Re: Will tree-ssa be GCC 3.5?

[Scott Robert Ladd <coyote at coyotegulch dot com>]

Robert Dewar wrote:
> If indeed you are coming to the table with $$$ for the support of G95
> development, that is a different matter entirely. In that case you 
> should ask for quotations on this work, and then USNL can look over
> the proposals and choose one that is credible.

Steven Bosscher issued an excellent reply to this.

A bit more background... Toon Moene brought up a report by the U.S. 
National Laboratories on the gfortran mailing list; I, in turn, took the 
same questions to comp.lang.fortran, which has a somewhat broader 
audience. You can find the report in question here:

http://www.llnl.gov/asci/pathforward_trilab/OSSODA_PF_RFI_V9e.pdf

Toon and I attempted to correct what we saw as misperceptions in the 
report. Subsequently, the writers of said report contacted both of us 
with specific questions. From my reply to them:

    There is no shortage of qualified people who would work on
    improving g95/gfortran; the limitation is funding. In my
    opinion, gfortran has more resources available than does
    Open64; development on gfortran is active and progessive.
    Furthermore, gfortran will be part of the GNU Compiler
    Collection; given the ubiquity of GCC, this bodes well for
    the general distribution of gfortran.

    g95/gfortran will "happen"; the question is how quickly, and
    with which features. Automatic vectorization, automatic
    parallelization, improved optimization, and OpenMP support
    *could* be implemented in g95; the resources exist, but
    developers would need funding and support.

As you can see, I am quite cognizant of the funding issue. Other issues 
include the quality of code emitted by gfortran, its ability to support 
high performance computing, and the timeliness of its release. I am 
attempting to answer those questions by asking polite questions.

Is tree-ssa going to become mainline, and if so, when under the current 
circumstances? In other words, is it worth expending tax dollars on the 
GCC project, or should the National labs support or produce another project?

..Scott

--
Scott Robert Ladd
Coyote Gulch Productions (http://www.coyotegulch.com)
Software Invention for High-Performance Computing