Flang Development

[Ben Brewer]

Hey John,
Craig <craig.g...@codethink.co.uk> and I have resumed work on
flang as part of our work for Codethink and have made some initial
patches, we hope to push more significant improvements over the coming
months.

We would make a pull request to llvm-flang but it turns out that
repository doesn't contain the full git history. For our repository we
re-based your changes over the top of the existing history so that it's
not lost and our patches are applied on top of that. We think
preserving the repository history is important as it helps us and future
potential contributors to understand the reason behind changes to the
code base.

We propose that you could pull our master and force push it over your
own, given it contains all your commits and commit messages as well as
the previous history.

Regards,
Ben Brewer

[Hal Finkel]

Hi Ben,

I see that there has been some recent rebasing work in https://github.com/llvm-flang/flang -- is all of this redundant with, or otherwise represented in https://github.com/CodethinkLabs/flang? If so, I recommend that we move forward with reinitializing llvm-flang against your repository.

Thanks again,
Hal

> --
> You received this message because you are subscribed to the Google
> Groups "flang-dev" group.
> To unsubscribe from this group and stop receiving emails from it,
> send an email to flang-dev+...@googlegroups.com.
> To post to this group, send email to flan...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/flang-dev/55D5E3DE.3040300%40codethink.co.uk.
> For more options, visit https://groups.google.com/d/optout.
>

--
Hal Finkel
Assistant Computational Scientist
Leadership Computing Facility
Argonne National Laboratory

[Ben Brewer]

Hey Hal,
Our rebasing work was to add the history back into the repository,
we also added a few fixes. It'd be wise to either apply the new patches
on top of our repository and move from there, or you rebase the official
repository in the same manner, either way it's pretty important to keep
the history.

Regards,
Ben

[Hal Finkel]

----- Original Message -----
> From: "Ben Brewer" <ben.b...@codethink.co.uk>
> To: "Hal Finkel" <hfi...@anl.gov>
> Cc: flan...@googlegroups.com, "Craig Griffiths" <craig.g...@codethink.co.uk>, "John Leidel"
> <john....@ttu.edu>
> Sent: Monday, October 26, 2015 5:17:50 AM
> Subject: Re: [flang-dev] Flang Development
>
> Hey Hal,
> Our rebasing work was to add the history back into the
> repository,
> we also added a few fixes. It'd be wise to either apply the new
> patches
> on top of our repository and move from there

Can you import the changesets and apply them to your repository? You might be in a better position to do that, especially since you've been actively working with the code in your repository.

>, or you rebase the
> official
> repository in the same manner, either way it's pretty important to
> keep
> the history.

I agree; keeping the history, when possible, is something we should do.

John, how would you like to move forward here?

Thanks again,
Hal

> https://groups.google.com/d/msgid/flang-dev/562DFDCE.20109%40codethink.co.uk.

> For more options, visit https://groups.google.com/d/optout.
>

[John Leidel]

What's the easiest (read, least destructive) way to accomplish this? 

To view this discussion on the web visit https://groups.google.com/d/msgid/flang-dev/21280980.121.1445876265425.JavaMail.javamailuser%40localhost.

[Craig Griffiths]

Hi all,

I have cherry picked the patches from Paul Osmialowski to the Codethink
labs repository, which should be up to date now:
https://github.com/CodethinkLabs/flang.git

John Leidel: If you want to use the tree with the full history you will
have to overwrite your current repository with ours.

Thanks,
Craig

[Hal Finkel]

----- Original Message -----
> From: "Craig Griffiths" <craig.g...@codethink.co.uk>
> To: "John Leidel" <john....@gmail.com>, "Hal Finkel" <hfi...@anl.gov>
> Cc: "Ben Brewer" <ben.b...@codethink.co.uk>, flan...@googlegroups.com, "John Leidel" <john....@ttu.edu>
> Sent: Tuesday, October 27, 2015 5:20:58 AM
> Subject: Re: [flang-dev] Flang Development
>

> Hi all,
>
> I have cherry picked the patches from Paul Osmialowski to the
> Codethink
> labs repository, which should be up to date now:
> https://github.com/CodethinkLabs/flang.git

Great, thanks!

>
> John Leidel: If you want to use the tree with the full history you
> will
> have to overwrite your current repository with ours.

FWIW, it might be easiest just to rename llvm-flang/flang to llvm-flang/flang-old and then fork CodethinkLabs/flang to be the new llvm-flang/flang.

-Hal

[Hal Finkel]

John, have you been able to make any progress on this?

-Hal

> --
> You received this message because you are subscribed to the Google
> Groups "flang-dev" group.
> To unsubscribe from this group and stop receiving emails from it,
> send an email to flang-dev+...@googlegroups.com.
> To post to this group, send email to flan...@googlegroups.com.
> To view this discussion on the web visit

> https://groups.google.com/d/msgid/flang-dev/22467609.224.1445963547289.JavaMail.javamailuser%40localhost.

> For more options, visit https://groups.google.com/d/optout.
>

[Ben Brewer]

We are pleased to announce the Open Fortran Compiler [1], a new Fortran front-end capable of parsing and performing semantic analysis on Fortran.  It currently parses all NIST F77 tests, and can re-print the parse trees using fixed & free-form compatible syntax. We can also print semantic errors and warnings, and translate some semantic trees to F90.

This is a result of writing tools to help deal with modernising and porting an extensive legacy Fortran codebase, starting first with work in the flang project earlier this year.  Some months ago, we realised that flang is not suited to the compilation of a significant portion of the Fortran in our codebase.

There are a few reasons for this, but the main blocker is the fact that flang uses the clang lexical analyser which can only work efficiently on Fortran 90 code; prior to Fortran 90 spaces were considered syntactically irrelevant, causing many ambiguities. It seems that Fortran and C differ too much for significant re-use of the clang project to be practical.

The initial focus has been on providing support for static code analysis of legacy FORTRAN 77 and earlier versions.  The immediate goal is to translate more older Fortran dialects into a Fortran 90 compatible form, so that older code bases can be used with more modern tooling.

A sensible future goal is pairing OFC with an llvm backend to allow direct compilation of older Fortran dialects to newer platforms.  OFC intends to support many code generators over time, and the project name was chosen with that end in mind.

Fortran code suitable for testing ofc has been pushed to a separate repository [2].  We would welcome any additional tests (especially for older dialects of Fortran), or assistance in integrating this with a back-end for the production of executable code.

If you’d like to discuss this in real-time please feel free to join us at #open-fortran-compiler on irc.freenode.net

[1] https://github.com/CodethinkLabs/ofc
[2] https://github.com/CodethinkLabs/ofc-tests

Regards,
Ben Brewer (aka flatmush)

[Ben Brewer]

P.S: Apologies for accidentally sending in-reply-to previously.

[John Leidel]

Hal, not yet.  Been busy with a number of other things surrounding SC.  I'll try to get this knocked out today.

[John Leidel]

This is done and issues have been migrated to the new repo.