Re: [Boost-build] fortran fixes and usage fixes for intel-linux.jam

Subject: Re: [Boost-build] fortran fixes and usage fixes for intel-linux.jam
From: Spencer E. Olson (olsonse_at_[hidden])
Date: 2010-01-19 12:13:50

Did anyone get a chance to review the patches that I sent earlier? 
Is there a prospect for getting these patches included?

I am using Boost.Build in a large high performance computing effort and would 
really benefit from having these changes included--the CMake evangelists that 
I work with are becoming quite loud and I'd like to avoid that path if 
possible.  

Thanks,

Spencer Olson

On Wednesday 06 January 2010 16:10, Spencer E. Olson wrote:
> Please find two patches attached (to be applied in the specified order)
> that fix some issues with Fortran and improve the usage for intel-linux.jam
>
> Fortran fix:
> 1.  Adds FORTRAN90 to one of the Fortran types for gfortran.jam
> 2.  Makes Fortran compiler a part of the particular toolset rather than a
> separate tool.  I realize that some people use one Fortran compiler with a
> different suite, but I think it will be more common to use Fortran
> compilers in correlation to the toolset= option.  This makes gfortran a
> part of the gcc toolset.  It also makes ifort be used as a part of the
> Intel suite (at least on intel-linux--I don't have Intel compilers on
> windows).
>
>     By keeping the Fortran compiler a part of the suite, the Boost.Build
> user does not need to do any complex module loading just to compile Fortran
> code. Furthermore, the Boost.Build user does not need to do any complicated
> jam code to switch between compilers depending on the toolset= option.
> The group I am with works routinely on systems with several compilers with
> diverse configuration (think large HPC machines-->1000 processors).  We
> routinely need to test our code as compiled by several different toolsets.
>    By adopting the changes included in these patches, Fortran compilation
> becomes much more seamless.  Perhaps the complicated module loading should
> be reserved for those that really want to use Fortran compiler X with
> toolset=Y. If I remember right, the Portland Group Fortran compiler IS a
> part of the PGI toolset.
>
> Intel-Linux fix:
> 1.  Improves the usage of the Intel toolset (on Linux at least) by honoring
> USER_OPTIONS for compiling and linking.
> 2.  Adds an <optimization>speed option
>
>
> I hope that this can be incorporated into the distribution--it will
> certainly help my efforts to use Boost.Build on the language-diverse
> numerical software that I work on.
>
> Spencer Olson