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?
Ian Lance Taylor wrote:
It's correct that ``the gcc developers'' and National Labs will never
meet. That's because ``the gcc developers'' have no way to even
accept funding, much less propose a schedule.
Granted. The discussion has wandered far fom my original intent, which
was merely to find out if tree-ssa was going to become mainline. I did
not raise the issues of funding or contracts.
However, there are many companies and individual contractors who work
with gcc (including my employer, Wasabi Systems). They are in a
position to accept funding and propose a schedule, just like any other
commercial software developer. When writing a contract such as this,
it is normal to stipulate that the work must be contributed back to
the mainline of gcc development, and I've seen contracts which even
make some portion of the payment contingent upon the work being
accepted back.
I'm quite aware of this. The problem is, the Labs were not interested in
makign such a contract available for gfortran development, due to
concerns about its quality and timeline.
Here's the sequence of events:
1) Through a message by Toon Moene, I discoevred a report written by
representatives of several U.S. National Laboratories. The report cites
the need for a quality Fortran 95 that addresses the HPC environments
used by the labs. They concluded that they would need to create a NEW
project, and fund it, given their impressions of the state of Open
Source compilers.
2) I was contacted by the National labs, in a private e-mail, regarding
my publicly-expressed opinions of their report. They asked some specific
questions, some of which I answered and for some of which I sought answers.
3) I did not know if tree-ssa was a purely theoretical project or the
future of GCC mainline, so I asked.
Certain people have presented chipped shoulders about version numbers
and funding, neither of which are germaine to the question "Will
tree-ssa become mainline?"
I don't know how National Labs handles their funding, but a typical
way to get them to fund gcc would be to have them issue an RFP for the
work they want done, and make sure that companies which invest in gcc
learn about the RFP and make proposals. Another typical way would be
to simply provide contact information for the right people at National
Labs, or simply announce that we should go through you.
No contract or RFP has been issued; they have some very specific intents
and guidelines. Quite honestly, I'm not certain their faith in GCC is
sufficient for them to issue a GCC-focused RFP. I was trying to bolster
their confidence, and asked a few questions that have been extrapolated
quite annoyingly.
--
Scott Robert Ladd
Coyote Gulch Productions (http://www.coyotegulch.com)
Software Invention for High-Performance Computing