[Flang] Add a Fortran Standards Support doc #132195

kiranchandramohan · 2025-03-20T12:08:50Z

No description provided.

mjklemm

LGTM, with the few nits that I have commented on.

I guess we should see if we can get a list of updates from J3/WG5 for 2018 and 2023 Fortran to populate the list. We also should list intrinsic functions explicitly if there are supported (or not).

mjklemm · 2025-03-20T14:32:02Z

flang/docs/FortranStandardsSupport.md

+Standards support is provided upto Fortran 2008 for now. It will be later extended for Fortran 2018 and Fortran 2023.
+
+The standards support information is provided as a table with three columns that are self explanatory. The Status column uses
+the letters **P**, **Y**, **N** for the various implementation status.


Suggested change

the letters **P**, **Y**, **N** for the various implementation status.

the letters **P**, **Y**, **N** for the implementation status:

mjklemm · 2025-03-20T14:33:26Z

flang/docs/FortranStandardsSupport.md

+
+-->
+
+# Flang's Fortran standards support


Suggested change

# Flang's Fortran standards support

# Flang Fortran Standards Support

tarunprabhu

Thanks Kiran. Just a few nits and comments.

tarunprabhu · 2025-03-20T15:00:40Z

flang/docs/FortranStandardsSupport.md

+This document summarizes Flang's Fortran standards support. The information is only provided as a guideline. The compiler emits
+TODOs/Not Yet Implemented messages for unimplemented features and that should be treated as the authoratative information. 


Suggested change

This document summarizes Flang's Fortran standards support. The information is only provided as a guideline. The compiler emits

TODOs/Not Yet Implemented messages for unimplemented features and that should be treated as the authoratative information.

This document summarizes Flang's Fortran standards support. The information is only provided as a guideline. The

TODOs/Not Yet Implemented messages emitted by the compiler for unimplemented features should be treated as authoritative.

tarunprabhu · 2025-03-20T15:01:18Z

flang/docs/FortranStandardsSupport.md

+
+This document summarizes Flang's Fortran standards support. The information is only provided as a guideline. The compiler emits
+TODOs/Not Yet Implemented messages for unimplemented features and that should be treated as the authoratative information. 
+Standards support is provided upto Fortran 2008 for now. It will be later extended for Fortran 2018 and Fortran 2023.


Suggested change

Standards support is provided upto Fortran 2008 for now. It will be later extended for Fortran 2018 and Fortran 2023.

Standards support is provided upto Fortran 2008 for now. It will be extended later for Fortran 2018 and Fortran 2023.

"upto" should be "up to"

I have removed this sentence.

tarunprabhu · 2025-03-20T15:05:46Z

flang/docs/FortranStandardsSupport.md

+- **N** : When the implementation is absent
+
+Note 1 : No distinction is made between the support in the Parser/Semantics and MLIR or Lowering support.
+Note 2 : Besides the features listed below a few intrinsics like MIN/MAX are not supported for a few cases with CHARACTER type.


Do we have a list of the intrinsics where support for CHARACTER types is incomplete? If it is not too difficult to identify these, a complete list may be better than alluding to "a few intrinsics".

I do not have a list of intrinsics here. The only ones I am actually aware or the MIN/MAX intrinsic. Since I was not 100% I am making a conservative statement here. Do we want to just say MIN/MAX only? Do you have additional info in the gfortran testsuite?

