfields, which broke the 
compilation of libquadmath. (See also discussion in another thread.)

Tobias

[1] http://www.equation.com/servlet/equation.cmd?fa=falsevirus

On Nov 16, 9:16=A0am, Tobias Burnus <bur...@net-b.de> wrote:

> Well, at least the version I (and Polyhedron) tried gave an internal
> compiler error when compiling the Polyhedron benchmark
> (http://www.polyhedron.com/compare0html) while the MinGW64 builds at
> sourceforge.net work. As the latter are done by the MinGW64 developers,
> I would favour them. I also heard from someone having link errors with
> the equation.com builds.

I would favor the ones done by MinGW64 developers as well, but I
haven't found a handy "installer" version that specifies 64-bit.

On 17/11/2011 10:22 a.m., Steven G. Kargl wrote:

> Oh my, "Binaries distributed at this site are a variant of GCC"
>
> E.C has never explained what is meant by "are a variant".
>
> E.C has never answered my emails about the source
> code they use (and I know at least one other person).
>
> E.C appears to violate the GPL that they so proudly
> acknowledge in the above blurb.
>

I wonder what could be their motivation.