[MPFR] Does anybody have MPFR/GMP compiled with Intel C++ on Windows?

Michael comtech.usa at gmail.com
Mon Sep 10 00:09:48 CEST 2007


Hi Brian,

Thanks a lot for your help!

I am happy to report that I am your MS version of GMP/MPFR build user.

If you could publish the Intel Version of GMP/MPFR build soon, that
would be great!

Thanks a lot and looking forward to use your new builds...

Best,

-M

On 9/9/07, Brian Gladman <brg at gladman.plus.com> wrote:
> ----- Original Message -----
> From: "Michael" <comtech.usa at gmail.com>
> To: <mpfr at loria.fr>; <gmp-discuss at swox.com>
> Sent: Sunday, September 09, 2007 5:14 PM
> Subject: [MPFR] Does anybody have MPFR/GMP compiled with Intel C++ on
> Windows?
>
>
> > Dear all,
> >
> > I am looking for the Intel C++ compilation of MPFR/GMP on Windows
> > integrated with Visual Studio 2003 or 2005. Does anybody have such a
> > compilation? I am curious about the performance comparison also.
> >
> > Thanks a lot for your help!
>
> I have a version of GMP that uses the Intel compiler running under Microsoft
> Visual Studio 2005.  But I have not yet compiled MPFR in this way (it should
> not be too hard now that GMP has been done).
>
> Interestingly with all the assembler support, it would not normally make
> much difference whether the Microsoft or the Intel compiler is used. But
> this turns out not to be true for x64 native Windows builds.
>
> Microsoft have made the 'braindead' decision not to provide x64 inline
> assembler support in their x64 compiler and this prevents taking advantage
> of a number of GMP inline assembler macros.
>
> Fortunately, however, the latest version of the Intel Windows compiler does
> have x64 inline assembler AND it can compile both Microsoft and GCC style
> inline assembler instructions.  As a result, a small change one gmp file
> allows the existing GCC style x64 inline assembler code in GMP to be used in
> a native Windows build.  And the resulting build does improve on the speed
> of the Microsoft build. In particular the division code for small operands
> is significantly faster.
>
> I have not yet published my Intel build files for GMP as this only became
> possible recently with the new version of the Intel compiler.  Also, as I
> said earlier, I have yet to converty the MPFR build.
>
>     Brian Gladman
>
>


More information about the gmp-discuss mailing list