I can check to see if there are any failures as a result of this. I don't know off the top of my head.

kkwli · 2025-03-20T15:00:23Z

flang/docs/FortranStandardsSupport.md

+---
+```
+
+This document summarizes Flang's Fortran standards support. The information is only provided as a guideline. The compiler emits


If we take the suggested change above, should we also remove 's here? Or just rephrase it to:

This document summarizes Fortran standards support in Flang.

kkwli · 2025-03-20T15:04:37Z

flang/docs/FortranStandardsSupport.md

+Note 1 : No distinction is made between the support in the Parser/Semantics and MLIR or Lowering support.
+Note 2 : Besides the features listed below a few intrinsics like MIN/MAX are not supported for a few cases with CHARACTER type.
+
+## Fortran 77


FORTRAN 77 - consistent with what we use in FortranFeatureHistory.md

kkwli · 2025-03-20T15:13:00Z

flang/docs/FortranStandardsSupport.md

+
+Note 1 : No distinction is made between the support in the Parser/Semantics and MLIR or Lowering support.
+Note 2 : Besides the features listed below a few intrinsics like MIN/MAX are not supported for a few cases with CHARACTER type.
+


Since all the contents in the first three sections are the same, I suggest to consolidate to the FORTRAN77, Fortran 90 and Fortran 95 section. In addition, if we take Michael's suggestion to populate the Fortran 2018 and Fortran 2023 sections, I suggest to re-order the sections to start with Fortran 2023 first. I think people are more interested in which newer features that flang support nowadays.

I have reorganized in descending order of standards. I do not have entries for 2018 and 2023 now. If it is OK, we can add that in subsequent PRs.
I have not consolidated FORTRAN77, Fortran 90 and Fortran 95. I can do so before submitting if there are no reports of unsupported features in this review.

DanielCChen

LGTM. A couple of nits added inline.

DanielCChen · 2025-03-20T15:39:48Z

flang/docs/FortranStandardsSupport.md

+
+| Feature                                                    | Status | Comments                                                |
+|------------------------------------------------------------|--------|---------------------------------------------------------|
+| Parameterized Derived Types                                | P      | PDT with length type is not supported. See [Proposal](ParameterizedDerivedTypes.md) |


length type => length type parameters

DanielCChen · 2025-03-20T15:41:15Z

flang/docs/FortranStandardsSupport.md

+| Parameterized Derived Types                                | P      | PDT with length type is not supported. See [Proposal](ParameterizedDerivedTypes.md) |
+| Assignment to allocatable                                  | P      | Assignment to whole allocatable in FORALL is not implemented       |
+| Pointer Assignment                                         | P      | Polymorphic assignment in FORALL is not implemented     |
+| The VOLATILE attribute                                     | P      | Volatile in procedure interfaces is not implemented     |


Volatile => VOLATILE
as it refers to the attribute.

ceseo · 2025-03-20T19:16:56Z

This looks very good! Do you plan to do the same for Flang+OpenMP as well?

tblah

Thanks for putting this together Kiran

flang/docs/FortranStandardsSupport.md

kiranchandramohan

Addressed review comments.

kiranchandramohan · 2025-03-21T12:48:45Z

flang/docs/FortranStandardsSupport.md

+
+-->
+
+# Flang's Fortran standards support


kiranchandramohan · 2025-03-21T12:49:04Z

flang/docs/FortranStandardsSupport.md

+---
+```
+
+This document summarizes Flang's Fortran standards support. The information is only provided as a guideline. The compiler emits


kiranchandramohan · 2025-03-21T12:49:14Z

flang/docs/FortranStandardsSupport.md

+This document summarizes Flang's Fortran standards support. The information is only provided as a guideline. The compiler emits
+TODOs/Not Yet Implemented messages for unimplemented features and that should be treated as the authoratative information. 


kiranchandramohan · 2025-03-21T12:49:24Z

flang/docs/FortranStandardsSupport.md

+
+This document summarizes Flang's Fortran standards support. The information is only provided as a guideline. The compiler emits
+TODOs/Not Yet Implemented messages for unimplemented features and that should be treated as the authoratative information. 
+Standards support is provided upto Fortran 2008 for now. It will be later extended for Fortran 2018 and Fortran 2023.


kiranchandramohan · 2025-03-21T12:49:37Z

flang/docs/FortranStandardsSupport.md

+Standards support is provided upto Fortran 2008 for now. It will be later extended for Fortran 2018 and Fortran 2023.
+
+The standards support information is provided as a table with three columns that are self explanatory. The Status column uses
+the letters **P**, **Y**, **N** for the various implementation status.


kiranchandramohan · 2025-03-21T12:50:56Z

flang/docs/FortranStandardsSupport.md

+- **N** : When the implementation is absent
+
+Note 1 : No distinction is made between the support in the Parser/Semantics and MLIR or Lowering support.
+Note 2 : Besides the features listed below a few intrinsics like MIN/MAX are not supported for a few cases with CHARACTER type.


