Re: Source-to-Source compilers?

Simone Pellegrini <>
Mon, 19 Jan 2009 03:59:59 -0800 (PST)

          From comp.compilers

Related articles
Source-to-Source compilers? (Simone Pellegrini) (2009-01-16)
Re: Source-to-Source compilers? (Hans-Peter Diettrich) (2009-01-17)
Re: Source-to-Source compilers? (2009-01-16)
Re: Source-to-Source compilers? (Ricardo Nabinger Sanchez) (2009-01-18)
Re: Source-to-Source compilers? (Simone Pellegrini) (2009-01-19)
Re: Source-to-Source compilers? (Simone Pellegrini) (2009-01-19)
Re: Source-to-Source compilers? (Simone Pellegrini) (2009-01-19)
Re: Source-to-Source compilers? (serge guelton) (2009-01-20)
| List of all articles for this month |

From: Simone Pellegrini <>
Newsgroups: comp.compilers
Date: Mon, 19 Jan 2009 03:59:59 -0800 (PST)
Organization: Compilers Central
References: 09-01-039 09-01-040
Keywords: tools, Fortran
Posted-Date: 19 Jan 2009 07:19:16 EST

On Jan 17, 3:01 am, Hans-Peter Diettrich <> wrote:
> Simone Pellegrini schrieb:
> > I am working in the optimization of parallel applications for multi-
> > core architectures. We don't want to auto-parallelize code but take a
> > parallel application (mainly written in C/C++ or Fortran) and produce
> > the optimized version via application of transformations.
> I.e. you want parallelization at compile time, not at runtime?

Not really, I assume the program I want to optimize is already
parallelized (openmp statements or MPI communications),
but I want to increase the performance just by applying well known
optimization transformations (like collective operations
or MPI statement movement... sync to async... and so on)
> > Now most of these transformation should be done at source level, so
> > our need is for something able to do source-to-source compilation. As
> > now, the only tools that fulfill our requirements are 2.
> IMO the transformation should be integrated into the single compilation
> pass, based on the AST. An (intermediate) source file could be provided
> for debugging, but is not required for the generation of executable code.

Yes that's right... but considering that these codes will run on
several architectures, and people should want to use
some kind of custom... high tuned compiler for the architecture the
have... I am sure they will not be happy
if the code is compiled by the tool I am developing. :)

Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.