This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Will tree-ssa be GCC 3.5?
- From: Scott Robert Ladd <coyote at coyotegulch dot com>
- To: Matt Fago <fago at caltech dot edu>
- Cc: gcc at gcc dot gnu dot org
- Date: Mon, 01 Dec 2003 12:38:10 -0500
- Subject: Re: Will tree-ssa be GCC 3.5?
- References: <3FCB7886.3080104@caltech.edu>
Matt Fago wrote:
It does seem that there are not enough informed supporters for GCC in
the labs, and I applaud you for your efforts. My own work as an ASCI
university collaborator has been hampered by the lack of a free F95
compiler, and I sincerely hope that you are able to convince ASCI to
support gfortran.
The question may well be settled, as I understand it. I don't know,
though, if the chosen solution involves gfortran, Open64, or another
project.
It would seem that the labs could best support gfortran by funding
several of their own programmers (or ASCI university collaborators) to
contribute to gfortran, rather than shoveling money at some third party.
That was my original assumption of their intent. My lowly goal was to
refute some of the views expressed in the ASCI paper, thus having those
Lab programmers work on gfortran to the benefit of us all.
Alas, I was not as mercenary in my thinking as some others; I looked
upon this as an opportunity to improve support for GCC as a whole. I
wasn't looking to how I could profit -- a serious failing in an American
capitalist, for certain. ;)
Given the ubiquitous success of GCC, it seems obvious (?) that only GCC
could hope to provide this result without essentially duplicating GCC in
its entirely. However, the statement regarding the GPL is worrisome and
out-of-place considering the stated desire for an open-source tool. What
good is an open-source front end if the back-ends are all proprietary?!
Given the proper development support one would hope that GCC could meet
the 90% performance requirement.
These are some of the issues I was addressing privately with my
correspondents at the Labs.
The proper question then seems to me to be:
Can the lab requirements be sufficiently reconciled
with the goals of gfortran?
Good point. They seem to have come to a conclusion, although I am not
privy to the specifics.
Thanks for the excellent post.
..Scott
--
Scott Robert Ladd
Coyote Gulch Productions (http://www.coyotegulch.com)
Software Invention for High-Performance Computing