I do not have a list of intrinsics here. The only ones I am actually aware or the MIN/MAX intrinsic. Since I was not 100% I am making a conservative statement here. Do we want to just say MIN/MAX only? Do you have additional info in the gfortran testsuite?

kiranchandramohan · 2025-03-21T12:51:13Z

flang/docs/FortranStandardsSupport.md

+Note 1 : No distinction is made between the support in the Parser/Semantics and MLIR or Lowering support.
+Note 2 : Besides the features listed below a few intrinsics like MIN/MAX are not supported for a few cases with CHARACTER type.
+
+## Fortran 77


kiranchandramohan · 2025-03-21T12:51:22Z

flang/docs/FortranStandardsSupport.md

+
+| Feature                                                    | Status | Comments                                                |
+|------------------------------------------------------------|--------|---------------------------------------------------------|
+| Parameterized Derived Types                                | P      | PDT with length type is not supported. See [Proposal](ParameterizedDerivedTypes.md) |


flang/docs/FortranStandardsSupport.md

kiranchandramohan · 2025-03-21T12:51:39Z

flang/docs/FortranStandardsSupport.md

+| Parameterized Derived Types                                | P      | PDT with length type is not supported. See [Proposal](ParameterizedDerivedTypes.md) |
+| Assignment to allocatable                                  | P      | Assignment to whole allocatable in FORALL is not implemented       |
+| Pointer Assignment                                         | P      | Polymorphic assignment in FORALL is not implemented     |
+| The VOLATILE attribute                                     | P      | Volatile in procedure interfaces is not implemented     |


kiranchandramohan · 2025-03-21T12:56:27Z

I guess we should see if we can get a list of updates from J3/WG5 for 2018 and 2023 Fortran to populate the list.

I am leaving this for a future patch. 2023 should be fairly straightforward. I do not have the data for Fortran 2018.

We also should list intrinsic functions explicitly if there are supported (or not).

Agree, ATM I do not have this data.

This looks very good! Do you plan to do the same for Flang+OpenMP as well?

Yes.

klausler · 2025-03-21T16:02:07Z

I guess we should see if we can get a list of updates from J3/WG5 for 2018 and 2023 Fortran to populate the list.

I am leaving this for a future patch. 2023 should be fairly straightforward. I do not have the data for Fortran 2018.

f18 has all of F'2018 except coshape intrinsic lowering.

F'2023 support is completely absent and nobody is working on it.

kiranchandramohan · 2025-03-21T17:11:08Z

Thanks @klausler for the info. I will add sections for Fortran 2018 and 2023.

I guess we should see if we can get a list of updates from J3/WG5 for 2018 and 2023 Fortran to populate the list.

I am leaving this for a future patch. 2023 should be fairly straightforward. I do not have the data for Fortran 2018.

f18 has all of F'2018 except coshape intrinsic lowering.

Is there a specific reason you called out the coshape intrinsic? I see the other coarray extensions in Fortran 2018 hitting TODOs in lowering.

F'2023 support is completely absent and nobody is working on it.

A few of them seem to have support
-> Allow much longer statement lines and overall statement length
-> selected_logical_kind
-> do-concurrent reduction

klausler · 2025-03-21T17:27:43Z

Thanks @klausler for the info. I will add sections for Fortran 2018 and 2023.

I guess we should see if we can get a list of updates from J3/WG5 for 2018 and 2023 Fortran to populate the list.

