A New Flang - Interactive Code Review and Design Discussion

Hi everyone!

The team developing Flang (https://github.com/flang-compiler/flang), an
LLVM frontend for Fortran, has been working diligently at improving the
existing code base, adding new features, etc. However, while this is
delivering a production-quality frontend, it is clear that the
incremental evolution of the existing C code base likely won't get us to
a "Clang for Fortran" for several reasons. What we want is a frontend
for Fortran that, like Clang, is suitable for both tooling and
traditional compilation, written in modern C++ making use of LLVM
support classes as appropriate, and otherwise takes advantage of the
collective frontend-development experience of the LLVM community. To
address this desire, as announced briefly at EuroLLVM, the NVIDIA team
has started working on a ground-up rewrite of Flang, and would like to
present their current progress to the community, explain how
peculiarities of Fortran have influenced the current experimental
design, and get feedback from the community early in the development
process.

As a result, we'd like to schedule a one-hour interactive webinar, where
NVIDIA can present their current progress and get feedback from the
community. To those of you who are contributors to any part of Clang,
your participation is specifically requested. I'm going to try to
schedule this in two parts, first by picking some candidate days, and
then picking a time from that set of candidate days. If you're
interested in joining, please fill out this Doodle poll:

https://doodle.com/poll/mxvu5abtevs5zsv7

It contains days from the weeks of June 25, July 2, and July 9. One week
from today, on June 6th, I'll plan to make a new poll with times based
on candidate days from this poll.

A link to the repository with the new code will be sent out before the
webinar. NVIDIA intends to record the meeting for the benefit of those
not able to participate in person.

Thanks again,
Hal

P.S. If you want to have a technical discussion about Flang itself,
please use the flang-dev mailing list
(http://lists.flang-compiler.org/mailman/listinfo), not this one. I'm
posting here only to reach Clang contributors. The goal is to create a
Fortran frontend that might eventually be accepted as a formal part of
the LLVM project.

Hi again,

Based on the previous poll, we'll hold the webinar either on June 25,
July 2, July 9, or July 11.

Please fill out the following poll so that we can pick an exact time for
the webinar (as soon as you can):

https://doodle.com/poll/4qstmakgw3fykuf4

Thanks,

Hal

Hi again,

Based on the previous poll, we'll hold the webinar either on June 25,
July 2, July 9, or July 11.

Please fill out the following poll so that we can pick an exact time for
the webinar (as soon as you can):

https://doodle.com/poll/4qstmakgw3fykuf4

Based on the current poll results, I'm recommending that we pick two
times for the webinar:

Monday, July 2. 12-1pm CDT / 10-11am PDT / 6-7pm BST.

Wednesday, July 11. 11-12pm CDT / 9-10am PDT / 5-6pm BST.

It looks like the first time is more convenient for those on the west
coast of the US and the second time is nearly the most-popular selection
(and perhaps slightly more convenient for those in Europe).

Thanks again,
Hal

Quick reminder: The first of these calls starts in 30 minutes.

If you missed the information on how to join, see:
http://lists.llvm.org/pipermail/cfe-dev/2018-June/058314.html

-Hal

If I join the call can I laugh and say I “told ya so” or is that insensitive and too soon?

Oh well. sucks I wasn’t actually able to help in developing a Fortran fe for clang. Gotta love corporate sabotage backed by government …

/me hides

If I join the call can I laugh and say I “told ya so” or is that insensitive and too soon?

Oh well. sucks I wasn’t actually able to help in developing a Fortran fe for clang. Gotta love corporate sabotage backed by government …

This is not appropriate for LLVM Project mailing lists. It is both off topic and inappropriate.

You have been warned about this kind of post before, please stop making these posts on LLVM mailing lists or we will have to moderate you.