I am leaving this for a future patch. 2023 should be fairly straightforward. I do not have the data for Fortran 2018.

f18 has all of F'2018 except coshape intrinsic lowering.

Is there a specific reason you called out the coshape intrinsic? I see the other coarray extensions in Fortran 2018 hitting TODOs in lowering.

I skimmed my notes of F'2018 additions and coshape stood out as the one item that I know isn't fully supported. If other coarray features from F'2018 are incomplete, please note them as well. Some were added as part of a TS between F'2008 and F'2018 so that makes them a little harder to find.

F'2023 support is completely absent and nobody is working on it.

A few of them seem to have support -> Allow much longer statement lines and overall statement length -> selected_logical_kind -> do-concurrent reduction

Fair enough. The lack of limits on continuation lines and statement length have always been present, so I never thought of them as new features. But F'2023 was a minor revision, and includes some weird features (e.g. @ subscripts) and a breaking change that don't appear in any compiler or (obviously) application yet, and that probably won't get any attention from me -- although I did add a portability warning for the breaking change. If you add a section for F'2023, it's going to be largely unpopulated.

kkwli

Thanks.

rouson · 2025-03-24T12:04:49Z

flang/docs/FortranStandardsSupport.md

+
+## Fortran 2018
+All features except those listed in the following table are supported. Almost all of the unsupported features are related to
+coarrays.


Suggested change

coarrays.

multi-image execution.

Many multi-image parallel features have do not necessarily involve coarrays (e.g., collective subroutines and teams).

Similarly to how the parallel support for do concurrent is described below as in progress, please state the support for multi-image features is in progress.

I now have the following sentence:

Almost all of the unsupported features are related to the additional parallel features (coarrays/teams/collectives).

The support status of collectives and teams are individually called out in the status table for Fortran 2018.

@kiranchandramohan thanks for making the edits. "Parallel features" could be interpreted to include parallelization of do concurrent so I'd still suggest "multi-image" to replace "parallel."

Done. Made the suggested change.

rouson · 2025-03-24T12:11:05Z

I guess we should see if we can get a list of updates from J3/WG5 for 2018 and 2023 Fortran to populate the list. We also should list intrinsic functions explicitly if there are supported (or not).

@mjklemm the list of changes is in the Introduction to the Fortran 2023 Interpretation Document.

rouson · 2025-03-24T12:21:22Z

flang/docs/FortranStandardsSupport.md

+
+## Fortran 2018
+All features except those listed in the following table are supported. Almost all of the unsupported features are related to
+coarrays.


Similarly to how the parallel support for do concurrent is described below as in progress, please state the support for multi-image features is in progress.

rouson · 2025-03-24T13:21:51Z

flang/docs/FortranStandardsSupport.md

+| Using integer arrays to specify subscripts                 | N      | |
+| Using integer arrays to specify rank and bound of an array | N      | |
+| Using an integer constant to specify rank                  | N      | |
+| Reduction specifier for do concurrent                      | Y      | |


I think this is still in progress. @ergawy can confirm. Should the status be N or P?

This is supported for the pure Fortran case, for the do concurrent -> OpenMP case (or do concurrent parallelization in general), this is not supported yet. So I think @kiranchandramohan is indeed correct here.

As @ergawy mentions, this is supported for the sequential Fortran case.
But I understand @rouson 's point that this might cause a user to think that reduction works in a parallel sense. So I am changing it to P (partial).

@kiranchandramohan actually I hadn't thought about the difference between serial support and parallel support. My point was that I didn't even realize that serial support for reduce existed yet on the main branch of llvm-project. I'm building flang now and will try it out. If serial support works, then I think either "Y" or "P" is fine and I leave it up to you to decide.

mleair

LGTM. Looks like you used the following document for the Fortran 2023 features:

https://wg5-fortran.org/N2201-N2250/N2212.pdf

You may also want to pull in the features from the Fortran 2008/2018 Compiler
Conformance charts that Ian Chivers and Jane Sleightholme were compiling. You can find them at https://www.fortranplus.co.uk/fortran-information/

kiranchandramohan · 2025-04-19T10:36:00Z

LGTM. Looks like you used the following document for the Fortran 2023 features:

https://wg5-fortran.org/N2201-N2250/N2212.pdf

You may also want to pull in the features from the Fortran 2008/2018 Compiler Conformance charts that Ian Chivers and Jane Sleightholme were compiling. You can find them at https://www.fortranplus.co.uk/fortran-information/

Thanks, Mark. I have added references to the What is new in Fortran 2023 and 2018 Fortran documents. I believe the Compiler Conformance charts are based on these documents.

I plan to submit this PR on Tuesday if there are no objections.

sscalpone · 2025-04-19T18:08:20Z

Hi Kiran, apologies for late comments. I hope they are relatively minor additions if you choose to incorporate them.

For the missing 2018 features, in the comment column, it would help to add "coarrays" or similar to highlight that the major missing feature is coarrays and a lot of gaps.

It might be nice to call out at the top of the doc that the major missing features are coarrays and PDT with length type parameters. If you agree these are the only major missing features.

Thank you!

kiranchandramohan · 2025-04-20T16:46:11Z

Thanks Steve for the review.

For the missing 2018 features, in the comment column, it would help to add "coarrays" or similar to highlight that the major missing feature is coarrays and a lot of gaps.

Added "Multi-image/Coarray feature" in the comment column.

It might be nice to call out at the top of the doc that the major missing features are coarrays and PDT with length type parameters. If you agree these are the only major missing features.

Added a Note in the first section.

Note: The two major missing features in Flang at present are coarrays and parameterized derived types (PDTs) with length type parameters.

Changes are in 04fb498

h-vetinari · 2025-04-20T22:36:23Z

flang/docs/FortranStandardsSupport.md

+- **Y** : When the implementation is complete
+- **N** : When the implementation is absent
+
+There is no distinction made between support in the Parser/Semantics and the MLIR or Lowering stages.


This sentence is now repeated (just above the P/Y/N there's an equivalent one).

Thanks. Removed the former occurence.

h-vetinari · 2025-04-20T22:38:59Z

flang/docs/FortranStandardsSupport.md

+the letters **P**, **Y**, **N** for the implementation status:
+- **P** : When the implementation is incomplete for a few cases
+- **Y** : When the implementation is complete
+- **N** : When the implementation is absent


I still think we should make this easier for readers (explain the letters, order them by degree of completion)

Suggested change

the letters **P**, **Y**, **N** for the implementation status:

- **P** : When the implementation is incomplete for a few cases

- **Y** : When the implementation is complete

- **N** : When the implementation is absent

the letters **Y**, **P**, **N** for the implementation status:

- **Y** (Yes): When the implementation is complete

- **P** (Partial): When the implementation is incomplete

- **N** (No): When the implementation is absent

Done. The following format was used.

- **Y** : Yes. When the implementation is complete

llvmbot added the flang Flang issues not falling into any other category label Mar 20, 2025

kiranchandramohan force-pushed the fortran-support branch from 3eb1775 to c03e566 Compare March 20, 2025 12:13

kiranchandramohan requested review from sscalpone, tblah, mjklemm, DanielCChen, kkwli, jeanPerier, rouson, pawosm-arm, klausler and kiranktp March 20, 2025 12:18

[Flang] Add a Fortran Standards Support doc

a0460e7

kiranchandramohan force-pushed the fortran-support branch from c03e566 to a0460e7 Compare March 20, 2025 12:20

mjklemm approved these changes Mar 20, 2025

View reviewed changes

tarunprabhu reviewed Mar 20, 2025

View reviewed changes

kkwli reviewed Mar 20, 2025

View reviewed changes

DanielCChen reviewed Mar 20, 2025

View reviewed changes

klausler removed their request for review March 20, 2025 16:09

tblah reviewed Mar 21, 2025

View reviewed changes

flang/docs/FortranStandardsSupport.md Outdated Show resolved Hide resolved

Address review comments

8499992

kiranchandramohan commented Mar 21, 2025

View reviewed changes

kiranchandramohan added 2 commits March 23, 2025 02:39

Add support tables for Fortran 2023 and Fortran 2018 standards

6c4b380

Move MIN MAX extension to Fortran 2003 table

84165a7

kkwli approved these changes Mar 23, 2025

View reviewed changes

Remove line about extending support for Fortran 2018 and Fortran 2023

8f9db84

rouson reviewed Mar 24, 2025

View reviewed changes

rouson suggested changes Mar 24, 2025

View reviewed changes

kiranchandramohan added 2 commits March 24, 2025 15:53

Address review comments

8eb16e4

Address review comment about multi-image execution

6781534

kiranchandramohan requested review from mleair and raghavendhra March 27, 2025 16:32

mleair reviewed Mar 27, 2025

View reviewed changes

tblah mentioned this pull request Apr 3, 2025

[ATfL] Add package documentation arm/arm-toolchain#248

Merged

kiranchandramohan mentioned this pull request Apr 7, 2025

[Flang] Fortran / OpenMP conformance page #62310

Closed

Partially address comments from Mark

66199ad

sscalpone requested a review from klausler April 19, 2025 17:59

Address Steve's comments

04fb498

kiranchandramohan force-pushed the fortran-support branch from 22bd5c0 to 04fb498 Compare April 20, 2025 16:44

h-vetinari reviewed Apr 20, 2025

View reviewed changes

kiranchandramohan added 2 commits April 21, 2025 08:58

Remove duplicate sentence

27260b6

Address comment from h.vetinari

c36aebc

kiranchandramohan merged commit 45a3056 into llvm:main Apr 24, 2025
12 checks passed

IanWood1 pushed a commit to IanWood1/llvm-project that referenced this pull request May 6, 2025

[Flang] Add a Fortran Standards Support doc (llvm#132195)

8743426

IanWood1 pushed a commit to IanWood1/llvm-project that referenced this pull request May 6, 2025

[Flang] Add a Fortran Standards Support doc (llvm#132195)

61bb0e9

IanWood1 pushed a commit to IanWood1/llvm-project that referenced this pull request May 6, 2025

[Flang] Add a Fortran Standards Support doc (llvm#132195)

cf00ded

Ankur-0429 pushed a commit to Ankur-0429/llvm-project that referenced this pull request May 9, 2025

[Flang] Add a Fortran Standards Support doc (llvm#132195)

6a87588

	the letters P, Y, N for the various implementation status.
	the letters P, Y, N for the implementation status:

	# Flang's Fortran standards support
	# Flang Fortran Standards Support

		This document summarizes Flang's Fortran standards support. The information is only provided as a guideline. The compiler emits
		TODOs/Not Yet Implemented messages for unimplemented features and that should be treated as the authoratative information.

	Standards support is provided upto Fortran 2008 for now. It will be later extended for Fortran 2018 and Fortran 2023.
	Standards support is provided upto Fortran 2008 for now. It will be extended later for Fortran 2018 and Fortran 2023.


		Note 1 : No distinction is made between the support in the Parser/Semantics and MLIR or Lowering support.
		Note 2 : Besides the features listed below a few intrinsics like MIN/MAX are not supported for a few cases with CHARACTER type.

[Flang] Add a Fortran Standards Support doc #132195

[Flang] Add a Fortran Standards Support doc #132195

Uh oh!

Conversation

kiranchandramohan commented Mar 20, 2025

Uh oh!

mjklemm left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

tarunprabhu left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

DanielCChen left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ceseo commented Mar 20, 2025

Uh oh!

tblah left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

kiranchandramohan left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

mjklemm left a comment •

edited

Loading